xref: /titanic_41/usr/src/uts/common/inet/udp/udp.c (revision d29b2c4438482eb00488be49a1f5d6835f455546)
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 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 const char udp_version[] = "%Z%%M%	%I%	%E% SMI";
30 
31 #include <sys/types.h>
32 #include <sys/stream.h>
33 #include <sys/dlpi.h>
34 #include <sys/pattr.h>
35 #include <sys/stropts.h>
36 #include <sys/strlog.h>
37 #include <sys/strsun.h>
38 #include <sys/time.h>
39 #define	_SUN_TPI_VERSION 2
40 #include <sys/tihdr.h>
41 #include <sys/timod.h>
42 #include <sys/ddi.h>
43 #include <sys/sunddi.h>
44 #include <sys/strsubr.h>
45 #include <sys/suntpi.h>
46 #include <sys/xti_inet.h>
47 #include <sys/cmn_err.h>
48 #include <sys/kmem.h>
49 #include <sys/policy.h>
50 #include <sys/ucred.h>
51 #include <sys/zone.h>
52 
53 #include <sys/socket.h>
54 #include <sys/sockio.h>
55 #include <sys/vtrace.h>
56 #include <sys/sdt.h>
57 #include <sys/debug.h>
58 #include <sys/isa_defs.h>
59 #include <sys/random.h>
60 #include <netinet/in.h>
61 #include <netinet/ip6.h>
62 #include <netinet/icmp6.h>
63 #include <netinet/udp.h>
64 #include <net/if.h>
65 #include <net/route.h>
66 
67 #include <inet/common.h>
68 #include <inet/ip.h>
69 #include <inet/ip_impl.h>
70 #include <inet/ip6.h>
71 #include <inet/ip_ire.h>
72 #include <inet/ip_if.h>
73 #include <inet/ip_multi.h>
74 #include <inet/ip_ndp.h>
75 #include <inet/mi.h>
76 #include <inet/mib2.h>
77 #include <inet/nd.h>
78 #include <inet/optcom.h>
79 #include <inet/snmpcom.h>
80 #include <inet/kstatcom.h>
81 #include <inet/udp_impl.h>
82 #include <inet/ipclassifier.h>
83 #include <inet/ipsec_impl.h>
84 #include <inet/ipp_common.h>
85 
86 /*
87  * The ipsec_info.h header file is here since it has the definition for the
88  * M_CTL message types used by IP to convey information to the ULP. The
89  * ipsec_info.h needs the pfkeyv2.h, hence the latter's presence.
90  */
91 #include <net/pfkeyv2.h>
92 #include <inet/ipsec_info.h>
93 
94 #include <sys/tsol/label.h>
95 #include <sys/tsol/tnet.h>
96 #include <rpc/pmap_prot.h>
97 
98 /*
99  * Synchronization notes:
100  *
101  * UDP uses a combination of its internal perimeter, a global lock and
102  * a set of bind hash locks to protect its data structures.  Please see
103  * the note above udp_mode_assertions for details about the internal
104  * perimeter.
105  *
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.  An UDP endpoint is removed from the bind hash list only
113  * when it is being unbound or being closed.  The per bucket lock also
114  * protects a UDP endpoint's state changes.
115  *
116  * Plumbing notes:
117  *
118  * Both udp and ip are merged, but the streams plumbing is kept unchanged
119  * in that udp is always pushed atop /dev/ip.  This is done to preserve
120  * backwards compatibility for certain applications which rely on such
121  * plumbing geometry to do things such as issuing I_POP on the stream
122  * in order to obtain direct access to /dev/ip, etc.
123  *
124  * All UDP processings happen in the /dev/ip instance; the udp module
125  * instance does not possess any state about the endpoint, and merely
126  * acts as a dummy module whose presence is to keep the streams plumbing
127  * appearance unchanged.  At open time /dev/ip allocates a conn_t that
128  * happens to embed a udp_t.  This stays dormant until the time udp is
129  * pushed, which indicates to /dev/ip that it must convert itself from
130  * an IP to a UDP endpoint.
131  *
132  * We only allow for the following plumbing cases:
133  *
134  * Normal:
135  *	/dev/ip is first opened and later udp is pushed directly on top.
136  *	This is the default action that happens when a udp socket or
137  *	/dev/udp is opened.  The conn_t created by /dev/ip instance is
138  *	now shared and is marked with IPCL_UDP.
139  *
140  * SNMP-only:
141  *	udp is pushed on top of a module other than /dev/ip.  When this
142  *	happens it will support only SNMP semantics.  A new conn_t is
143  *	allocated and marked with IPCL_UDPMOD.
144  *
145  * The above cases imply that we don't support any intermediate module to
146  * reside in between /dev/ip and udp -- in fact, we never supported such
147  * scenario in the past as the inter-layer communication semantics have
148  * always been private.  Also note that the normal case allows for SNMP
149  * requests to be processed in addition to the rest of UDP operations.
150  *
151  * The normal case plumbing is depicted by the following diagram:
152  *
153  *	+---------------+---------------+
154  *	|		|		| udp
155  *	|     udp_wq	|    udp_rq	|
156  *	|		|    UDP_RD	|
157  *	|		|		|
158  *	+---------------+---------------+
159  *		|		^
160  *		v		|
161  *	+---------------+---------------+
162  *	|		|		| /dev/ip
163  *	|     ip_wq	|     ip_rq	| conn_t
164  *	|     UDP_WR	|		|
165  *	|		|		|
166  *	+---------------+---------------+
167  *
168  * Messages arriving at udp_wq from above will end up in ip_wq before
169  * it gets processed, i.e. udp write entry points will advance udp_wq
170  * and use its q_next value as ip_wq in order to use the conn_t that
171  * is stored in its q_ptr.  Likewise, messages generated by ip to the
172  * module above udp will appear as if they are originated from udp_rq,
173  * i.e. putnext() calls to the module above udp is done using the
174  * udp_rq instead of ip_rq in order to avoid udp_rput() which does
175  * nothing more than calling putnext().
176  *
177  * The above implies the following rule of thumb:
178  *
179  *   1. udp_t is obtained from conn_t, which is created by the /dev/ip
180  *	instance and is stored in q_ptr of both ip_wq and ip_rq.  There
181  *	is no direct reference to conn_t from either udp_wq or udp_rq.
182  *
183  *   2. Write-side entry points of udp can obtain the conn_t via the
184  *	Q_TO_CONN() macro, using the queue value obtain from UDP_WR().
185  *
186  *   3. While in /dev/ip context, putnext() to the module above udp can
187  *	be done by supplying the queue value obtained from UDP_RD().
188  *
189  */
190 
191 static queue_t *UDP_WR(queue_t *);
192 static queue_t *UDP_RD(queue_t *);
193 
194 struct kmem_cache *udp_cache;
195 
196 /* For /etc/system control */
197 uint_t udp_bind_fanout_size = UDP_BIND_FANOUT_SIZE;
198 
199 #define	NDD_TOO_QUICK_MSG \
200 	"ndd get info rate too high for non-privileged users, try again " \
201 	"later.\n"
202 #define	NDD_OUT_OF_BUF_MSG	"<< Out of buffer >>\n"
203 
204 /* Option processing attrs */
205 typedef struct udpattrs_s {
206 	union {
207 		ip6_pkt_t	*udpattr_ipp6;	/* For V6 */
208 		ip4_pkt_t 	*udpattr_ipp4;	/* For V4 */
209 	} udpattr_ippu;
210 #define	udpattr_ipp6 udpattr_ippu.udpattr_ipp6
211 #define	udpattr_ipp4 udpattr_ippu.udpattr_ipp4
212 	mblk_t		*udpattr_mb;
213 	boolean_t	udpattr_credset;
214 } udpattrs_t;
215 
216 static void	udp_addr_req(queue_t *q, mblk_t *mp);
217 static void	udp_bind(queue_t *q, mblk_t *mp);
218 static void	udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp);
219 static void	udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock);
220 static int	udp_build_hdrs(queue_t *q, udp_t *udp);
221 static void	udp_capability_req(queue_t *q, mblk_t *mp);
222 static int	udp_close(queue_t *q);
223 static void	udp_connect(queue_t *q, mblk_t *mp);
224 static void	udp_disconnect(queue_t *q, mblk_t *mp);
225 static void	udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error,
226 		    int sys_error);
227 static void	udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive,
228 		    t_scalar_t tlierr, int unixerr);
229 static int	udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp,
230 		    cred_t *cr);
231 static int	udp_extra_priv_ports_add(queue_t *q, mblk_t *mp,
232 		    char *value, caddr_t cp, cred_t *cr);
233 static int	udp_extra_priv_ports_del(queue_t *q, mblk_t *mp,
234 		    char *value, caddr_t cp, cred_t *cr);
235 static void	udp_icmp_error(queue_t *q, mblk_t *mp);
236 static void	udp_icmp_error_ipv6(queue_t *q, mblk_t *mp);
237 static void	udp_info_req(queue_t *q, mblk_t *mp);
238 static mblk_t	*udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim,
239 		    t_scalar_t addr_length);
240 static int	udp_open(queue_t *q, dev_t *devp, int flag, int sflag,
241 		    cred_t *credp);
242 static  int	udp_unitdata_opt_process(queue_t *q, mblk_t *mp,
243 		    int *errorp, udpattrs_t *udpattrs);
244 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name);
245 static int	udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr);
246 static boolean_t udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt);
247 static int	udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
248 		    cred_t *cr);
249 static void	udp_report_item(mblk_t *mp, udp_t *udp);
250 static void	udp_rput(queue_t *q, mblk_t *mp);
251 static void	udp_rput_other(queue_t *, mblk_t *);
252 static int	udp_rinfop(queue_t *q, infod_t *dp);
253 static int	udp_rrw(queue_t *q, struiod_t *dp);
254 static	void	udp_rput_bind_ack(queue_t *q, mblk_t *mp);
255 static int	udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp,
256 		    cred_t *cr);
257 static void	udp_send_data(udp_t *, queue_t *, mblk_t *, ipha_t *);
258 static void	udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr,
259 		    t_scalar_t destlen, t_scalar_t err);
260 static void	udp_unbind(queue_t *q, mblk_t *mp);
261 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port,
262     boolean_t random);
263 static void	udp_wput(queue_t *q, mblk_t *mp);
264 static mblk_t	*udp_output_v4(conn_t *, mblk_t *, ipaddr_t, uint16_t, uint_t,
265 		    int *, boolean_t);
266 static mblk_t	*udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6,
267 		    int *error);
268 static void	udp_wput_other(queue_t *q, mblk_t *mp);
269 static void	udp_wput_iocdata(queue_t *q, mblk_t *mp);
270 static void	udp_output(conn_t *connp, mblk_t *mp, struct sockaddr *addr,
271 		    socklen_t addrlen);
272 static size_t	udp_set_rcv_hiwat(udp_t *udp, size_t size);
273 
274 static void	*udp_stack_init(netstackid_t stackid, netstack_t *ns);
275 static void	udp_stack_fini(netstackid_t stackid, void *arg);
276 
277 static void	*udp_kstat_init(netstackid_t stackid);
278 static void	udp_kstat_fini(netstackid_t stackid, kstat_t *ksp);
279 static void	*udp_kstat2_init(netstackid_t, udp_stat_t *);
280 static void	udp_kstat2_fini(netstackid_t, kstat_t *);
281 static int	udp_kstat_update(kstat_t *kp, int rw);
282 static void	udp_input_wrapper(void *arg, mblk_t *mp, void *arg2);
283 static void	udp_rput_other_wrapper(void *arg, mblk_t *mp, void *arg2);
284 static void	udp_wput_other_wrapper(void *arg, mblk_t *mp, void *arg2);
285 static void	udp_resume_bind_cb(void *arg, mblk_t *mp, void *arg2);
286 
287 static void	udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp,
288 		    uint_t pkt_len);
289 static void	udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing);
290 static void	udp_enter(conn_t *, mblk_t *, sqproc_t, uint8_t);
291 static void	udp_exit(conn_t *);
292 static void	udp_become_writer(conn_t *, mblk_t *, sqproc_t, uint8_t);
293 #ifdef DEBUG
294 static void	udp_mode_assertions(udp_t *, int);
295 #endif /* DEBUG */
296 
297 major_t UDP6_MAJ;
298 #define	UDP6 "udp6"
299 
300 #define	UDP_RECV_HIWATER	(56 * 1024)
301 #define	UDP_RECV_LOWATER	128
302 #define	UDP_XMIT_HIWATER	(56 * 1024)
303 #define	UDP_XMIT_LOWATER	1024
304 
305 static struct module_info udp_info =  {
306 	UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER
307 };
308 
309 static struct qinit udp_rinit = {
310 	(pfi_t)udp_rput, NULL, udp_open, udp_close, NULL,
311 	&udp_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD
312 };
313 
314 static struct qinit udp_winit = {
315 	(pfi_t)udp_wput, NULL, NULL, NULL, NULL,
316 	&udp_info, NULL, NULL, NULL, STRUIOT_NONE
317 };
318 
319 /* Support for just SNMP if UDP is not pushed directly over device IP */
320 struct qinit udp_snmp_rinit = {
321 	(pfi_t)putnext, NULL, udp_open, ip_snmpmod_close, NULL,
322 	&udp_info, NULL, NULL, NULL, STRUIOT_NONE
323 };
324 
325 struct qinit udp_snmp_winit = {
326 	(pfi_t)ip_snmpmod_wput, NULL, udp_open, ip_snmpmod_close, NULL,
327 	&udp_info, NULL, NULL, NULL, STRUIOT_NONE
328 };
329 
330 struct streamtab udpinfo = {
331 	&udp_rinit, &udp_winit
332 };
333 
334 static	sin_t	sin_null;	/* Zero address for quick clears */
335 static	sin6_t	sin6_null;	/* Zero address for quick clears */
336 
337 #define	UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH)
338 
339 /* Default structure copied into T_INFO_ACK messages */
340 static struct T_info_ack udp_g_t_info_ack_ipv4 = {
341 	T_INFO_ACK,
342 	UDP_MAXPACKET_IPV4,	/* TSDU_size. Excl. headers */
343 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
344 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
345 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
346 	sizeof (sin_t),	/* ADDR_size. */
347 	0,		/* OPT_size - not initialized here */
348 	UDP_MAXPACKET_IPV4,	/* TIDU_size.  Excl. headers */
349 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
350 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
351 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
352 };
353 
354 #define	UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN)
355 
356 static	struct T_info_ack udp_g_t_info_ack_ipv6 = {
357 	T_INFO_ACK,
358 	UDP_MAXPACKET_IPV6,	/* TSDU_size.  Excl. headers */
359 	T_INVALID,	/* ETSU_size.  udp does not support expedited data. */
360 	T_INVALID,	/* CDATA_size. udp does not support connect data. */
361 	T_INVALID,	/* DDATA_size. udp does not support disconnect data. */
362 	sizeof (sin6_t), /* ADDR_size. */
363 	0,		/* OPT_size - not initialized here */
364 	UDP_MAXPACKET_IPV6,	/* TIDU_size. Excl. headers */
365 	T_CLTS,		/* SERV_type.  udp supports connection-less. */
366 	TS_UNBND,	/* CURRENT_state.  This is set from udp_state. */
367 	(XPG4_1|SENDZERO) /* PROVIDER_flag */
368 };
369 
370 /* largest UDP port number */
371 #define	UDP_MAX_PORT	65535
372 
373 /*
374  * Table of ND variables supported by udp.  These are loaded into us_nd
375  * in udp_open.
376  * All of these are alterable, within the min/max values given, at run time.
377  */
378 /* BEGIN CSTYLED */
379 udpparam_t udp_param_arr[] = {
380  /*min		max		value		name */
381  { 0L,		256,		32,		"udp_wroff_extra" },
382  { 1L,		255,		255,		"udp_ipv4_ttl" },
383  { 0,		IPV6_MAX_HOPS,	IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"},
384  { 1024,	(32 * 1024),	1024,		"udp_smallest_nonpriv_port" },
385  { 0,		1,		1,		"udp_do_checksum" },
386  { 1024,	UDP_MAX_PORT,	(32 * 1024),	"udp_smallest_anon_port" },
387  { 1024,	UDP_MAX_PORT,	UDP_MAX_PORT,	"udp_largest_anon_port" },
388  { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER,	"udp_xmit_hiwat"},
389  { 0,		     (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"},
390  { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER,	"udp_recv_hiwat"},
391  { 65536,	(1<<30),	2*1024*1024,	"udp_max_buf"},
392  { 100,		60000,		1000,		"udp_ndd_get_info_interval"},
393 };
394 /* END CSTYLED */
395 
396 /* Setable in /etc/system */
397 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */
398 uint32_t udp_random_anon_port = 1;
399 
400 /*
401  * Hook functions to enable cluster networking.
402  * On non-clustered systems these vectors must always be NULL
403  */
404 
405 void (*cl_inet_bind)(uchar_t protocol, sa_family_t addr_family,
406     uint8_t *laddrp, in_port_t lport) = NULL;
407 void (*cl_inet_unbind)(uint8_t protocol, sa_family_t addr_family,
408     uint8_t *laddrp, in_port_t lport) = NULL;
409 
410 typedef union T_primitives *t_primp_t;
411 
412 #define	UDP_ENQUEUE_MP(udp, mp, proc, tag) {			\
413 	ASSERT((mp)->b_prev == NULL && (mp)->b_queue == NULL);	\
414 	ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock));	\
415 	(mp)->b_queue = (queue_t *)((uintptr_t)tag);		\
416 	(mp)->b_prev = (mblk_t *)proc;				\
417 	if ((udp)->udp_mphead == NULL)				\
418 		(udp)->udp_mphead = (mp);			\
419 	else							\
420 		(udp)->udp_mptail->b_next = (mp);		\
421 	(udp)->udp_mptail = (mp);				\
422 	(udp)->udp_mpcount++;					\
423 }
424 
425 #define	UDP_READERS_INCREF(udp) {				\
426 	ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock));	\
427 	(udp)->udp_reader_count++;				\
428 }
429 
430 #define	UDP_READERS_DECREF(udp) {				\
431 	ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock));	\
432 	(udp)->udp_reader_count--;				\
433 	if ((udp)->udp_reader_count == 0)			\
434 		cv_broadcast(&(udp)->udp_connp->conn_cv);	\
435 }
436 
437 #define	UDP_SQUEUE_DECREF(udp) {				\
438 	ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock));	\
439 	(udp)->udp_squeue_count--;				\
440 	if ((udp)->udp_squeue_count == 0)			\
441 		cv_broadcast(&(udp)->udp_connp->conn_cv);	\
442 }
443 
444 /*
445  * Notes on UDP endpoint synchronization:
446  *
447  * UDP needs exclusive operation on a per endpoint basis, when executing
448  * functions that modify the endpoint state.  udp_rput_other() deals with
449  * packets with IP options, and processing these packets end up having
450  * to update the endpoint's option related state.  udp_wput_other() deals
451  * with control operations from the top, e.g. connect() that needs to
452  * update the endpoint state.  These could be synchronized using locks,
453  * but the current version uses squeues for this purpose.  squeues may
454  * give performance improvement for certain cases such as connected UDP
455  * sockets; thus the framework allows for using squeues.
456  *
457  * The perimeter routines are described as follows:
458  *
459  * udp_enter():
460  *	Enter the UDP endpoint perimeter.
461  *
462  * udp_become_writer():
463  *	Become exclusive on the UDP endpoint.  Specifies a function
464  *	that will be called exclusively either immediately or later
465  *	when the perimeter is available exclusively.
466  *
467  * udp_exit():
468  *	Exit the UDP perimeter.
469  *
470  * Entering UDP from the top or from the bottom must be done using
471  * udp_enter().  No lock must be held while attempting to enter the UDP
472  * perimeter.  When finished, udp_exit() must be called to get out of
473  * the perimeter.
474  *
475  * UDP operates in either MT_HOT mode or in SQUEUE mode.  In MT_HOT mode,
476  * multiple threads may enter a UDP endpoint concurrently.  This is used
477  * for sending and/or receiving normal data.  Control operations and other
478  * special cases call udp_become_writer() to become exclusive on a per
479  * endpoint basis and this results in transitioning to SQUEUE mode.  squeue
480  * by definition serializes access to the conn_t.  When there are no more
481  * pending messages on the squeue for the UDP connection, the endpoint
482  * reverts to MT_HOT mode.  During the interregnum when not all MT threads
483  * of an endpoint have finished, messages are queued in the UDP endpoint
484  * and the UDP is in UDP_MT_QUEUED mode or UDP_QUEUED_SQUEUE mode.
485  *
486  * These modes have the following analogs:
487  *
488  *	UDP_MT_HOT/udp_reader_count==0		none
489  *	UDP_MT_HOT/udp_reader_count>0		RW_READ_LOCK
490  *	UDP_MT_QUEUED				RW_WRITE_WANTED
491  *	UDP_SQUEUE or UDP_QUEUED_SQUEUE		RW_WRITE_LOCKED
492  *
493  * Stable modes:	UDP_MT_HOT, UDP_SQUEUE
494  * Transient modes:	UDP_MT_QUEUED, UDP_QUEUED_SQUEUE
495  *
496  * While in stable modes, UDP keeps track of the number of threads
497  * operating on the endpoint.  The udp_reader_count variable represents
498  * the number of threads entering the endpoint as readers while it is
499  * in UDP_MT_HOT mode.  Transitioning to UDP_SQUEUE happens when there
500  * is only a single reader, i.e. when this counter drops to 1.  Likewise,
501  * udp_squeue_count represents the number of threads operating on the
502  * endpoint's squeue while it is in UDP_SQUEUE mode.  The mode transition
503  * to UDP_MT_HOT happens after the last thread exits the endpoint, i.e.
504  * when this counter drops to 0.
505  *
506  * The default mode is set to UDP_MT_HOT and UDP alternates between
507  * UDP_MT_HOT and UDP_SQUEUE as shown in the state transition below.
508  *
509  * Mode transition:
510  * ----------------------------------------------------------------
511  * old mode		Event				New mode
512  * ----------------------------------------------------------------
513  * UDP_MT_HOT		Call to udp_become_writer()	UDP_SQUEUE
514  *			and udp_reader_count == 1
515  *
516  * UDP_MT_HOT		Call to udp_become_writer()	UDP_MT_QUEUED
517  *			and udp_reader_count > 1
518  *
519  * UDP_MT_QUEUED	udp_reader_count drops to zero	UDP_QUEUED_SQUEUE
520  *
521  * UDP_QUEUED_SQUEUE	All messages enqueued on the	UDP_SQUEUE
522  *			internal UDP queue successfully
523  *			moved to squeue AND udp_squeue_count != 0
524  *
525  * UDP_QUEUED_SQUEUE	All messages enqueued on the	UDP_MT_HOT
526  *			internal UDP queue successfully
527  *			moved to squeue AND udp_squeue_count
528  *			drops to zero
529  *
530  * UDP_SQUEUE		udp_squeue_count drops to zero	UDP_MT_HOT
531  * ----------------------------------------------------------------
532  */
533 
534 static queue_t *
535 UDP_WR(queue_t *q)
536 {
537 	ASSERT(q->q_ptr == NULL && _OTHERQ(q)->q_ptr == NULL);
538 	ASSERT(WR(q)->q_next != NULL && WR(q)->q_next->q_ptr != NULL);
539 	ASSERT(IPCL_IS_UDP(Q_TO_CONN(WR(q)->q_next)));
540 
541 	return (_WR(q)->q_next);
542 }
543 
544 static queue_t *
545 UDP_RD(queue_t *q)
546 {
547 	ASSERT(q->q_ptr != NULL && _OTHERQ(q)->q_ptr != NULL);
548 	ASSERT(IPCL_IS_UDP(Q_TO_CONN(q)));
549 	ASSERT(RD(q)->q_next != NULL && RD(q)->q_next->q_ptr == NULL);
550 
551 	return (_RD(q)->q_next);
552 }
553 
554 #ifdef DEBUG
555 #define	UDP_MODE_ASSERTIONS(udp, caller) udp_mode_assertions(udp, caller)
556 #else
557 #define	UDP_MODE_ASSERTIONS(udp, caller)
558 #endif
559 
560 /* Invariants */
561 #ifdef DEBUG
562 
563 uint32_t udp_count[4];
564 
565 /* Context of udp_mode_assertions */
566 #define	UDP_ENTER		1
567 #define	UDP_BECOME_WRITER	2
568 #define	UDP_EXIT		3
569 
570 static void
571 udp_mode_assertions(udp_t *udp, int caller)
572 {
573 	ASSERT(MUTEX_HELD(&udp->udp_connp->conn_lock));
574 
575 	switch (udp->udp_mode) {
576 	case UDP_MT_HOT:
577 		/*
578 		 * Messages have not yet been enqueued on the internal queue,
579 		 * otherwise we would have switched to UDP_MT_QUEUED. Likewise
580 		 * by definition, there can't be any messages enqueued on the
581 		 * squeue. The UDP could be quiescent, so udp_reader_count
582 		 * could be zero at entry.
583 		 */
584 		ASSERT(udp->udp_mphead == NULL && udp->udp_mpcount == 0 &&
585 		    udp->udp_squeue_count == 0);
586 		ASSERT(caller == UDP_ENTER || udp->udp_reader_count != 0);
587 		udp_count[0]++;
588 		break;
589 
590 	case UDP_MT_QUEUED:
591 		/*
592 		 * The last MT thread to exit the udp perimeter empties the
593 		 * internal queue and then switches the UDP to
594 		 * UDP_QUEUED_SQUEUE mode. Since we are still in UDP_MT_QUEUED
595 		 * mode, it means there must be at least 1 MT thread still in
596 		 * the perimeter and at least 1 message on the internal queue.
597 		 */
598 		ASSERT(udp->udp_reader_count >= 1 && udp->udp_mphead != NULL &&
599 		    udp->udp_mpcount != 0 && udp->udp_squeue_count == 0);
600 		udp_count[1]++;
601 		break;
602 
603 	case UDP_QUEUED_SQUEUE:
604 		/*
605 		 * The switch has happened from MT to SQUEUE. So there can't
606 		 * any MT threads. Messages could still pile up on the internal
607 		 * queue until the transition is complete and we move to
608 		 * UDP_SQUEUE mode. We can't assert on nonzero udp_squeue_count
609 		 * since the squeue could drain any time.
610 		 */
611 		ASSERT(udp->udp_reader_count == 0);
612 		udp_count[2]++;
613 		break;
614 
615 	case UDP_SQUEUE:
616 		/*
617 		 * The transition is complete. Thre can't be any messages on
618 		 * the internal queue. The udp could be quiescent or the squeue
619 		 * could drain any time, so we can't assert on nonzero
620 		 * udp_squeue_count during entry. Nor can we assert that
621 		 * udp_reader_count is zero, since, a reader thread could have
622 		 * directly become writer in line by calling udp_become_writer
623 		 * without going through the queued states.
624 		 */
625 		ASSERT(udp->udp_mphead == NULL && udp->udp_mpcount == 0);
626 		ASSERT(caller == UDP_ENTER || udp->udp_squeue_count != 0);
627 		udp_count[3]++;
628 		break;
629 	}
630 }
631 #endif
632 
633 #define	_UDP_ENTER(connp, mp, proc, tag) {				\
634 	udp_t *_udp = (connp)->conn_udp;				\
635 									\
636 	mutex_enter(&(connp)->conn_lock);				\
637 	if ((connp)->conn_state_flags & CONN_CLOSING) {			\
638 		mutex_exit(&(connp)->conn_lock);			\
639 		freemsg(mp);						\
640 	} else {							\
641 		UDP_MODE_ASSERTIONS(_udp, UDP_ENTER);			\
642 									\
643 		switch (_udp->udp_mode) {				\
644 		case UDP_MT_HOT:					\
645 			/* We can execute as reader right away. */	\
646 			UDP_READERS_INCREF(_udp);			\
647 			mutex_exit(&(connp)->conn_lock);		\
648 			(*(proc))(connp, mp, (connp)->conn_sqp);	\
649 			break;						\
650 									\
651 		case UDP_SQUEUE:					\
652 			/*						\
653 			 * We are in squeue mode, send the		\
654 			 * packet to the squeue				\
655 			 */						\
656 			_udp->udp_squeue_count++;			\
657 			CONN_INC_REF_LOCKED(connp);			\
658 			mutex_exit(&(connp)->conn_lock);		\
659 			squeue_enter((connp)->conn_sqp, mp, proc,	\
660 			    connp, tag);				\
661 			break;						\
662 									\
663 		case UDP_MT_QUEUED:					\
664 		case UDP_QUEUED_SQUEUE:					\
665 			/*						\
666 			 * Some messages may have been enqueued		\
667 			 * ahead of us.  Enqueue the new message	\
668 			 * at the tail of the internal queue to		\
669 			 * preserve message ordering.			\
670 			 */						\
671 			UDP_ENQUEUE_MP(_udp, mp, proc, tag);		\
672 			mutex_exit(&(connp)->conn_lock);		\
673 			break;						\
674 		}							\
675 	}								\
676 }
677 
678 static void
679 udp_enter(conn_t *connp, mblk_t *mp, sqproc_t proc, uint8_t tag)
680 {
681 	_UDP_ENTER(connp, mp, proc, tag);
682 }
683 
684 static void
685 udp_become_writer(conn_t *connp, mblk_t *mp, sqproc_t proc, uint8_t tag)
686 {
687 	udp_t	*udp;
688 
689 	udp = connp->conn_udp;
690 
691 	mutex_enter(&connp->conn_lock);
692 
693 	UDP_MODE_ASSERTIONS(udp, UDP_BECOME_WRITER);
694 
695 	switch (udp->udp_mode) {
696 	case UDP_MT_HOT:
697 		if (udp->udp_reader_count == 1) {
698 			/*
699 			 * We are the only MT thread. Switch to squeue mode
700 			 * immediately.
701 			 */
702 			udp->udp_mode = UDP_SQUEUE;
703 			udp->udp_squeue_count = 1;
704 			CONN_INC_REF_LOCKED(connp);
705 			mutex_exit(&connp->conn_lock);
706 			squeue_enter(connp->conn_sqp, mp, proc, connp, tag);
707 			return;
708 		}
709 		/* FALLTHRU */
710 
711 	case UDP_MT_QUEUED:
712 		/* Enqueue the packet internally in UDP */
713 		udp->udp_mode = UDP_MT_QUEUED;
714 		UDP_ENQUEUE_MP(udp, mp, proc, tag);
715 		mutex_exit(&connp->conn_lock);
716 		return;
717 
718 	case UDP_SQUEUE:
719 	case UDP_QUEUED_SQUEUE:
720 		/*
721 		 * We are already exclusive. i.e. we are already
722 		 * writer. Simply call the desired function.
723 		 */
724 		udp->udp_squeue_count++;
725 		mutex_exit(&connp->conn_lock);
726 		(*proc)(connp, mp, connp->conn_sqp);
727 		return;
728 	}
729 }
730 
731 /*
732  * Transition from MT mode to SQUEUE mode, when the last MT thread
733  * is exiting the UDP perimeter. Move all messages from the internal
734  * udp queue to the squeue. A better way would be to move all the
735  * messages in one shot, this needs more support from the squeue framework
736  */
737 static void
738 udp_switch_to_squeue(udp_t *udp)
739 {
740 	mblk_t *mp;
741 	mblk_t	*mp_next;
742 	sqproc_t proc;
743 	uint8_t	tag;
744 	conn_t	*connp = udp->udp_connp;
745 
746 	ASSERT(MUTEX_HELD(&connp->conn_lock));
747 	ASSERT(udp->udp_mode == UDP_MT_QUEUED);
748 	while (udp->udp_mphead != NULL) {
749 		mp = udp->udp_mphead;
750 		udp->udp_mphead = NULL;
751 		udp->udp_mptail = NULL;
752 		udp->udp_mpcount = 0;
753 		udp->udp_mode = UDP_QUEUED_SQUEUE;
754 		mutex_exit(&connp->conn_lock);
755 		/*
756 		 * It is best not to hold any locks across the calls
757 		 * to squeue functions. Since we drop the lock we
758 		 * need to go back and check the udp_mphead once again
759 		 * after the squeue_fill and hence the while loop at
760 		 * the top of this function
761 		 */
762 		for (; mp != NULL; mp = mp_next) {
763 			mp_next = mp->b_next;
764 			proc = (sqproc_t)mp->b_prev;
765 			tag = (uint8_t)((uintptr_t)mp->b_queue);
766 			mp->b_next = NULL;
767 			mp->b_prev = NULL;
768 			mp->b_queue = NULL;
769 			CONN_INC_REF(connp);
770 			udp->udp_squeue_count++;
771 			squeue_fill(connp->conn_sqp, mp, proc, connp,
772 			    tag);
773 		}
774 		mutex_enter(&connp->conn_lock);
775 	}
776 	/*
777 	 * udp_squeue_count of zero implies that the squeue has drained
778 	 * even before we arrived here (i.e. after the squeue_fill above)
779 	 */
780 	udp->udp_mode = (udp->udp_squeue_count != 0) ?
781 	    UDP_SQUEUE : UDP_MT_HOT;
782 }
783 
784 #define	_UDP_EXIT(connp) {						\
785 	udp_t *_udp = (connp)->conn_udp;				\
786 									\
787 	mutex_enter(&(connp)->conn_lock);				\
788 	UDP_MODE_ASSERTIONS(_udp, UDP_EXIT);				\
789 									\
790 	switch (_udp->udp_mode) {					\
791 	case UDP_MT_HOT:						\
792 		UDP_READERS_DECREF(_udp);				\
793 		mutex_exit(&(connp)->conn_lock);			\
794 		break;							\
795 									\
796 	case UDP_SQUEUE:						\
797 		UDP_SQUEUE_DECREF(_udp);				\
798 		if (_udp->udp_squeue_count == 0)			\
799 		    _udp->udp_mode = UDP_MT_HOT;			\
800 		mutex_exit(&(connp)->conn_lock);			\
801 		break;							\
802 									\
803 	case UDP_MT_QUEUED:						\
804 		/*							\
805 		 * If this is the last MT thread, we need to		\
806 		 * switch to squeue mode				\
807 		 */							\
808 		UDP_READERS_DECREF(_udp);				\
809 		if (_udp->udp_reader_count == 0)			\
810 			udp_switch_to_squeue(_udp);			\
811 		mutex_exit(&(connp)->conn_lock);			\
812 		break;							\
813 									\
814 	case UDP_QUEUED_SQUEUE:						\
815 		UDP_SQUEUE_DECREF(_udp);				\
816 		/*							\
817 		 * Even if the udp_squeue_count drops to zero, we	\
818 		 * don't want to change udp_mode to UDP_MT_HOT here.	\
819 		 * The thread in udp_switch_to_squeue will take care	\
820 		 * of the transition to UDP_MT_HOT, after emptying	\
821 		 * any more new messages that have been enqueued in	\
822 		 * udp_mphead.						\
823 		 */							\
824 		mutex_exit(&(connp)->conn_lock);			\
825 		break;							\
826 	}								\
827 }
828 
829 static void
830 udp_exit(conn_t *connp)
831 {
832 	_UDP_EXIT(connp);
833 }
834 
835 /*
836  * Return the next anonymous port in the privileged port range for
837  * bind checking.
838  *
839  * Trusted Extension (TX) notes: TX allows administrator to mark or
840  * reserve ports as Multilevel ports (MLP). MLP has special function
841  * on TX systems. Once a port is made MLP, it's not available as
842  * ordinary port. This creates "holes" in the port name space. It
843  * may be necessary to skip the "holes" find a suitable anon port.
844  */
845 static in_port_t
846 udp_get_next_priv_port(udp_t *udp)
847 {
848 	static in_port_t next_priv_port = IPPORT_RESERVED - 1;
849 	in_port_t nextport;
850 	boolean_t restart = B_FALSE;
851 	udp_stack_t *us = udp->udp_us;
852 
853 retry:
854 	if (next_priv_port < us->us_min_anonpriv_port ||
855 	    next_priv_port >= IPPORT_RESERVED) {
856 		next_priv_port = IPPORT_RESERVED - 1;
857 		if (restart)
858 			return (0);
859 		restart = B_TRUE;
860 	}
861 
862 	if (is_system_labeled() &&
863 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
864 	    next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) {
865 		next_priv_port = nextport;
866 		goto retry;
867 	}
868 
869 	return (next_priv_port--);
870 }
871 
872 /* UDP bind hash report triggered via the Named Dispatch mechanism. */
873 /* ARGSUSED */
874 static int
875 udp_bind_hash_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
876 {
877 	udp_fanout_t	*udpf;
878 	int		i;
879 	zoneid_t	zoneid;
880 	conn_t		*connp;
881 	udp_t		*udp;
882 	udp_stack_t	*us;
883 
884 	connp = Q_TO_CONN(q);
885 	udp = connp->conn_udp;
886 	us = udp->udp_us;
887 
888 	/* Refer to comments in udp_status_report(). */
889 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
890 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
891 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
892 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
893 			return (0);
894 		}
895 	}
896 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
897 		/* The following may work even if we cannot get a large buf. */
898 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
899 		return (0);
900 	}
901 
902 	(void) mi_mpprintf(mp,
903 	    "UDP     " MI_COL_HDRPAD_STR
904 	/*   12345678[89ABCDEF] */
905 	    " zone lport src addr        dest addr       port  state");
906 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
907 
908 	zoneid = connp->conn_zoneid;
909 
910 	for (i = 0; i < us->us_bind_fanout_size; i++) {
911 		udpf = &us->us_bind_fanout[i];
912 		mutex_enter(&udpf->uf_lock);
913 
914 		/* Print the hash index. */
915 		udp = udpf->uf_udp;
916 		if (zoneid != GLOBAL_ZONEID) {
917 			/* skip to first entry in this zone; might be none */
918 			while (udp != NULL &&
919 			    udp->udp_connp->conn_zoneid != zoneid)
920 				udp = udp->udp_bind_hash;
921 		}
922 		if (udp != NULL) {
923 			uint_t print_len, buf_len;
924 
925 			buf_len = mp->b_cont->b_datap->db_lim -
926 			    mp->b_cont->b_wptr;
927 			print_len = snprintf((char *)mp->b_cont->b_wptr,
928 			    buf_len, "%d\n", i);
929 			if (print_len < buf_len) {
930 				mp->b_cont->b_wptr += print_len;
931 			} else {
932 				mp->b_cont->b_wptr += buf_len;
933 			}
934 			for (; udp != NULL; udp = udp->udp_bind_hash) {
935 				if (zoneid == GLOBAL_ZONEID ||
936 				    zoneid == udp->udp_connp->conn_zoneid)
937 					udp_report_item(mp->b_cont, udp);
938 			}
939 		}
940 		mutex_exit(&udpf->uf_lock);
941 	}
942 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
943 	return (0);
944 }
945 
946 /*
947  * Hash list removal routine for udp_t structures.
948  */
949 static void
950 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock)
951 {
952 	udp_t	*udpnext;
953 	kmutex_t *lockp;
954 	udp_stack_t *us = udp->udp_us;
955 
956 	if (udp->udp_ptpbhn == NULL)
957 		return;
958 
959 	/*
960 	 * Extract the lock pointer in case there are concurrent
961 	 * hash_remove's for this instance.
962 	 */
963 	ASSERT(udp->udp_port != 0);
964 	if (!caller_holds_lock) {
965 		lockp = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
966 		    us->us_bind_fanout_size)].uf_lock;
967 		ASSERT(lockp != NULL);
968 		mutex_enter(lockp);
969 	}
970 	if (udp->udp_ptpbhn != NULL) {
971 		udpnext = udp->udp_bind_hash;
972 		if (udpnext != NULL) {
973 			udpnext->udp_ptpbhn = udp->udp_ptpbhn;
974 			udp->udp_bind_hash = NULL;
975 		}
976 		*udp->udp_ptpbhn = udpnext;
977 		udp->udp_ptpbhn = NULL;
978 	}
979 	if (!caller_holds_lock) {
980 		mutex_exit(lockp);
981 	}
982 }
983 
984 static void
985 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp)
986 {
987 	udp_t	**udpp;
988 	udp_t	*udpnext;
989 
990 	ASSERT(MUTEX_HELD(&uf->uf_lock));
991 	if (udp->udp_ptpbhn != NULL) {
992 		udp_bind_hash_remove(udp, B_TRUE);
993 	}
994 	udpp = &uf->uf_udp;
995 	udpnext = udpp[0];
996 	if (udpnext != NULL) {
997 		/*
998 		 * If the new udp bound to the INADDR_ANY address
999 		 * and the first one in the list is not bound to
1000 		 * INADDR_ANY we skip all entries until we find the
1001 		 * first one bound to INADDR_ANY.
1002 		 * This makes sure that applications binding to a
1003 		 * specific address get preference over those binding to
1004 		 * INADDR_ANY.
1005 		 */
1006 		if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) &&
1007 		    !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) {
1008 			while ((udpnext = udpp[0]) != NULL &&
1009 			    !V6_OR_V4_INADDR_ANY(
1010 			    udpnext->udp_bound_v6src)) {
1011 				udpp = &(udpnext->udp_bind_hash);
1012 			}
1013 			if (udpnext != NULL)
1014 				udpnext->udp_ptpbhn = &udp->udp_bind_hash;
1015 		} else {
1016 			udpnext->udp_ptpbhn = &udp->udp_bind_hash;
1017 		}
1018 	}
1019 	udp->udp_bind_hash = udpnext;
1020 	udp->udp_ptpbhn = udpp;
1021 	udpp[0] = udp;
1022 }
1023 
1024 /*
1025  * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message
1026  * passed to udp_wput.
1027  * It associates a port number and local address with the stream.
1028  * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP
1029  * protocol type (IPPROTO_UDP) placed in the message following the address.
1030  * A T_BIND_ACK message is passed upstream when ip acknowledges the request.
1031  * (Called as writer.)
1032  *
1033  * Note that UDP over IPv4 and IPv6 sockets can use the same port number
1034  * without setting SO_REUSEADDR. This is needed so that they
1035  * can be viewed as two independent transport protocols.
1036  * However, anonymouns ports are allocated from the same range to avoid
1037  * duplicating the us->us_next_port_to_try.
1038  */
1039 static void
1040 udp_bind(queue_t *q, mblk_t *mp)
1041 {
1042 	sin_t		*sin;
1043 	sin6_t		*sin6;
1044 	mblk_t		*mp1;
1045 	in_port_t	port;		/* Host byte order */
1046 	in_port_t	requested_port;	/* Host byte order */
1047 	struct T_bind_req *tbr;
1048 	int		count;
1049 	in6_addr_t	v6src;
1050 	boolean_t	bind_to_req_port_only;
1051 	int		loopmax;
1052 	udp_fanout_t	*udpf;
1053 	in_port_t	lport;		/* Network byte order */
1054 	zoneid_t	zoneid;
1055 	conn_t		*connp;
1056 	udp_t		*udp;
1057 	boolean_t	is_inaddr_any;
1058 	mlp_type_t	addrtype, mlptype;
1059 	udp_stack_t	*us;
1060 
1061 	connp = Q_TO_CONN(q);
1062 	udp = connp->conn_udp;
1063 	us = udp->udp_us;
1064 	if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) {
1065 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1066 		    "udp_bind: bad req, len %u",
1067 		    (uint_t)(mp->b_wptr - mp->b_rptr));
1068 		udp_err_ack(q, mp, TPROTO, 0);
1069 		return;
1070 	}
1071 
1072 	if (udp->udp_state != TS_UNBND) {
1073 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1074 		    "udp_bind: bad state, %u", udp->udp_state);
1075 		udp_err_ack(q, mp, TOUTSTATE, 0);
1076 		return;
1077 	}
1078 	/*
1079 	 * Reallocate the message to make sure we have enough room for an
1080 	 * address and the protocol type.
1081 	 */
1082 	mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1);
1083 	if (!mp1) {
1084 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1085 		return;
1086 	}
1087 
1088 	mp = mp1;
1089 	tbr = (struct T_bind_req *)mp->b_rptr;
1090 	switch (tbr->ADDR_length) {
1091 	case 0:			/* Request for a generic port */
1092 		tbr->ADDR_offset = sizeof (struct T_bind_req);
1093 		if (udp->udp_family == AF_INET) {
1094 			tbr->ADDR_length = sizeof (sin_t);
1095 			sin = (sin_t *)&tbr[1];
1096 			*sin = sin_null;
1097 			sin->sin_family = AF_INET;
1098 			mp->b_wptr = (uchar_t *)&sin[1];
1099 		} else {
1100 			ASSERT(udp->udp_family == AF_INET6);
1101 			tbr->ADDR_length = sizeof (sin6_t);
1102 			sin6 = (sin6_t *)&tbr[1];
1103 			*sin6 = sin6_null;
1104 			sin6->sin6_family = AF_INET6;
1105 			mp->b_wptr = (uchar_t *)&sin6[1];
1106 		}
1107 		port = 0;
1108 		break;
1109 
1110 	case sizeof (sin_t):	/* Complete IPv4 address */
1111 		sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset,
1112 		    sizeof (sin_t));
1113 		if (sin == NULL || !OK_32PTR((char *)sin)) {
1114 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1115 			return;
1116 		}
1117 		if (udp->udp_family != AF_INET ||
1118 		    sin->sin_family != AF_INET) {
1119 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1120 			return;
1121 		}
1122 		port = ntohs(sin->sin_port);
1123 		break;
1124 
1125 	case sizeof (sin6_t):	/* complete IPv6 address */
1126 		sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset,
1127 		    sizeof (sin6_t));
1128 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
1129 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1130 			return;
1131 		}
1132 		if (udp->udp_family != AF_INET6 ||
1133 		    sin6->sin6_family != AF_INET6) {
1134 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1135 			return;
1136 		}
1137 		port = ntohs(sin6->sin6_port);
1138 		break;
1139 
1140 	default:		/* Invalid request */
1141 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1142 		    "udp_bind: bad ADDR_length length %u", tbr->ADDR_length);
1143 		udp_err_ack(q, mp, TBADADDR, 0);
1144 		return;
1145 	}
1146 
1147 	requested_port = port;
1148 
1149 	if (requested_port == 0 || tbr->PRIM_type == O_T_BIND_REQ)
1150 		bind_to_req_port_only = B_FALSE;
1151 	else			/* T_BIND_REQ and requested_port != 0 */
1152 		bind_to_req_port_only = B_TRUE;
1153 
1154 	if (requested_port == 0) {
1155 		/*
1156 		 * If the application passed in zero for the port number, it
1157 		 * doesn't care which port number we bind to. Get one in the
1158 		 * valid range.
1159 		 */
1160 		if (udp->udp_anon_priv_bind) {
1161 			port = udp_get_next_priv_port(udp);
1162 		} else {
1163 			port = udp_update_next_port(udp,
1164 			    us->us_next_port_to_try, B_TRUE);
1165 		}
1166 	} else {
1167 		/*
1168 		 * If the port is in the well-known privileged range,
1169 		 * make sure the caller was privileged.
1170 		 */
1171 		int i;
1172 		boolean_t priv = B_FALSE;
1173 
1174 		if (port < us->us_smallest_nonpriv_port) {
1175 			priv = B_TRUE;
1176 		} else {
1177 			for (i = 0; i < us->us_num_epriv_ports; i++) {
1178 				if (port == us->us_epriv_ports[i]) {
1179 					priv = B_TRUE;
1180 					break;
1181 				}
1182 			}
1183 		}
1184 
1185 		if (priv) {
1186 			cred_t *cr = DB_CREDDEF(mp, connp->conn_cred);
1187 
1188 			if (secpolicy_net_privaddr(cr, port) != 0) {
1189 				udp_err_ack(q, mp, TACCES, 0);
1190 				return;
1191 			}
1192 		}
1193 	}
1194 
1195 	if (port == 0) {
1196 		udp_err_ack(q, mp, TNOADDR, 0);
1197 		return;
1198 	}
1199 
1200 	/*
1201 	 * Copy the source address into our udp structure. This address
1202 	 * may still be zero; if so, IP will fill in the correct address
1203 	 * each time an outbound packet is passed to it.
1204 	 */
1205 	if (udp->udp_family == AF_INET) {
1206 		ASSERT(sin != NULL);
1207 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
1208 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
1209 		    udp->udp_ip_snd_options_len;
1210 		IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src);
1211 	} else {
1212 		ASSERT(sin6 != NULL);
1213 		v6src = sin6->sin6_addr;
1214 		if (IN6_IS_ADDR_V4MAPPED(&v6src)) {
1215 			udp->udp_ipversion = IPV4_VERSION;
1216 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
1217 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
1218 		} else {
1219 			udp->udp_ipversion = IPV6_VERSION;
1220 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
1221 		}
1222 	}
1223 
1224 	/*
1225 	 * If udp_reuseaddr is not set, then we have to make sure that
1226 	 * the IP address and port number the application requested
1227 	 * (or we selected for the application) is not being used by
1228 	 * another stream.  If another stream is already using the
1229 	 * requested IP address and port, the behavior depends on
1230 	 * "bind_to_req_port_only". If set the bind fails; otherwise we
1231 	 * search for any an unused port to bind to the the stream.
1232 	 *
1233 	 * As per the BSD semantics, as modified by the Deering multicast
1234 	 * changes, if udp_reuseaddr is set, then we allow multiple binds
1235 	 * to the same port independent of the local IP address.
1236 	 *
1237 	 * This is slightly different than in SunOS 4.X which did not
1238 	 * support IP multicast. Note that the change implemented by the
1239 	 * Deering multicast code effects all binds - not only binding
1240 	 * to IP multicast addresses.
1241 	 *
1242 	 * Note that when binding to port zero we ignore SO_REUSEADDR in
1243 	 * order to guarantee a unique port.
1244 	 */
1245 
1246 	count = 0;
1247 	if (udp->udp_anon_priv_bind) {
1248 		/*
1249 		 * loopmax = (IPPORT_RESERVED-1) -
1250 		 *    us->us_min_anonpriv_port + 1
1251 		 */
1252 		loopmax = IPPORT_RESERVED - us->us_min_anonpriv_port;
1253 	} else {
1254 		loopmax = us->us_largest_anon_port -
1255 		    us->us_smallest_anon_port + 1;
1256 	}
1257 
1258 	is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src);
1259 	zoneid = connp->conn_zoneid;
1260 
1261 	for (;;) {
1262 		udp_t		*udp1;
1263 		boolean_t	found_exclbind = B_FALSE;
1264 
1265 		/*
1266 		 * Walk through the list of udp streams bound to
1267 		 * requested port with the same IP address.
1268 		 */
1269 		lport = htons(port);
1270 		udpf = &us->us_bind_fanout[UDP_BIND_HASH(lport,
1271 		    us->us_bind_fanout_size)];
1272 		mutex_enter(&udpf->uf_lock);
1273 		for (udp1 = udpf->uf_udp; udp1 != NULL;
1274 		    udp1 = udp1->udp_bind_hash) {
1275 			if (lport != udp1->udp_port)
1276 				continue;
1277 
1278 			/*
1279 			 * On a labeled system, we must treat bindings to ports
1280 			 * on shared IP addresses by sockets with MAC exemption
1281 			 * privilege as being in all zones, as there's
1282 			 * otherwise no way to identify the right receiver.
1283 			 */
1284 			if (!(IPCL_ZONE_MATCH(udp1->udp_connp, zoneid) ||
1285 			    IPCL_ZONE_MATCH(connp,
1286 			    udp1->udp_connp->conn_zoneid)) &&
1287 			    !udp->udp_mac_exempt && !udp1->udp_mac_exempt)
1288 				continue;
1289 
1290 			/*
1291 			 * If UDP_EXCLBIND is set for either the bound or
1292 			 * binding endpoint, the semantics of bind
1293 			 * is changed according to the following chart.
1294 			 *
1295 			 * spec = specified address (v4 or v6)
1296 			 * unspec = unspecified address (v4 or v6)
1297 			 * A = specified addresses are different for endpoints
1298 			 *
1299 			 * bound	bind to		allowed?
1300 			 * -------------------------------------
1301 			 * unspec	unspec		no
1302 			 * unspec	spec		no
1303 			 * spec		unspec		no
1304 			 * spec		spec		yes if A
1305 			 *
1306 			 * For labeled systems, SO_MAC_EXEMPT behaves the same
1307 			 * as UDP_EXCLBIND, except that zoneid is ignored.
1308 			 */
1309 			if (udp1->udp_exclbind || udp->udp_exclbind ||
1310 			    udp1->udp_mac_exempt || udp->udp_mac_exempt) {
1311 				if (V6_OR_V4_INADDR_ANY(
1312 				    udp1->udp_bound_v6src) ||
1313 				    is_inaddr_any ||
1314 				    IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
1315 				    &v6src)) {
1316 					found_exclbind = B_TRUE;
1317 					break;
1318 				}
1319 				continue;
1320 			}
1321 
1322 			/*
1323 			 * Check ipversion to allow IPv4 and IPv6 sockets to
1324 			 * have disjoint port number spaces.
1325 			 */
1326 			if (udp->udp_ipversion != udp1->udp_ipversion) {
1327 
1328 				/*
1329 				 * On the first time through the loop, if the
1330 				 * the user intentionally specified a
1331 				 * particular port number, then ignore any
1332 				 * bindings of the other protocol that may
1333 				 * conflict. This allows the user to bind IPv6
1334 				 * alone and get both v4 and v6, or bind both
1335 				 * both and get each seperately. On subsequent
1336 				 * times through the loop, we're checking a
1337 				 * port that we chose (not the user) and thus
1338 				 * we do not allow casual duplicate bindings.
1339 				 */
1340 				if (count == 0 && requested_port != 0)
1341 					continue;
1342 			}
1343 
1344 			/*
1345 			 * No difference depending on SO_REUSEADDR.
1346 			 *
1347 			 * If existing port is bound to a
1348 			 * non-wildcard IP address and
1349 			 * the requesting stream is bound to
1350 			 * a distinct different IP addresses
1351 			 * (non-wildcard, also), keep going.
1352 			 */
1353 			if (!is_inaddr_any &&
1354 			    !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) &&
1355 			    !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src,
1356 			    &v6src)) {
1357 				continue;
1358 			}
1359 			break;
1360 		}
1361 
1362 		if (!found_exclbind &&
1363 		    (udp->udp_reuseaddr && requested_port != 0)) {
1364 			break;
1365 		}
1366 
1367 		if (udp1 == NULL) {
1368 			/*
1369 			 * No other stream has this IP address
1370 			 * and port number. We can use it.
1371 			 */
1372 			break;
1373 		}
1374 		mutex_exit(&udpf->uf_lock);
1375 		if (bind_to_req_port_only) {
1376 			/*
1377 			 * We get here only when requested port
1378 			 * is bound (and only first  of the for()
1379 			 * loop iteration).
1380 			 *
1381 			 * The semantics of this bind request
1382 			 * require it to fail so we return from
1383 			 * the routine (and exit the loop).
1384 			 *
1385 			 */
1386 			udp_err_ack(q, mp, TADDRBUSY, 0);
1387 			return;
1388 		}
1389 
1390 		if (udp->udp_anon_priv_bind) {
1391 			port = udp_get_next_priv_port(udp);
1392 		} else {
1393 			if ((count == 0) && (requested_port != 0)) {
1394 				/*
1395 				 * If the application wants us to find
1396 				 * a port, get one to start with. Set
1397 				 * requested_port to 0, so that we will
1398 				 * update us->us_next_port_to_try below.
1399 				 */
1400 				port = udp_update_next_port(udp,
1401 				    us->us_next_port_to_try, B_TRUE);
1402 				requested_port = 0;
1403 			} else {
1404 				port = udp_update_next_port(udp, port + 1,
1405 				    B_FALSE);
1406 			}
1407 		}
1408 
1409 		if (port == 0 || ++count >= loopmax) {
1410 			/*
1411 			 * We've tried every possible port number and
1412 			 * there are none available, so send an error
1413 			 * to the user.
1414 			 */
1415 			udp_err_ack(q, mp, TNOADDR, 0);
1416 			return;
1417 		}
1418 	}
1419 
1420 	/*
1421 	 * Copy the source address into our udp structure.  This address
1422 	 * may still be zero; if so, ip will fill in the correct address
1423 	 * each time an outbound packet is passed to it.
1424 	 * If we are binding to a broadcast or multicast address udp_rput
1425 	 * will clear the source address when it receives the T_BIND_ACK.
1426 	 */
1427 	udp->udp_v6src = udp->udp_bound_v6src = v6src;
1428 	udp->udp_port = lport;
1429 	/*
1430 	 * Now reset the the next anonymous port if the application requested
1431 	 * an anonymous port, or we handed out the next anonymous port.
1432 	 */
1433 	if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) {
1434 		us->us_next_port_to_try = port + 1;
1435 	}
1436 
1437 	/* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */
1438 	if (udp->udp_family == AF_INET) {
1439 		sin->sin_port = udp->udp_port;
1440 	} else {
1441 		int error;
1442 
1443 		sin6->sin6_port = udp->udp_port;
1444 		/* Rebuild the header template */
1445 		error = udp_build_hdrs(q, udp);
1446 		if (error != 0) {
1447 			mutex_exit(&udpf->uf_lock);
1448 			udp_err_ack(q, mp, TSYSERR, error);
1449 			return;
1450 		}
1451 	}
1452 	udp->udp_state = TS_IDLE;
1453 	udp_bind_hash_insert(udpf, udp);
1454 	mutex_exit(&udpf->uf_lock);
1455 
1456 	if (cl_inet_bind) {
1457 		/*
1458 		 * Running in cluster mode - register bind information
1459 		 */
1460 		if (udp->udp_ipversion == IPV4_VERSION) {
1461 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET,
1462 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
1463 			    (in_port_t)udp->udp_port);
1464 		} else {
1465 			(*cl_inet_bind)(IPPROTO_UDP, AF_INET6,
1466 			    (uint8_t *)&(udp->udp_v6src),
1467 			    (in_port_t)udp->udp_port);
1468 		}
1469 
1470 	}
1471 
1472 	connp->conn_anon_port = (is_system_labeled() && requested_port == 0);
1473 	if (is_system_labeled() && (!connp->conn_anon_port ||
1474 	    connp->conn_anon_mlp)) {
1475 		uint16_t mlpport;
1476 		cred_t *cr = connp->conn_cred;
1477 		zone_t *zone;
1478 
1479 		zone = crgetzone(cr);
1480 		connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth :
1481 		    mlptSingle;
1482 		addrtype = tsol_mlp_addr_type(zone->zone_id, IPV6_VERSION,
1483 		    &v6src, udp->udp_us->us_netstack->netstack_ip);
1484 		if (addrtype == mlptSingle) {
1485 			udp_err_ack(q, mp, TNOADDR, 0);
1486 			connp->conn_anon_port = B_FALSE;
1487 			connp->conn_mlp_type = mlptSingle;
1488 			return;
1489 		}
1490 		mlpport = connp->conn_anon_port ? PMAPPORT : port;
1491 		mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport,
1492 		    addrtype);
1493 		if (mlptype != mlptSingle &&
1494 		    (connp->conn_mlp_type == mlptSingle ||
1495 		    secpolicy_net_bindmlp(cr) != 0)) {
1496 			if (udp->udp_debug) {
1497 				(void) strlog(UDP_MOD_ID, 0, 1,
1498 				    SL_ERROR|SL_TRACE,
1499 				    "udp_bind: no priv for multilevel port %d",
1500 				    mlpport);
1501 			}
1502 			udp_err_ack(q, mp, TACCES, 0);
1503 			connp->conn_anon_port = B_FALSE;
1504 			connp->conn_mlp_type = mlptSingle;
1505 			return;
1506 		}
1507 
1508 		/*
1509 		 * If we're specifically binding a shared IP address and the
1510 		 * port is MLP on shared addresses, then check to see if this
1511 		 * zone actually owns the MLP.  Reject if not.
1512 		 */
1513 		if (mlptype == mlptShared && addrtype == mlptShared) {
1514 			/*
1515 			 * No need to handle exclusive-stack zones since
1516 			 * ALL_ZONES only applies to the shared stack.
1517 			 */
1518 			zoneid_t mlpzone;
1519 
1520 			mlpzone = tsol_mlp_findzone(IPPROTO_UDP,
1521 			    htons(mlpport));
1522 			if (connp->conn_zoneid != mlpzone) {
1523 				if (udp->udp_debug) {
1524 					(void) strlog(UDP_MOD_ID, 0, 1,
1525 					    SL_ERROR|SL_TRACE,
1526 					    "udp_bind: attempt to bind port "
1527 					    "%d on shared addr in zone %d "
1528 					    "(should be %d)",
1529 					    mlpport, connp->conn_zoneid,
1530 					    mlpzone);
1531 				}
1532 				udp_err_ack(q, mp, TACCES, 0);
1533 				connp->conn_anon_port = B_FALSE;
1534 				connp->conn_mlp_type = mlptSingle;
1535 				return;
1536 			}
1537 		}
1538 		if (connp->conn_anon_port) {
1539 			int error;
1540 
1541 			error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp,
1542 			    port, B_TRUE);
1543 			if (error != 0) {
1544 				if (udp->udp_debug) {
1545 					(void) strlog(UDP_MOD_ID, 0, 1,
1546 					    SL_ERROR|SL_TRACE,
1547 					    "udp_bind: cannot establish anon "
1548 					    "MLP for port %d", port);
1549 				}
1550 				udp_err_ack(q, mp, TACCES, 0);
1551 				connp->conn_anon_port = B_FALSE;
1552 				connp->conn_mlp_type = mlptSingle;
1553 				return;
1554 			}
1555 		}
1556 		connp->conn_mlp_type = mlptype;
1557 	}
1558 
1559 	/* Pass the protocol number in the message following the address. */
1560 	*mp->b_wptr++ = IPPROTO_UDP;
1561 	if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
1562 		/*
1563 		 * Append a request for an IRE if udp_v6src not
1564 		 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address).
1565 		 */
1566 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
1567 		if (!mp->b_cont) {
1568 			udp_err_ack(q, mp, TSYSERR, ENOMEM);
1569 			return;
1570 		}
1571 		mp->b_cont->b_wptr += sizeof (ire_t);
1572 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
1573 	}
1574 	if (udp->udp_family == AF_INET6)
1575 		mp = ip_bind_v6(q, mp, connp, NULL);
1576 	else
1577 		mp = ip_bind_v4(q, mp, connp);
1578 
1579 	if (mp != NULL)
1580 		udp_rput_other(_RD(q), mp);
1581 	else
1582 		CONN_INC_REF(connp);
1583 }
1584 
1585 
1586 void
1587 udp_resume_bind(conn_t *connp, mblk_t *mp)
1588 {
1589 	udp_enter(connp, mp, udp_resume_bind_cb, SQTAG_BIND_RETRY);
1590 }
1591 
1592 /*
1593  * This is called from ip_wput_nondata to resume a deferred UDP bind.
1594  */
1595 /* ARGSUSED */
1596 static void
1597 udp_resume_bind_cb(void *arg, mblk_t *mp, void *arg2)
1598 {
1599 	conn_t *connp = arg;
1600 
1601 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
1602 
1603 	udp_rput_other(connp->conn_rq, mp);
1604 
1605 	CONN_OPER_PENDING_DONE(connp);
1606 	udp_exit(connp);
1607 }
1608 
1609 /*
1610  * This routine handles each T_CONN_REQ message passed to udp.  It
1611  * associates a default destination address with the stream.
1612  *
1613  * This routine sends down a T_BIND_REQ to IP with the following mblks:
1614  *	T_BIND_REQ	- specifying local and remote address/port
1615  *	IRE_DB_REQ_TYPE	- to get an IRE back containing ire_type and src
1616  *	T_OK_ACK	- for the T_CONN_REQ
1617  *	T_CONN_CON	- to keep the TPI user happy
1618  *
1619  * The connect completes in udp_rput.
1620  * When a T_BIND_ACK is received information is extracted from the IRE
1621  * and the two appended messages are sent to the TPI user.
1622  * Should udp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert
1623  * it to an error ack for the appropriate primitive.
1624  */
1625 static void
1626 udp_connect(queue_t *q, mblk_t *mp)
1627 {
1628 	sin6_t	*sin6;
1629 	sin_t	*sin;
1630 	struct T_conn_req	*tcr;
1631 	in6_addr_t v6dst;
1632 	ipaddr_t v4dst;
1633 	uint16_t dstport;
1634 	uint32_t flowinfo;
1635 	mblk_t	*mp1, *mp2;
1636 	udp_fanout_t	*udpf;
1637 	udp_t	*udp, *udp1;
1638 	udp_stack_t	*us;
1639 
1640 	udp = Q_TO_UDP(q);
1641 
1642 	tcr = (struct T_conn_req *)mp->b_rptr;
1643 	us = udp->udp_us;
1644 
1645 	/* A bit of sanity checking */
1646 	if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) {
1647 		udp_err_ack(q, mp, TPROTO, 0);
1648 		return;
1649 	}
1650 	/*
1651 	 * This UDP must have bound to a port already before doing
1652 	 * a connect.
1653 	 */
1654 	if (udp->udp_state == TS_UNBND) {
1655 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
1656 		    "udp_connect: bad state, %u", udp->udp_state);
1657 		udp_err_ack(q, mp, TOUTSTATE, 0);
1658 		return;
1659 	}
1660 	ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL);
1661 
1662 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
1663 	    us->us_bind_fanout_size)];
1664 
1665 	if (udp->udp_state == TS_DATA_XFER) {
1666 		/* Already connected - clear out state */
1667 		mutex_enter(&udpf->uf_lock);
1668 		udp->udp_v6src = udp->udp_bound_v6src;
1669 		udp->udp_state = TS_IDLE;
1670 		mutex_exit(&udpf->uf_lock);
1671 	}
1672 
1673 	if (tcr->OPT_length != 0) {
1674 		udp_err_ack(q, mp, TBADOPT, 0);
1675 		return;
1676 	}
1677 
1678 	/*
1679 	 * Determine packet type based on type of address passed in
1680 	 * the request should contain an IPv4 or IPv6 address.
1681 	 * Make sure that address family matches the type of
1682 	 * family of the the address passed down
1683 	 */
1684 	switch (tcr->DEST_length) {
1685 	default:
1686 		udp_err_ack(q, mp, TBADADDR, 0);
1687 		return;
1688 
1689 	case sizeof (sin_t):
1690 		sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset,
1691 		    sizeof (sin_t));
1692 		if (sin == NULL || !OK_32PTR((char *)sin)) {
1693 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1694 			return;
1695 		}
1696 		if (udp->udp_family != AF_INET ||
1697 		    sin->sin_family != AF_INET) {
1698 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1699 			return;
1700 		}
1701 		v4dst = sin->sin_addr.s_addr;
1702 		dstport = sin->sin_port;
1703 		IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
1704 		ASSERT(udp->udp_ipversion == IPV4_VERSION);
1705 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE +
1706 		    udp->udp_ip_snd_options_len;
1707 		break;
1708 
1709 	case sizeof (sin6_t):
1710 		sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset,
1711 		    sizeof (sin6_t));
1712 		if (sin6 == NULL || !OK_32PTR((char *)sin6)) {
1713 			udp_err_ack(q, mp, TSYSERR, EINVAL);
1714 			return;
1715 		}
1716 		if (udp->udp_family != AF_INET6 ||
1717 		    sin6->sin6_family != AF_INET6) {
1718 			udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT);
1719 			return;
1720 		}
1721 		v6dst = sin6->sin6_addr;
1722 		if (IN6_IS_ADDR_V4MAPPED(&v6dst)) {
1723 			IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst);
1724 			udp->udp_ipversion = IPV4_VERSION;
1725 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
1726 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
1727 			flowinfo = 0;
1728 		} else {
1729 			udp->udp_ipversion = IPV6_VERSION;
1730 			udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len;
1731 			flowinfo = sin6->sin6_flowinfo;
1732 		}
1733 		dstport = sin6->sin6_port;
1734 		break;
1735 	}
1736 	if (dstport == 0) {
1737 		udp_err_ack(q, mp, TBADADDR, 0);
1738 		return;
1739 	}
1740 
1741 	/*
1742 	 * Create a default IP header with no IP options.
1743 	 */
1744 	udp->udp_dstport = dstport;
1745 	if (udp->udp_ipversion == IPV4_VERSION) {
1746 		/*
1747 		 * Interpret a zero destination to mean loopback.
1748 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
1749 		 * generate the T_CONN_CON.
1750 		 */
1751 		if (v4dst == INADDR_ANY) {
1752 			v4dst = htonl(INADDR_LOOPBACK);
1753 			IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst);
1754 			if (udp->udp_family == AF_INET) {
1755 				sin->sin_addr.s_addr = v4dst;
1756 			} else {
1757 				sin6->sin6_addr = v6dst;
1758 			}
1759 		}
1760 		udp->udp_v6dst = v6dst;
1761 		udp->udp_flowinfo = 0;
1762 
1763 		/*
1764 		 * If the destination address is multicast and
1765 		 * an outgoing multicast interface has been set,
1766 		 * use the address of that interface as our
1767 		 * source address if no source address has been set.
1768 		 */
1769 		if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY &&
1770 		    CLASSD(v4dst) &&
1771 		    udp->udp_multicast_if_addr != INADDR_ANY) {
1772 			IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr,
1773 			    &udp->udp_v6src);
1774 		}
1775 	} else {
1776 		ASSERT(udp->udp_ipversion == IPV6_VERSION);
1777 		/*
1778 		 * Interpret a zero destination to mean loopback.
1779 		 * Update the T_CONN_REQ (sin/sin6) since it is used to
1780 		 * generate the T_CONN_CON.
1781 		 */
1782 		if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) {
1783 			v6dst = ipv6_loopback;
1784 			sin6->sin6_addr = v6dst;
1785 		}
1786 		udp->udp_v6dst = v6dst;
1787 		udp->udp_flowinfo = flowinfo;
1788 		/*
1789 		 * If the destination address is multicast and
1790 		 * an outgoing multicast interface has been set,
1791 		 * then the ip bind logic will pick the correct source
1792 		 * address (i.e. matching the outgoing multicast interface).
1793 		 */
1794 	}
1795 
1796 	/*
1797 	 * Verify that the src/port/dst/port and zoneid is unique for all
1798 	 * connections in TS_DATA_XFER
1799 	 */
1800 	mutex_enter(&udpf->uf_lock);
1801 	for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) {
1802 		if (udp1->udp_state != TS_DATA_XFER)
1803 			continue;
1804 		if (udp->udp_port != udp1->udp_port ||
1805 		    udp->udp_ipversion != udp1->udp_ipversion ||
1806 		    dstport != udp1->udp_dstport ||
1807 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) ||
1808 		    !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst) ||
1809 		    !(IPCL_ZONE_MATCH(udp->udp_connp,
1810 		    udp1->udp_connp->conn_zoneid) ||
1811 		    IPCL_ZONE_MATCH(udp1->udp_connp,
1812 		    udp->udp_connp->conn_zoneid)))
1813 			continue;
1814 		mutex_exit(&udpf->uf_lock);
1815 		udp_err_ack(q, mp, TBADADDR, 0);
1816 		return;
1817 	}
1818 	udp->udp_state = TS_DATA_XFER;
1819 	mutex_exit(&udpf->uf_lock);
1820 
1821 	/*
1822 	 * Send down bind to IP to verify that there is a route
1823 	 * and to determine the source address.
1824 	 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput.
1825 	 */
1826 	if (udp->udp_family == AF_INET)
1827 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa_conn_t));
1828 	else
1829 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa6_conn_t));
1830 	if (mp1 == NULL) {
1831 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1832 bind_failed:
1833 		mutex_enter(&udpf->uf_lock);
1834 		udp->udp_state = TS_IDLE;
1835 		mutex_exit(&udpf->uf_lock);
1836 		return;
1837 	}
1838 
1839 	/*
1840 	 * We also have to send a connection confirmation to
1841 	 * keep TLI happy. Prepare it for udp_rput.
1842 	 */
1843 	if (udp->udp_family == AF_INET)
1844 		mp2 = mi_tpi_conn_con(NULL, (char *)sin,
1845 		    sizeof (*sin), NULL, 0);
1846 	else
1847 		mp2 = mi_tpi_conn_con(NULL, (char *)sin6,
1848 		    sizeof (*sin6), NULL, 0);
1849 	if (mp2 == NULL) {
1850 		freemsg(mp1);
1851 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
1852 		goto bind_failed;
1853 	}
1854 
1855 	mp = mi_tpi_ok_ack_alloc(mp);
1856 	if (mp == NULL) {
1857 		/* Unable to reuse the T_CONN_REQ for the ack. */
1858 		freemsg(mp2);
1859 		udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM);
1860 		goto bind_failed;
1861 	}
1862 
1863 	/* Hang onto the T_OK_ACK and T_CONN_CON for later. */
1864 	linkb(mp1, mp);
1865 	linkb(mp1, mp2);
1866 
1867 	mblk_setcred(mp1, udp->udp_connp->conn_cred);
1868 	if (udp->udp_family == AF_INET)
1869 		mp1 = ip_bind_v4(q, mp1, udp->udp_connp);
1870 	else
1871 		mp1 = ip_bind_v6(q, mp1, udp->udp_connp, NULL);
1872 
1873 	if (mp1 != NULL)
1874 		udp_rput_other(_RD(q), mp1);
1875 	else
1876 		CONN_INC_REF(udp->udp_connp);
1877 }
1878 
1879 static int
1880 udp_close(queue_t *q)
1881 {
1882 	conn_t	*connp = Q_TO_CONN(UDP_WR(q));
1883 	udp_t	*udp;
1884 	queue_t	*ip_rq = RD(UDP_WR(q));
1885 
1886 	ASSERT(connp != NULL && IPCL_IS_UDP(connp));
1887 	udp = connp->conn_udp;
1888 
1889 	ip_quiesce_conn(connp);
1890 	/*
1891 	 * Disable read-side synchronous stream
1892 	 * interface and drain any queued data.
1893 	 */
1894 	udp_rcv_drain(q, udp, B_TRUE);
1895 	ASSERT(!udp->udp_direct_sockfs);
1896 
1897 	qprocsoff(q);
1898 
1899 	/* restore IP module's high and low water marks to default values */
1900 	ip_rq->q_hiwat = ip_rq->q_qinfo->qi_minfo->mi_hiwat;
1901 	WR(ip_rq)->q_hiwat = WR(ip_rq)->q_qinfo->qi_minfo->mi_hiwat;
1902 	WR(ip_rq)->q_lowat = WR(ip_rq)->q_qinfo->qi_minfo->mi_lowat;
1903 
1904 	ASSERT(udp->udp_rcv_cnt == 0);
1905 	ASSERT(udp->udp_rcv_msgcnt == 0);
1906 	ASSERT(udp->udp_rcv_list_head == NULL);
1907 	ASSERT(udp->udp_rcv_list_tail == NULL);
1908 
1909 	udp_close_free(connp);
1910 
1911 	/*
1912 	 * Restore connp as an IP endpoint.
1913 	 * Locking required to prevent a race with udp_snmp_get()/
1914 	 * ipcl_get_next_conn(), which selects conn_t which are
1915 	 * IPCL_UDP and not CONN_CONDEMNED.
1916 	 */
1917 	mutex_enter(&connp->conn_lock);
1918 	connp->conn_flags &= ~IPCL_UDP;
1919 	connp->conn_state_flags &=
1920 	    ~(CONN_CLOSING | CONN_CONDEMNED | CONN_QUIESCED);
1921 	connp->conn_ulp_labeled = B_FALSE;
1922 	mutex_exit(&connp->conn_lock);
1923 
1924 	return (0);
1925 }
1926 
1927 /*
1928  * Called in the close path from IP (ip_quiesce_conn) to quiesce the conn
1929  */
1930 void
1931 udp_quiesce_conn(conn_t *connp)
1932 {
1933 	udp_t	*udp = connp->conn_udp;
1934 
1935 	if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) {
1936 		/*
1937 		 * Running in cluster mode - register unbind information
1938 		 */
1939 		if (udp->udp_ipversion == IPV4_VERSION) {
1940 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
1941 			    (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))),
1942 			    (in_port_t)udp->udp_port);
1943 		} else {
1944 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
1945 			    (uint8_t *)(&(udp->udp_v6src)),
1946 			    (in_port_t)udp->udp_port);
1947 		}
1948 	}
1949 
1950 	udp_bind_hash_remove(udp, B_FALSE);
1951 
1952 	mutex_enter(&connp->conn_lock);
1953 	while (udp->udp_reader_count != 0 || udp->udp_squeue_count != 0 ||
1954 	    udp->udp_mode != UDP_MT_HOT) {
1955 		cv_wait(&connp->conn_cv, &connp->conn_lock);
1956 	}
1957 	mutex_exit(&connp->conn_lock);
1958 }
1959 
1960 void
1961 udp_close_free(conn_t *connp)
1962 {
1963 	udp_t *udp = connp->conn_udp;
1964 
1965 	/* If there are any options associated with the stream, free them. */
1966 	if (udp->udp_ip_snd_options) {
1967 		mi_free((char *)udp->udp_ip_snd_options);
1968 		udp->udp_ip_snd_options = NULL;
1969 	}
1970 
1971 	if (udp->udp_ip_rcv_options) {
1972 		mi_free((char *)udp->udp_ip_rcv_options);
1973 		udp->udp_ip_rcv_options = NULL;
1974 	}
1975 
1976 	/* Free memory associated with sticky options */
1977 	if (udp->udp_sticky_hdrs_len != 0) {
1978 		kmem_free(udp->udp_sticky_hdrs,
1979 		    udp->udp_sticky_hdrs_len);
1980 		udp->udp_sticky_hdrs = NULL;
1981 		udp->udp_sticky_hdrs_len = 0;
1982 	}
1983 
1984 	ip6_pkt_free(&udp->udp_sticky_ipp);
1985 
1986 	udp->udp_connp = NULL;
1987 	netstack_rele(udp->udp_us->us_netstack);
1988 
1989 	connp->conn_udp = NULL;
1990 	kmem_cache_free(udp_cache, udp);
1991 }
1992 
1993 /*
1994  * This routine handles each T_DISCON_REQ message passed to udp
1995  * as an indicating that UDP is no longer connected. This results
1996  * in sending a T_BIND_REQ to IP to restore the binding to just
1997  * the local address/port.
1998  *
1999  * This routine sends down a T_BIND_REQ to IP with the following mblks:
2000  *	T_BIND_REQ	- specifying just the local address/port
2001  *	T_OK_ACK	- for the T_DISCON_REQ
2002  *
2003  * The disconnect completes in udp_rput.
2004  * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user.
2005  * Should udp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert
2006  * it to an error ack for the appropriate primitive.
2007  */
2008 static void
2009 udp_disconnect(queue_t *q, mblk_t *mp)
2010 {
2011 	udp_t	*udp = Q_TO_UDP(q);
2012 	mblk_t	*mp1;
2013 	udp_fanout_t *udpf;
2014 	udp_stack_t *us;
2015 
2016 	us = udp->udp_us;
2017 	if (udp->udp_state != TS_DATA_XFER) {
2018 		(void) mi_strlog(q, 1, SL_ERROR|SL_TRACE,
2019 		    "udp_disconnect: bad state, %u", udp->udp_state);
2020 		udp_err_ack(q, mp, TOUTSTATE, 0);
2021 		return;
2022 	}
2023 	udpf = &us->us_bind_fanout[UDP_BIND_HASH(udp->udp_port,
2024 	    us->us_bind_fanout_size)];
2025 	mutex_enter(&udpf->uf_lock);
2026 	udp->udp_v6src = udp->udp_bound_v6src;
2027 	udp->udp_state = TS_IDLE;
2028 	mutex_exit(&udpf->uf_lock);
2029 
2030 	/*
2031 	 * Send down bind to IP to remove the full binding and revert
2032 	 * to the local address binding.
2033 	 */
2034 	if (udp->udp_family == AF_INET)
2035 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin_t));
2036 	else
2037 		mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin6_t));
2038 	if (mp1 == NULL) {
2039 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
2040 		return;
2041 	}
2042 	mp = mi_tpi_ok_ack_alloc(mp);
2043 	if (mp == NULL) {
2044 		/* Unable to reuse the T_DISCON_REQ for the ack. */
2045 		udp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM);
2046 		return;
2047 	}
2048 
2049 	if (udp->udp_family == AF_INET6) {
2050 		int error;
2051 
2052 		/* Rebuild the header template */
2053 		error = udp_build_hdrs(q, udp);
2054 		if (error != 0) {
2055 			udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error);
2056 			freemsg(mp1);
2057 			return;
2058 		}
2059 	}
2060 	mutex_enter(&udpf->uf_lock);
2061 	udp->udp_discon_pending = 1;
2062 	mutex_exit(&udpf->uf_lock);
2063 
2064 	/* Append the T_OK_ACK to the T_BIND_REQ for udp_rput */
2065 	linkb(mp1, mp);
2066 
2067 	if (udp->udp_family == AF_INET6)
2068 		mp1 = ip_bind_v6(q, mp1, udp->udp_connp, NULL);
2069 	else
2070 		mp1 = ip_bind_v4(q, mp1, udp->udp_connp);
2071 
2072 	if (mp1 != NULL)
2073 		udp_rput_other(_RD(q), mp1);
2074 	else
2075 		CONN_INC_REF(udp->udp_connp);
2076 }
2077 
2078 /* This routine creates a T_ERROR_ACK message and passes it upstream. */
2079 static void
2080 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error)
2081 {
2082 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
2083 		putnext(UDP_RD(q), mp);
2084 }
2085 
2086 /* Shorthand to generate and send TPI error acks to our client */
2087 static void
2088 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error,
2089     int sys_error)
2090 {
2091 	struct T_error_ack	*teackp;
2092 
2093 	if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack),
2094 	    M_PCPROTO, T_ERROR_ACK)) != NULL) {
2095 		teackp = (struct T_error_ack *)mp->b_rptr;
2096 		teackp->ERROR_prim = primitive;
2097 		teackp->TLI_error = t_error;
2098 		teackp->UNIX_error = sys_error;
2099 		putnext(UDP_RD(q), mp);
2100 	}
2101 }
2102 
2103 /*ARGSUSED*/
2104 static int
2105 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
2106 {
2107 	int i;
2108 	udp_t		*udp = Q_TO_UDP(q);
2109 	udp_stack_t *us = udp->udp_us;
2110 
2111 	for (i = 0; i < us->us_num_epriv_ports; i++) {
2112 		if (us->us_epriv_ports[i] != 0)
2113 			(void) mi_mpprintf(mp, "%d ", us->us_epriv_ports[i]);
2114 	}
2115 	return (0);
2116 }
2117 
2118 /* ARGSUSED */
2119 static int
2120 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
2121     cred_t *cr)
2122 {
2123 	long	new_value;
2124 	int	i;
2125 	udp_t		*udp = Q_TO_UDP(q);
2126 	udp_stack_t *us = udp->udp_us;
2127 
2128 	/*
2129 	 * Fail the request if the new value does not lie within the
2130 	 * port number limits.
2131 	 */
2132 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
2133 	    new_value <= 0 || new_value >= 65536) {
2134 		return (EINVAL);
2135 	}
2136 
2137 	/* Check if the value is already in the list */
2138 	for (i = 0; i < us->us_num_epriv_ports; i++) {
2139 		if (new_value == us->us_epriv_ports[i]) {
2140 			return (EEXIST);
2141 		}
2142 	}
2143 	/* Find an empty slot */
2144 	for (i = 0; i < us->us_num_epriv_ports; i++) {
2145 		if (us->us_epriv_ports[i] == 0)
2146 			break;
2147 	}
2148 	if (i == us->us_num_epriv_ports) {
2149 		return (EOVERFLOW);
2150 	}
2151 
2152 	/* Set the new value */
2153 	us->us_epriv_ports[i] = (in_port_t)new_value;
2154 	return (0);
2155 }
2156 
2157 /* ARGSUSED */
2158 static int
2159 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp,
2160     cred_t *cr)
2161 {
2162 	long	new_value;
2163 	int	i;
2164 	udp_t		*udp = Q_TO_UDP(q);
2165 	udp_stack_t *us = udp->udp_us;
2166 
2167 	/*
2168 	 * Fail the request if the new value does not lie within the
2169 	 * port number limits.
2170 	 */
2171 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
2172 	    new_value <= 0 || new_value >= 65536) {
2173 		return (EINVAL);
2174 	}
2175 
2176 	/* Check that the value is already in the list */
2177 	for (i = 0; i < us->us_num_epriv_ports; i++) {
2178 		if (us->us_epriv_ports[i] == new_value)
2179 			break;
2180 	}
2181 	if (i == us->us_num_epriv_ports) {
2182 		return (ESRCH);
2183 	}
2184 
2185 	/* Clear the value */
2186 	us->us_epriv_ports[i] = 0;
2187 	return (0);
2188 }
2189 
2190 /* At minimum we need 4 bytes of UDP header */
2191 #define	ICMP_MIN_UDP_HDR	4
2192 
2193 /*
2194  * udp_icmp_error is called by udp_rput to process ICMP msgs. passed up by IP.
2195  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
2196  * Assumes that IP has pulled up everything up to and including the ICMP header.
2197  * An M_CTL could potentially come here from some other module (i.e. if UDP
2198  * is pushed on some module other than IP). Thus, if we find that the M_CTL
2199  * does not have enough ICMP information , following STREAMS conventions,
2200  * we send it upstream assuming it is an M_CTL we don't understand.
2201  */
2202 static void
2203 udp_icmp_error(queue_t *q, mblk_t *mp)
2204 {
2205 	icmph_t *icmph;
2206 	ipha_t	*ipha;
2207 	int	iph_hdr_length;
2208 	udpha_t	*udpha;
2209 	sin_t	sin;
2210 	sin6_t	sin6;
2211 	mblk_t	*mp1;
2212 	int	error = 0;
2213 	size_t	mp_size = MBLKL(mp);
2214 	udp_t	*udp = Q_TO_UDP(q);
2215 
2216 	/*
2217 	 * Assume IP provides aligned packets - otherwise toss
2218 	 */
2219 	if (!OK_32PTR(mp->b_rptr)) {
2220 		freemsg(mp);
2221 		return;
2222 	}
2223 
2224 	/*
2225 	 * Verify that we have a complete IP header and the application has
2226 	 * asked for errors. If not, send it upstream.
2227 	 */
2228 	if (!udp->udp_dgram_errind || mp_size < sizeof (ipha_t)) {
2229 noticmpv4:
2230 		putnext(UDP_RD(q), mp);
2231 		return;
2232 	}
2233 
2234 	ipha = (ipha_t *)mp->b_rptr;
2235 	/*
2236 	 * Verify IP version. Anything other than IPv4 or IPv6 packet is sent
2237 	 * upstream. ICMPv6  is handled in udp_icmp_error_ipv6.
2238 	 */
2239 	switch (IPH_HDR_VERSION(ipha)) {
2240 	case IPV6_VERSION:
2241 		udp_icmp_error_ipv6(q, mp);
2242 		return;
2243 	case IPV4_VERSION:
2244 		break;
2245 	default:
2246 		goto noticmpv4;
2247 	}
2248 
2249 	/* Skip past the outer IP and ICMP headers */
2250 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
2251 	icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length];
2252 	/*
2253 	 * If we don't have the correct outer IP header length or if the ULP
2254 	 * is not IPPROTO_ICMP or if we don't have a complete inner IP header
2255 	 * send the packet upstream.
2256 	 */
2257 	if (iph_hdr_length < sizeof (ipha_t) ||
2258 	    ipha->ipha_protocol != IPPROTO_ICMP ||
2259 	    (ipha_t *)&icmph[1] + 1 > (ipha_t *)mp->b_wptr) {
2260 		goto noticmpv4;
2261 	}
2262 	ipha = (ipha_t *)&icmph[1];
2263 
2264 	/* Skip past the inner IP and find the ULP header */
2265 	iph_hdr_length = IPH_HDR_LENGTH(ipha);
2266 	udpha = (udpha_t *)((char *)ipha + iph_hdr_length);
2267 	/*
2268 	 * If we don't have the correct inner IP header length or if the ULP
2269 	 * is not IPPROTO_UDP or if we don't have at least ICMP_MIN_UDP_HDR
2270 	 * bytes of UDP header, send it upstream.
2271 	 */
2272 	if (iph_hdr_length < sizeof (ipha_t) ||
2273 	    ipha->ipha_protocol != IPPROTO_UDP ||
2274 	    (uchar_t *)udpha + ICMP_MIN_UDP_HDR > mp->b_wptr) {
2275 		goto noticmpv4;
2276 	}
2277 
2278 	switch (icmph->icmph_type) {
2279 	case ICMP_DEST_UNREACHABLE:
2280 		switch (icmph->icmph_code) {
2281 		case ICMP_FRAGMENTATION_NEEDED:
2282 			/*
2283 			 * IP has already adjusted the path MTU.
2284 			 * XXX Somehow pass MTU indication to application?
2285 			 */
2286 			break;
2287 		case ICMP_PORT_UNREACHABLE:
2288 		case ICMP_PROTOCOL_UNREACHABLE:
2289 			error = ECONNREFUSED;
2290 			break;
2291 		default:
2292 			/* Transient errors */
2293 			break;
2294 		}
2295 		break;
2296 	default:
2297 		/* Transient errors */
2298 		break;
2299 	}
2300 	if (error == 0) {
2301 		freemsg(mp);
2302 		return;
2303 	}
2304 
2305 	switch (udp->udp_family) {
2306 	case AF_INET:
2307 		sin = sin_null;
2308 		sin.sin_family = AF_INET;
2309 		sin.sin_addr.s_addr = ipha->ipha_dst;
2310 		sin.sin_port = udpha->uha_dst_port;
2311 		mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0,
2312 		    error);
2313 		break;
2314 	case AF_INET6:
2315 		sin6 = sin6_null;
2316 		sin6.sin6_family = AF_INET6;
2317 		IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr);
2318 		sin6.sin6_port = udpha->uha_dst_port;
2319 
2320 		mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t),
2321 		    NULL, 0, error);
2322 		break;
2323 	}
2324 	if (mp1)
2325 		putnext(UDP_RD(q), mp1);
2326 	freemsg(mp);
2327 }
2328 
2329 /*
2330  * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6.
2331  * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors.
2332  * Assumes that IP has pulled up all the extension headers as well as the
2333  * ICMPv6 header.
2334  * An M_CTL could potentially come here from some other module (i.e. if UDP
2335  * is pushed on some module other than IP). Thus, if we find that the M_CTL
2336  * does not have enough ICMP information , following STREAMS conventions,
2337  * we send it upstream assuming it is an M_CTL we don't understand. The reason
2338  * it might get here is if the non-ICMP M_CTL accidently has 6 in the version
2339  * field (when cast to ipha_t in udp_icmp_error).
2340  */
2341 static void
2342 udp_icmp_error_ipv6(queue_t *q, mblk_t *mp)
2343 {
2344 	icmp6_t		*icmp6;
2345 	ip6_t		*ip6h, *outer_ip6h;
2346 	uint16_t	hdr_length;
2347 	uint8_t		*nexthdrp;
2348 	udpha_t		*udpha;
2349 	sin6_t		sin6;
2350 	mblk_t		*mp1;
2351 	int		error = 0;
2352 	size_t		mp_size = MBLKL(mp);
2353 	udp_t		*udp = Q_TO_UDP(q);
2354 
2355 	/*
2356 	 * Verify that we have a complete IP header. If not, send it upstream.
2357 	 */
2358 	if (mp_size < sizeof (ip6_t)) {
2359 noticmpv6:
2360 		putnext(UDP_RD(q), mp);
2361 		return;
2362 	}
2363 
2364 	outer_ip6h = (ip6_t *)mp->b_rptr;
2365 	/*
2366 	 * Verify this is an ICMPV6 packet, else send it upstream
2367 	 */
2368 	if (outer_ip6h->ip6_nxt == IPPROTO_ICMPV6) {
2369 		hdr_length = IPV6_HDR_LEN;
2370 	} else if (!ip_hdr_length_nexthdr_v6(mp, outer_ip6h, &hdr_length,
2371 	    &nexthdrp) ||
2372 	    *nexthdrp != IPPROTO_ICMPV6) {
2373 		goto noticmpv6;
2374 	}
2375 	icmp6 = (icmp6_t *)&mp->b_rptr[hdr_length];
2376 	ip6h = (ip6_t *)&icmp6[1];
2377 	/*
2378 	 * Verify we have a complete ICMP and inner IP header.
2379 	 */
2380 	if ((uchar_t *)&ip6h[1] > mp->b_wptr)
2381 		goto noticmpv6;
2382 
2383 	if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &hdr_length, &nexthdrp))
2384 		goto noticmpv6;
2385 	udpha = (udpha_t *)((char *)ip6h + hdr_length);
2386 	/*
2387 	 * Validate inner header. If the ULP is not IPPROTO_UDP or if we don't
2388 	 * have at least ICMP_MIN_UDP_HDR bytes of  UDP header send the
2389 	 * packet upstream.
2390 	 */
2391 	if ((*nexthdrp != IPPROTO_UDP) ||
2392 	    ((uchar_t *)udpha + ICMP_MIN_UDP_HDR) > mp->b_wptr) {
2393 		goto noticmpv6;
2394 	}
2395 
2396 	switch (icmp6->icmp6_type) {
2397 	case ICMP6_DST_UNREACH:
2398 		switch (icmp6->icmp6_code) {
2399 		case ICMP6_DST_UNREACH_NOPORT:
2400 			error = ECONNREFUSED;
2401 			break;
2402 		case ICMP6_DST_UNREACH_ADMIN:
2403 		case ICMP6_DST_UNREACH_NOROUTE:
2404 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
2405 		case ICMP6_DST_UNREACH_ADDR:
2406 			/* Transient errors */
2407 			break;
2408 		default:
2409 			break;
2410 		}
2411 		break;
2412 	case ICMP6_PACKET_TOO_BIG: {
2413 		struct T_unitdata_ind	*tudi;
2414 		struct T_opthdr		*toh;
2415 		size_t			udi_size;
2416 		mblk_t			*newmp;
2417 		t_scalar_t		opt_length = sizeof (struct T_opthdr) +
2418 		    sizeof (struct ip6_mtuinfo);
2419 		sin6_t			*sin6;
2420 		struct ip6_mtuinfo	*mtuinfo;
2421 
2422 		/*
2423 		 * If the application has requested to receive path mtu
2424 		 * information, send up an empty message containing an
2425 		 * IPV6_PATHMTU ancillary data item.
2426 		 */
2427 		if (!udp->udp_ipv6_recvpathmtu)
2428 			break;
2429 
2430 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) +
2431 		    opt_length;
2432 		if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) {
2433 			BUMP_MIB(&udp->udp_mib, udpInErrors);
2434 			break;
2435 		}
2436 
2437 		/*
2438 		 * newmp->b_cont is left to NULL on purpose.  This is an
2439 		 * empty message containing only ancillary data.
2440 		 */
2441 		newmp->b_datap->db_type = M_PROTO;
2442 		tudi = (struct T_unitdata_ind *)newmp->b_rptr;
2443 		newmp->b_wptr = (uchar_t *)tudi + udi_size;
2444 		tudi->PRIM_type = T_UNITDATA_IND;
2445 		tudi->SRC_length = sizeof (sin6_t);
2446 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
2447 		tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t);
2448 		tudi->OPT_length = opt_length;
2449 
2450 		sin6 = (sin6_t *)&tudi[1];
2451 		bzero(sin6, sizeof (sin6_t));
2452 		sin6->sin6_family = AF_INET6;
2453 		sin6->sin6_addr = udp->udp_v6dst;
2454 
2455 		toh = (struct T_opthdr *)&sin6[1];
2456 		toh->level = IPPROTO_IPV6;
2457 		toh->name = IPV6_PATHMTU;
2458 		toh->len = opt_length;
2459 		toh->status = 0;
2460 
2461 		mtuinfo = (struct ip6_mtuinfo *)&toh[1];
2462 		bzero(mtuinfo, sizeof (struct ip6_mtuinfo));
2463 		mtuinfo->ip6m_addr.sin6_family = AF_INET6;
2464 		mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst;
2465 		mtuinfo->ip6m_mtu = icmp6->icmp6_mtu;
2466 		/*
2467 		 * We've consumed everything we need from the original
2468 		 * message.  Free it, then send our empty message.
2469 		 */
2470 		freemsg(mp);
2471 		putnext(UDP_RD(q), newmp);
2472 		return;
2473 	}
2474 	case ICMP6_TIME_EXCEEDED:
2475 		/* Transient errors */
2476 		break;
2477 	case ICMP6_PARAM_PROB:
2478 		/* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */
2479 		if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER &&
2480 		    (uchar_t *)ip6h + icmp6->icmp6_pptr ==
2481 		    (uchar_t *)nexthdrp) {
2482 			error = ECONNREFUSED;
2483 			break;
2484 		}
2485 		break;
2486 	}
2487 	if (error == 0) {
2488 		freemsg(mp);
2489 		return;
2490 	}
2491 
2492 	sin6 = sin6_null;
2493 	sin6.sin6_family = AF_INET6;
2494 	sin6.sin6_addr = ip6h->ip6_dst;
2495 	sin6.sin6_port = udpha->uha_dst_port;
2496 	sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK;
2497 
2498 	mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0,
2499 	    error);
2500 	if (mp1)
2501 		putnext(UDP_RD(q), mp1);
2502 	freemsg(mp);
2503 }
2504 
2505 /*
2506  * This routine responds to T_ADDR_REQ messages.  It is called by udp_wput.
2507  * The local address is filled in if endpoint is bound. The remote address
2508  * is filled in if remote address has been precified ("connected endpoint")
2509  * (The concept of connected CLTS sockets is alien to published TPI
2510  *  but we support it anyway).
2511  */
2512 static void
2513 udp_addr_req(queue_t *q, mblk_t *mp)
2514 {
2515 	sin_t	*sin;
2516 	sin6_t	*sin6;
2517 	mblk_t	*ackmp;
2518 	struct T_addr_ack *taa;
2519 	udp_t	*udp = Q_TO_UDP(q);
2520 
2521 	/* Make it large enough for worst case */
2522 	ackmp = reallocb(mp, sizeof (struct T_addr_ack) +
2523 	    2 * sizeof (sin6_t), 1);
2524 	if (ackmp == NULL) {
2525 		udp_err_ack(q, mp, TSYSERR, ENOMEM);
2526 		return;
2527 	}
2528 	taa = (struct T_addr_ack *)ackmp->b_rptr;
2529 
2530 	bzero(taa, sizeof (struct T_addr_ack));
2531 	ackmp->b_wptr = (uchar_t *)&taa[1];
2532 
2533 	taa->PRIM_type = T_ADDR_ACK;
2534 	ackmp->b_datap->db_type = M_PCPROTO;
2535 	/*
2536 	 * Note: Following code assumes 32 bit alignment of basic
2537 	 * data structures like sin_t and struct T_addr_ack.
2538 	 */
2539 	if (udp->udp_state != TS_UNBND) {
2540 		/*
2541 		 * Fill in local address first
2542 		 */
2543 		taa->LOCADDR_offset = sizeof (*taa);
2544 		if (udp->udp_family == AF_INET) {
2545 			taa->LOCADDR_length = sizeof (sin_t);
2546 			sin = (sin_t *)&taa[1];
2547 			/* Fill zeroes and then initialize non-zero fields */
2548 			*sin = sin_null;
2549 			sin->sin_family = AF_INET;
2550 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
2551 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
2552 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src,
2553 				    sin->sin_addr.s_addr);
2554 			} else {
2555 				/*
2556 				 * INADDR_ANY
2557 				 * udp_v6src is not set, we might be bound to
2558 				 * broadcast/multicast. Use udp_bound_v6src as
2559 				 * local address instead (that could
2560 				 * also still be INADDR_ANY)
2561 				 */
2562 				IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src,
2563 				    sin->sin_addr.s_addr);
2564 			}
2565 			sin->sin_port = udp->udp_port;
2566 			ackmp->b_wptr = (uchar_t *)&sin[1];
2567 			if (udp->udp_state == TS_DATA_XFER) {
2568 				/*
2569 				 * connected, fill remote address too
2570 				 */
2571 				taa->REMADDR_length = sizeof (sin_t);
2572 				/* assumed 32-bit alignment */
2573 				taa->REMADDR_offset = taa->LOCADDR_offset +
2574 				    taa->LOCADDR_length;
2575 
2576 				sin = (sin_t *)(ackmp->b_rptr +
2577 				    taa->REMADDR_offset);
2578 				/* initialize */
2579 				*sin = sin_null;
2580 				sin->sin_family = AF_INET;
2581 				sin->sin_addr.s_addr =
2582 				    V4_PART_OF_V6(udp->udp_v6dst);
2583 				sin->sin_port = udp->udp_dstport;
2584 				ackmp->b_wptr = (uchar_t *)&sin[1];
2585 			}
2586 		} else {
2587 			taa->LOCADDR_length = sizeof (sin6_t);
2588 			sin6 = (sin6_t *)&taa[1];
2589 			/* Fill zeroes and then initialize non-zero fields */
2590 			*sin6 = sin6_null;
2591 			sin6->sin6_family = AF_INET6;
2592 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
2593 				sin6->sin6_addr = udp->udp_v6src;
2594 			} else {
2595 				/*
2596 				 * UNSPECIFIED
2597 				 * udp_v6src is not set, we might be bound to
2598 				 * broadcast/multicast. Use udp_bound_v6src as
2599 				 * local address instead (that could
2600 				 * also still be UNSPECIFIED)
2601 				 */
2602 				sin6->sin6_addr =
2603 				    udp->udp_bound_v6src;
2604 			}
2605 			sin6->sin6_port = udp->udp_port;
2606 			ackmp->b_wptr = (uchar_t *)&sin6[1];
2607 			if (udp->udp_state == TS_DATA_XFER) {
2608 				/*
2609 				 * connected, fill remote address too
2610 				 */
2611 				taa->REMADDR_length = sizeof (sin6_t);
2612 				/* assumed 32-bit alignment */
2613 				taa->REMADDR_offset = taa->LOCADDR_offset +
2614 				    taa->LOCADDR_length;
2615 
2616 				sin6 = (sin6_t *)(ackmp->b_rptr +
2617 				    taa->REMADDR_offset);
2618 				/* initialize */
2619 				*sin6 = sin6_null;
2620 				sin6->sin6_family = AF_INET6;
2621 				sin6->sin6_addr = udp->udp_v6dst;
2622 				sin6->sin6_port =  udp->udp_dstport;
2623 				ackmp->b_wptr = (uchar_t *)&sin6[1];
2624 			}
2625 			ackmp->b_wptr = (uchar_t *)&sin6[1];
2626 		}
2627 	}
2628 	ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim);
2629 	putnext(UDP_RD(q), ackmp);
2630 }
2631 
2632 static void
2633 udp_copy_info(struct T_info_ack *tap, udp_t *udp)
2634 {
2635 	if (udp->udp_family == AF_INET) {
2636 		*tap = udp_g_t_info_ack_ipv4;
2637 	} else {
2638 		*tap = udp_g_t_info_ack_ipv6;
2639 	}
2640 	tap->CURRENT_state = udp->udp_state;
2641 	tap->OPT_size = udp_max_optsize;
2642 }
2643 
2644 /*
2645  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
2646  * udp_wput.  Much of the T_CAPABILITY_ACK information is copied from
2647  * udp_g_t_info_ack.  The current state of the stream is copied from
2648  * udp_state.
2649  */
2650 static void
2651 udp_capability_req(queue_t *q, mblk_t *mp)
2652 {
2653 	t_uscalar_t		cap_bits1;
2654 	struct T_capability_ack	*tcap;
2655 	udp_t	*udp = Q_TO_UDP(q);
2656 
2657 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
2658 
2659 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
2660 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
2661 	if (!mp)
2662 		return;
2663 
2664 	tcap = (struct T_capability_ack *)mp->b_rptr;
2665 	tcap->CAP_bits1 = 0;
2666 
2667 	if (cap_bits1 & TC1_INFO) {
2668 		udp_copy_info(&tcap->INFO_ack, udp);
2669 		tcap->CAP_bits1 |= TC1_INFO;
2670 	}
2671 
2672 	putnext(UDP_RD(q), mp);
2673 }
2674 
2675 /*
2676  * This routine responds to T_INFO_REQ messages.  It is called by udp_wput.
2677  * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack.
2678  * The current state of the stream is copied from udp_state.
2679  */
2680 static void
2681 udp_info_req(queue_t *q, mblk_t *mp)
2682 {
2683 	udp_t *udp = Q_TO_UDP(q);
2684 
2685 	/* Create a T_INFO_ACK message. */
2686 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
2687 	    T_INFO_ACK);
2688 	if (!mp)
2689 		return;
2690 	udp_copy_info((struct T_info_ack *)mp->b_rptr, udp);
2691 	putnext(UDP_RD(q), mp);
2692 }
2693 
2694 /*
2695  * IP recognizes seven kinds of bind requests:
2696  *
2697  * - A zero-length address binds only to the protocol number.
2698  *
2699  * - A 4-byte address is treated as a request to
2700  * validate that the address is a valid local IPv4
2701  * address, appropriate for an application to bind to.
2702  * IP does the verification, but does not make any note
2703  * of the address at this time.
2704  *
2705  * - A 16-byte address contains is treated as a request
2706  * to validate a local IPv6 address, as the 4-byte
2707  * address case above.
2708  *
2709  * - A 16-byte sockaddr_in to validate the local IPv4 address and also
2710  * use it for the inbound fanout of packets.
2711  *
2712  * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also
2713  * use it for the inbound fanout of packets.
2714  *
2715  * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout
2716  * information consisting of local and remote addresses
2717  * and ports.  In this case, the addresses are both
2718  * validated as appropriate for this operation, and, if
2719  * so, the information is retained for use in the
2720  * inbound fanout.
2721  *
2722  * - A 36-byte address address (ipa6_conn_t) containing complete IPv6
2723  * fanout information, like the 12-byte case above.
2724  *
2725  * IP will also fill in the IRE request mblk with information
2726  * regarding our peer.  In all cases, we notify IP of our protocol
2727  * type by appending a single protocol byte to the bind request.
2728  */
2729 static mblk_t *
2730 udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, t_scalar_t addr_length)
2731 {
2732 	char	*cp;
2733 	mblk_t	*mp;
2734 	struct T_bind_req *tbr;
2735 	ipa_conn_t	*ac;
2736 	ipa6_conn_t	*ac6;
2737 	sin_t		*sin;
2738 	sin6_t		*sin6;
2739 
2740 	ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ);
2741 
2742 	mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI);
2743 	if (!mp)
2744 		return (mp);
2745 	mp->b_datap->db_type = M_PROTO;
2746 	tbr = (struct T_bind_req *)mp->b_rptr;
2747 	tbr->PRIM_type = bind_prim;
2748 	tbr->ADDR_offset = sizeof (*tbr);
2749 	tbr->CONIND_number = 0;
2750 	tbr->ADDR_length = addr_length;
2751 	cp = (char *)&tbr[1];
2752 	switch (addr_length) {
2753 	case sizeof (ipa_conn_t):
2754 		ASSERT(udp->udp_family == AF_INET);
2755 		/* Append a request for an IRE */
2756 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2757 		if (!mp->b_cont) {
2758 			freemsg(mp);
2759 			return (NULL);
2760 		}
2761 		mp->b_cont->b_wptr += sizeof (ire_t);
2762 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2763 
2764 		/* cp known to be 32 bit aligned */
2765 		ac = (ipa_conn_t *)cp;
2766 		ac->ac_laddr = V4_PART_OF_V6(udp->udp_v6src);
2767 		ac->ac_faddr = V4_PART_OF_V6(udp->udp_v6dst);
2768 		ac->ac_fport = udp->udp_dstport;
2769 		ac->ac_lport = udp->udp_port;
2770 		break;
2771 
2772 	case sizeof (ipa6_conn_t):
2773 		ASSERT(udp->udp_family == AF_INET6);
2774 		/* Append a request for an IRE */
2775 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2776 		if (!mp->b_cont) {
2777 			freemsg(mp);
2778 			return (NULL);
2779 		}
2780 		mp->b_cont->b_wptr += sizeof (ire_t);
2781 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2782 
2783 		/* cp known to be 32 bit aligned */
2784 		ac6 = (ipa6_conn_t *)cp;
2785 		ac6->ac6_laddr = udp->udp_v6src;
2786 		ac6->ac6_faddr = udp->udp_v6dst;
2787 		ac6->ac6_fport = udp->udp_dstport;
2788 		ac6->ac6_lport = udp->udp_port;
2789 		break;
2790 
2791 	case sizeof (sin_t):
2792 		ASSERT(udp->udp_family == AF_INET);
2793 		/* Append a request for an IRE */
2794 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2795 		if (!mp->b_cont) {
2796 			freemsg(mp);
2797 			return (NULL);
2798 		}
2799 		mp->b_cont->b_wptr += sizeof (ire_t);
2800 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2801 
2802 		sin = (sin_t *)cp;
2803 		*sin = sin_null;
2804 		sin->sin_family = AF_INET;
2805 		sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_bound_v6src);
2806 		sin->sin_port = udp->udp_port;
2807 		break;
2808 
2809 	case sizeof (sin6_t):
2810 		ASSERT(udp->udp_family == AF_INET6);
2811 		/* Append a request for an IRE */
2812 		mp->b_cont = allocb(sizeof (ire_t), BPRI_HI);
2813 		if (!mp->b_cont) {
2814 			freemsg(mp);
2815 			return (NULL);
2816 		}
2817 		mp->b_cont->b_wptr += sizeof (ire_t);
2818 		mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE;
2819 
2820 		sin6 = (sin6_t *)cp;
2821 		*sin6 = sin6_null;
2822 		sin6->sin6_family = AF_INET6;
2823 		sin6->sin6_addr = udp->udp_bound_v6src;
2824 		sin6->sin6_port = udp->udp_port;
2825 		break;
2826 	}
2827 	/* Add protocol number to end */
2828 	cp[addr_length] = (char)IPPROTO_UDP;
2829 	mp->b_wptr = (uchar_t *)&cp[addr_length + 1];
2830 	return (mp);
2831 }
2832 
2833 /*
2834  * This is the open routine for udp.  It allocates a udp_t structure for
2835  * the stream and, on the first open of the module, creates an ND table.
2836  */
2837 /* ARGSUSED */
2838 static int
2839 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
2840 {
2841 	int	err;
2842 	udp_t	*udp;
2843 	conn_t *connp;
2844 	queue_t	*ip_wq;
2845 	zoneid_t zoneid;
2846 	netstack_t *ns;
2847 	udp_stack_t *us;
2848 
2849 	TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q);
2850 
2851 	/* If the stream is already open, return immediately. */
2852 	if (q->q_ptr != NULL)
2853 		return (0);
2854 
2855 	/* If this is not a push of udp as a module, fail. */
2856 	if (sflag != MODOPEN)
2857 		return (EINVAL);
2858 
2859 	ns = netstack_find_by_cred(credp);
2860 	ASSERT(ns != NULL);
2861 	us = ns->netstack_udp;
2862 	ASSERT(us != NULL);
2863 
2864 	/*
2865 	 * For exclusive stacks we set the zoneid to zero
2866 	 * to make UDP operate as if in the global zone.
2867 	 */
2868 	if (us->us_netstack->netstack_stackid != GLOBAL_NETSTACKID)
2869 		zoneid = GLOBAL_ZONEID;
2870 	else
2871 		zoneid = crgetzoneid(credp);
2872 
2873 	q->q_hiwat = us->us_recv_hiwat;
2874 	WR(q)->q_hiwat = us->us_xmit_hiwat;
2875 	WR(q)->q_lowat = us->us_xmit_lowat;
2876 
2877 	/* Insert ourselves in the stream since we're about to walk q_next */
2878 	qprocson(q);
2879 
2880 	udp = kmem_cache_alloc(udp_cache, KM_SLEEP);
2881 	bzero(udp, sizeof (*udp));
2882 
2883 	/*
2884 	 * UDP is supported only as a module and it has to be pushed directly
2885 	 * above the device instance of IP. If UDP is pushed anywhere else
2886 	 * on a stream, it will support just T_SVR4_OPTMGMT_REQ for the
2887 	 * sake of MIB browsers and fail everything else.
2888 	 */
2889 	ip_wq = WR(q)->q_next;
2890 	if (NOT_OVER_IP(ip_wq)) {
2891 		/* Support just SNMP for MIB browsers */
2892 		connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP,
2893 		    us->us_netstack);
2894 		connp->conn_rq = q;
2895 		connp->conn_wq = WR(q);
2896 		connp->conn_flags |= IPCL_UDPMOD;
2897 		connp->conn_cred = credp;
2898 		connp->conn_zoneid = zoneid;
2899 		connp->conn_udp = udp;
2900 		udp->udp_us = us;
2901 		udp->udp_connp = connp;
2902 		q->q_ptr = WR(q)->q_ptr = connp;
2903 		crhold(credp);
2904 		q->q_qinfo = &udp_snmp_rinit;
2905 		WR(q)->q_qinfo = &udp_snmp_winit;
2906 		return (0);
2907 	}
2908 
2909 	/*
2910 	 * Initialize the udp_t structure for this stream.
2911 	 */
2912 	q = RD(ip_wq);
2913 	connp = Q_TO_CONN(q);
2914 	mutex_enter(&connp->conn_lock);
2915 	connp->conn_proto = IPPROTO_UDP;
2916 	connp->conn_flags |= IPCL_UDP;
2917 	connp->conn_sqp = IP_SQUEUE_GET(lbolt);
2918 	connp->conn_udp = udp;
2919 
2920 	/* Set the initial state of the stream and the privilege status. */
2921 	udp->udp_connp = connp;
2922 	udp->udp_state = TS_UNBND;
2923 	udp->udp_mode = UDP_MT_HOT;
2924 	if (getmajor(*devp) == (major_t)UDP6_MAJ) {
2925 		udp->udp_family = AF_INET6;
2926 		udp->udp_ipversion = IPV6_VERSION;
2927 		udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
2928 		udp->udp_ttl = us->us_ipv6_hoplimit;
2929 		connp->conn_af_isv6 = B_TRUE;
2930 		connp->conn_flags |= IPCL_ISV6;
2931 	} else {
2932 		udp->udp_family = AF_INET;
2933 		udp->udp_ipversion = IPV4_VERSION;
2934 		udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE;
2935 		udp->udp_ttl = us->us_ipv4_ttl;
2936 		connp->conn_af_isv6 = B_FALSE;
2937 		connp->conn_flags &= ~IPCL_ISV6;
2938 	}
2939 
2940 	udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL;
2941 	connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP;
2942 	connp->conn_zoneid = zoneid;
2943 
2944 	udp->udp_open_time = lbolt64;
2945 	udp->udp_open_pid = curproc->p_pid;
2946 
2947 	/*
2948 	 * If the caller has the process-wide flag set, then default to MAC
2949 	 * exempt mode.  This allows read-down to unlabeled hosts.
2950 	 */
2951 	if (getpflags(NET_MAC_AWARE, credp) != 0)
2952 		udp->udp_mac_exempt = B_TRUE;
2953 
2954 	if (connp->conn_flags & IPCL_SOCKET) {
2955 		udp->udp_issocket = B_TRUE;
2956 		udp->udp_direct_sockfs = B_TRUE;
2957 	}
2958 
2959 	connp->conn_ulp_labeled = is_system_labeled();
2960 
2961 	mutex_exit(&connp->conn_lock);
2962 	udp->udp_us = us;
2963 
2964 	/*
2965 	 * The transmit hiwat/lowat is only looked at on IP's queue.
2966 	 * Store in q_hiwat in order to return on SO_SNDBUF/SO_RCVBUF
2967 	 * getsockopts.
2968 	 */
2969 	q->q_hiwat = us->us_recv_hiwat;
2970 	WR(q)->q_hiwat = us->us_xmit_hiwat;
2971 	WR(q)->q_lowat = us->us_xmit_lowat;
2972 
2973 	if (udp->udp_family == AF_INET6) {
2974 		/* Build initial header template for transmit */
2975 		if ((err = udp_build_hdrs(q, udp)) != 0) {
2976 			/* XXX missing free of connp? crfree? netstack_rele? */
2977 			qprocsoff(UDP_RD(q));
2978 			udp->udp_connp = NULL;
2979 			connp->conn_udp = NULL;
2980 			kmem_cache_free(udp_cache, udp);
2981 			return (err);
2982 		}
2983 	}
2984 
2985 	/* Set the Stream head write offset and high watermark. */
2986 	(void) mi_set_sth_wroff(UDP_RD(q),
2987 	    udp->udp_max_hdr_len + us->us_wroff_extra);
2988 	(void) mi_set_sth_hiwat(UDP_RD(q), udp_set_rcv_hiwat(udp, q->q_hiwat));
2989 
2990 	return (0);
2991 }
2992 
2993 /*
2994  * Which UDP options OK to set through T_UNITDATA_REQ...
2995  */
2996 /* ARGSUSED */
2997 static boolean_t
2998 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name)
2999 {
3000 	return (B_TRUE);
3001 }
3002 
3003 /*
3004  * This routine gets default values of certain options whose default
3005  * values are maintained by protcol specific code
3006  */
3007 /* ARGSUSED */
3008 int
3009 udp_opt_default(queue_t	*q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
3010 {
3011 	udp_t		*udp = Q_TO_UDP(q);
3012 	udp_stack_t *us = udp->udp_us;
3013 	int *i1 = (int *)ptr;
3014 
3015 	switch (level) {
3016 	case IPPROTO_IP:
3017 		switch (name) {
3018 		case IP_MULTICAST_TTL:
3019 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL;
3020 			return (sizeof (uchar_t));
3021 		case IP_MULTICAST_LOOP:
3022 			*ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP;
3023 			return (sizeof (uchar_t));
3024 		}
3025 		break;
3026 	case IPPROTO_IPV6:
3027 		switch (name) {
3028 		case IPV6_MULTICAST_HOPS:
3029 			*i1 = IP_DEFAULT_MULTICAST_TTL;
3030 			return (sizeof (int));
3031 		case IPV6_MULTICAST_LOOP:
3032 			*i1 = IP_DEFAULT_MULTICAST_LOOP;
3033 			return (sizeof (int));
3034 		case IPV6_UNICAST_HOPS:
3035 			*i1 = us->us_ipv6_hoplimit;
3036 			return (sizeof (int));
3037 		}
3038 		break;
3039 	}
3040 	return (-1);
3041 }
3042 
3043 /*
3044  * This routine retrieves the current status of socket options
3045  * and expects the caller to pass in the queue pointer of the
3046  * upper instance.  It returns the size of the option retrieved.
3047  */
3048 int
3049 udp_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr)
3050 {
3051 	int	*i1 = (int *)ptr;
3052 	conn_t	*connp;
3053 	udp_t	*udp;
3054 	ip6_pkt_t *ipp;
3055 	int	len;
3056 	udp_stack_t	*us;
3057 
3058 	q = UDP_WR(q);
3059 	connp = Q_TO_CONN(q);
3060 	udp = connp->conn_udp;
3061 	ipp = &udp->udp_sticky_ipp;
3062 	us = udp->udp_us;
3063 
3064 	switch (level) {
3065 	case SOL_SOCKET:
3066 		switch (name) {
3067 		case SO_DEBUG:
3068 			*i1 = udp->udp_debug;
3069 			break;	/* goto sizeof (int) option return */
3070 		case SO_REUSEADDR:
3071 			*i1 = udp->udp_reuseaddr;
3072 			break;	/* goto sizeof (int) option return */
3073 		case SO_TYPE:
3074 			*i1 = SOCK_DGRAM;
3075 			break;	/* goto sizeof (int) option return */
3076 
3077 		/*
3078 		 * The following three items are available here,
3079 		 * but are only meaningful to IP.
3080 		 */
3081 		case SO_DONTROUTE:
3082 			*i1 = udp->udp_dontroute;
3083 			break;	/* goto sizeof (int) option return */
3084 		case SO_USELOOPBACK:
3085 			*i1 = udp->udp_useloopback;
3086 			break;	/* goto sizeof (int) option return */
3087 		case SO_BROADCAST:
3088 			*i1 = udp->udp_broadcast;
3089 			break;	/* goto sizeof (int) option return */
3090 
3091 		case SO_SNDBUF:
3092 			*i1 = q->q_hiwat;
3093 			break;	/* goto sizeof (int) option return */
3094 		case SO_RCVBUF:
3095 			*i1 = RD(q)->q_hiwat;
3096 			break;	/* goto sizeof (int) option return */
3097 		case SO_DGRAM_ERRIND:
3098 			*i1 = udp->udp_dgram_errind;
3099 			break;	/* goto sizeof (int) option return */
3100 		case SO_RECVUCRED:
3101 			*i1 = udp->udp_recvucred;
3102 			break;	/* goto sizeof (int) option return */
3103 		case SO_TIMESTAMP:
3104 			*i1 = udp->udp_timestamp;
3105 			break;	/* goto sizeof (int) option return */
3106 		case SO_ANON_MLP:
3107 			*i1 = udp->udp_anon_mlp;
3108 			break;	/* goto sizeof (int) option return */
3109 		case SO_MAC_EXEMPT:
3110 			*i1 = udp->udp_mac_exempt;
3111 			break;	/* goto sizeof (int) option return */
3112 		case SO_ALLZONES:
3113 			*i1 = connp->conn_allzones;
3114 			break;	/* goto sizeof (int) option return */
3115 		case SO_EXCLBIND:
3116 			*i1 = udp->udp_exclbind ? SO_EXCLBIND : 0;
3117 			break;
3118 		case SO_PROTOTYPE:
3119 			*i1 = IPPROTO_UDP;
3120 			break;
3121 		case SO_DOMAIN:
3122 			*i1 = udp->udp_family;
3123 			break;
3124 		default:
3125 			return (-1);
3126 		}
3127 		break;
3128 	case IPPROTO_IP:
3129 		if (udp->udp_family != AF_INET)
3130 			return (-1);
3131 		switch (name) {
3132 		case IP_OPTIONS:
3133 		case T_IP_OPTIONS:
3134 			len = udp->udp_ip_rcv_options_len - udp->udp_label_len;
3135 			if (len > 0) {
3136 				bcopy(udp->udp_ip_rcv_options +
3137 				    udp->udp_label_len, ptr, len);
3138 			}
3139 			return (len);
3140 		case IP_TOS:
3141 		case T_IP_TOS:
3142 			*i1 = (int)udp->udp_type_of_service;
3143 			break;	/* goto sizeof (int) option return */
3144 		case IP_TTL:
3145 			*i1 = (int)udp->udp_ttl;
3146 			break;	/* goto sizeof (int) option return */
3147 		case IP_NEXTHOP:
3148 		case IP_RECVPKTINFO:
3149 			/*
3150 			 * This also handles IP_PKTINFO.
3151 			 * IP_PKTINFO and IP_RECVPKTINFO have the same value.
3152 			 * Differentiation is based on the size of the argument
3153 			 * passed in.
3154 			 * This option is handled in IP which will return an
3155 			 * error for IP_PKTINFO as it's not supported as a
3156 			 * sticky option.
3157 			 */
3158 			return (-EINVAL);
3159 		case IP_MULTICAST_IF:
3160 			/* 0 address if not set */
3161 			*(ipaddr_t *)ptr = udp->udp_multicast_if_addr;
3162 			return (sizeof (ipaddr_t));
3163 		case IP_MULTICAST_TTL:
3164 			*(uchar_t *)ptr = udp->udp_multicast_ttl;
3165 			return (sizeof (uchar_t));
3166 		case IP_MULTICAST_LOOP:
3167 			*ptr = connp->conn_multicast_loop;
3168 			return (sizeof (uint8_t));
3169 		case IP_RECVOPTS:
3170 			*i1 = udp->udp_recvopts;
3171 			break;	/* goto sizeof (int) option return */
3172 		case IP_RECVDSTADDR:
3173 			*i1 = udp->udp_recvdstaddr;
3174 			break;	/* goto sizeof (int) option return */
3175 		case IP_RECVIF:
3176 			*i1 = udp->udp_recvif;
3177 			break;	/* goto sizeof (int) option return */
3178 		case IP_RECVSLLA:
3179 			*i1 = udp->udp_recvslla;
3180 			break;	/* goto sizeof (int) option return */
3181 		case IP_RECVTTL:
3182 			*i1 = udp->udp_recvttl;
3183 			break;	/* goto sizeof (int) option return */
3184 		case IP_ADD_MEMBERSHIP:
3185 		case IP_DROP_MEMBERSHIP:
3186 		case IP_BLOCK_SOURCE:
3187 		case IP_UNBLOCK_SOURCE:
3188 		case IP_ADD_SOURCE_MEMBERSHIP:
3189 		case IP_DROP_SOURCE_MEMBERSHIP:
3190 		case MCAST_JOIN_GROUP:
3191 		case MCAST_LEAVE_GROUP:
3192 		case MCAST_BLOCK_SOURCE:
3193 		case MCAST_UNBLOCK_SOURCE:
3194 		case MCAST_JOIN_SOURCE_GROUP:
3195 		case MCAST_LEAVE_SOURCE_GROUP:
3196 		case IP_DONTFAILOVER_IF:
3197 			/* cannot "get" the value for these */
3198 			return (-1);
3199 		case IP_BOUND_IF:
3200 			/* Zero if not set */
3201 			*i1 = udp->udp_bound_if;
3202 			break;	/* goto sizeof (int) option return */
3203 		case IP_UNSPEC_SRC:
3204 			*i1 = udp->udp_unspec_source;
3205 			break;	/* goto sizeof (int) option return */
3206 		case IP_XMIT_IF:
3207 			*i1 = udp->udp_xmit_if;
3208 			break; /* goto sizeof (int) option return */
3209 		default:
3210 			return (-1);
3211 		}
3212 		break;
3213 	case IPPROTO_IPV6:
3214 		if (udp->udp_family != AF_INET6)
3215 			return (-1);
3216 		switch (name) {
3217 		case IPV6_UNICAST_HOPS:
3218 			*i1 = (unsigned int)udp->udp_ttl;
3219 			break;	/* goto sizeof (int) option return */
3220 		case IPV6_MULTICAST_IF:
3221 			/* 0 index if not set */
3222 			*i1 = udp->udp_multicast_if_index;
3223 			break;	/* goto sizeof (int) option return */
3224 		case IPV6_MULTICAST_HOPS:
3225 			*i1 = udp->udp_multicast_ttl;
3226 			break;	/* goto sizeof (int) option return */
3227 		case IPV6_MULTICAST_LOOP:
3228 			*i1 = connp->conn_multicast_loop;
3229 			break;	/* goto sizeof (int) option return */
3230 		case IPV6_JOIN_GROUP:
3231 		case IPV6_LEAVE_GROUP:
3232 		case MCAST_JOIN_GROUP:
3233 		case MCAST_LEAVE_GROUP:
3234 		case MCAST_BLOCK_SOURCE:
3235 		case MCAST_UNBLOCK_SOURCE:
3236 		case MCAST_JOIN_SOURCE_GROUP:
3237 		case MCAST_LEAVE_SOURCE_GROUP:
3238 			/* cannot "get" the value for these */
3239 			return (-1);
3240 		case IPV6_BOUND_IF:
3241 			/* Zero if not set */
3242 			*i1 = udp->udp_bound_if;
3243 			break;	/* goto sizeof (int) option return */
3244 		case IPV6_UNSPEC_SRC:
3245 			*i1 = udp->udp_unspec_source;
3246 			break;	/* goto sizeof (int) option return */
3247 		case IPV6_RECVPKTINFO:
3248 			*i1 = udp->udp_ip_recvpktinfo;
3249 			break;	/* goto sizeof (int) option return */
3250 		case IPV6_RECVTCLASS:
3251 			*i1 = udp->udp_ipv6_recvtclass;
3252 			break;	/* goto sizeof (int) option return */
3253 		case IPV6_RECVPATHMTU:
3254 			*i1 = udp->udp_ipv6_recvpathmtu;
3255 			break;	/* goto sizeof (int) option return */
3256 		case IPV6_RECVHOPLIMIT:
3257 			*i1 = udp->udp_ipv6_recvhoplimit;
3258 			break;	/* goto sizeof (int) option return */
3259 		case IPV6_RECVHOPOPTS:
3260 			*i1 = udp->udp_ipv6_recvhopopts;
3261 			break;	/* goto sizeof (int) option return */
3262 		case IPV6_RECVDSTOPTS:
3263 			*i1 = udp->udp_ipv6_recvdstopts;
3264 			break;	/* goto sizeof (int) option return */
3265 		case _OLD_IPV6_RECVDSTOPTS:
3266 			*i1 = udp->udp_old_ipv6_recvdstopts;
3267 			break;	/* goto sizeof (int) option return */
3268 		case IPV6_RECVRTHDRDSTOPTS:
3269 			*i1 = udp->udp_ipv6_recvrthdrdstopts;
3270 			break;	/* goto sizeof (int) option return */
3271 		case IPV6_RECVRTHDR:
3272 			*i1 = udp->udp_ipv6_recvrthdr;
3273 			break;	/* goto sizeof (int) option return */
3274 		case IPV6_PKTINFO: {
3275 			/* XXX assumes that caller has room for max size! */
3276 			struct in6_pktinfo *pkti;
3277 
3278 			pkti = (struct in6_pktinfo *)ptr;
3279 			if (ipp->ipp_fields & IPPF_IFINDEX)
3280 				pkti->ipi6_ifindex = ipp->ipp_ifindex;
3281 			else
3282 				pkti->ipi6_ifindex = 0;
3283 			if (ipp->ipp_fields & IPPF_ADDR)
3284 				pkti->ipi6_addr = ipp->ipp_addr;
3285 			else
3286 				pkti->ipi6_addr = ipv6_all_zeros;
3287 			return (sizeof (struct in6_pktinfo));
3288 		}
3289 		case IPV6_TCLASS:
3290 			if (ipp->ipp_fields & IPPF_TCLASS)
3291 				*i1 = ipp->ipp_tclass;
3292 			else
3293 				*i1 = IPV6_FLOW_TCLASS(
3294 				    IPV6_DEFAULT_VERS_AND_FLOW);
3295 			break;	/* goto sizeof (int) option return */
3296 		case IPV6_NEXTHOP: {
3297 			sin6_t *sin6 = (sin6_t *)ptr;
3298 
3299 			if (!(ipp->ipp_fields & IPPF_NEXTHOP))
3300 				return (0);
3301 			*sin6 = sin6_null;
3302 			sin6->sin6_family = AF_INET6;
3303 			sin6->sin6_addr = ipp->ipp_nexthop;
3304 			return (sizeof (sin6_t));
3305 		}
3306 		case IPV6_HOPOPTS:
3307 			if (!(ipp->ipp_fields & IPPF_HOPOPTS))
3308 				return (0);
3309 			if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6)
3310 				return (0);
3311 			/*
3312 			 * The cipso/label option is added by kernel.
3313 			 * User is not usually aware of this option.
3314 			 * We copy out the hbh opt after the label option.
3315 			 */
3316 			bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6,
3317 			    ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6);
3318 			if (udp->udp_label_len_v6 > 0) {
3319 				ptr[0] = ((char *)ipp->ipp_hopopts)[0];
3320 				ptr[1] = (ipp->ipp_hopoptslen -
3321 				    udp->udp_label_len_v6 + 7) / 8 - 1;
3322 			}
3323 			return (ipp->ipp_hopoptslen - udp->udp_label_len_v6);
3324 		case IPV6_RTHDRDSTOPTS:
3325 			if (!(ipp->ipp_fields & IPPF_RTDSTOPTS))
3326 				return (0);
3327 			bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen);
3328 			return (ipp->ipp_rtdstoptslen);
3329 		case IPV6_RTHDR:
3330 			if (!(ipp->ipp_fields & IPPF_RTHDR))
3331 				return (0);
3332 			bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen);
3333 			return (ipp->ipp_rthdrlen);
3334 		case IPV6_DSTOPTS:
3335 			if (!(ipp->ipp_fields & IPPF_DSTOPTS))
3336 				return (0);
3337 			bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen);
3338 			return (ipp->ipp_dstoptslen);
3339 		case IPV6_PATHMTU:
3340 			return (ip_fill_mtuinfo(&udp->udp_v6dst,
3341 			    udp->udp_dstport, (struct ip6_mtuinfo *)ptr,
3342 			    us->us_netstack));
3343 		default:
3344 			return (-1);
3345 		}
3346 		break;
3347 	case IPPROTO_UDP:
3348 		switch (name) {
3349 		case UDP_ANONPRIVBIND:
3350 			*i1 = udp->udp_anon_priv_bind;
3351 			break;
3352 		case UDP_EXCLBIND:
3353 			*i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0;
3354 			break;
3355 		case UDP_RCVHDR:
3356 			*i1 = udp->udp_rcvhdr ? 1 : 0;
3357 			break;
3358 		case UDP_NAT_T_ENDPOINT:
3359 			*i1 = udp->udp_nat_t_endpoint;
3360 			break;
3361 		default:
3362 			return (-1);
3363 		}
3364 		break;
3365 	default:
3366 		return (-1);
3367 	}
3368 	return (sizeof (int));
3369 }
3370 
3371 /*
3372  * This routine sets socket options; it expects the caller
3373  * to pass in the queue pointer of the upper instance.
3374  */
3375 /* ARGSUSED */
3376 int
3377 udp_opt_set(queue_t *q, uint_t optset_context, int level,
3378     int name, uint_t inlen, uchar_t *invalp, uint_t *outlenp,
3379     uchar_t *outvalp, void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3380 {
3381 	udpattrs_t *attrs = thisdg_attrs;
3382 	int	*i1 = (int *)invalp;
3383 	boolean_t onoff = (*i1 == 0) ? 0 : 1;
3384 	boolean_t checkonly;
3385 	int	error;
3386 	conn_t	*connp;
3387 	udp_t	*udp;
3388 	uint_t	newlen;
3389 	udp_stack_t *us;
3390 
3391 	q = UDP_WR(q);
3392 	connp = Q_TO_CONN(q);
3393 	udp = connp->conn_udp;
3394 	us = udp->udp_us;
3395 
3396 	switch (optset_context) {
3397 	case SETFN_OPTCOM_CHECKONLY:
3398 		checkonly = B_TRUE;
3399 		/*
3400 		 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ
3401 		 * inlen != 0 implies value supplied and
3402 		 * 	we have to "pretend" to set it.
3403 		 * inlen == 0 implies that there is no
3404 		 * 	value part in T_CHECK request and just validation
3405 		 * done elsewhere should be enough, we just return here.
3406 		 */
3407 		if (inlen == 0) {
3408 			*outlenp = 0;
3409 			return (0);
3410 		}
3411 		break;
3412 	case SETFN_OPTCOM_NEGOTIATE:
3413 		checkonly = B_FALSE;
3414 		break;
3415 	case SETFN_UD_NEGOTIATE:
3416 	case SETFN_CONN_NEGOTIATE:
3417 		checkonly = B_FALSE;
3418 		/*
3419 		 * Negotiating local and "association-related" options
3420 		 * through T_UNITDATA_REQ.
3421 		 *
3422 		 * Following routine can filter out ones we do not
3423 		 * want to be "set" this way.
3424 		 */
3425 		if (!udp_opt_allow_udr_set(level, name)) {
3426 			*outlenp = 0;
3427 			return (EINVAL);
3428 		}
3429 		break;
3430 	default:
3431 		/*
3432 		 * We should never get here
3433 		 */
3434 		*outlenp = 0;
3435 		return (EINVAL);
3436 	}
3437 
3438 	ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) ||
3439 	    (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0));
3440 
3441 	/*
3442 	 * For fixed length options, no sanity check
3443 	 * of passed in length is done. It is assumed *_optcom_req()
3444 	 * routines do the right thing.
3445 	 */
3446 
3447 	switch (level) {
3448 	case SOL_SOCKET:
3449 		switch (name) {
3450 		case SO_REUSEADDR:
3451 			if (!checkonly)
3452 				udp->udp_reuseaddr = onoff;
3453 			break;
3454 		case SO_DEBUG:
3455 			if (!checkonly)
3456 				udp->udp_debug = onoff;
3457 			break;
3458 		/*
3459 		 * The following three items are available here,
3460 		 * but are only meaningful to IP.
3461 		 */
3462 		case SO_DONTROUTE:
3463 			if (!checkonly)
3464 				udp->udp_dontroute = onoff;
3465 			break;
3466 		case SO_USELOOPBACK:
3467 			if (!checkonly)
3468 				udp->udp_useloopback = onoff;
3469 			break;
3470 		case SO_BROADCAST:
3471 			if (!checkonly)
3472 				udp->udp_broadcast = onoff;
3473 			break;
3474 
3475 		case SO_SNDBUF:
3476 			if (*i1 > us->us_max_buf) {
3477 				*outlenp = 0;
3478 				return (ENOBUFS);
3479 			}
3480 			if (!checkonly) {
3481 				q->q_hiwat = *i1;
3482 				WR(UDP_RD(q))->q_hiwat = *i1;
3483 			}
3484 			break;
3485 		case SO_RCVBUF:
3486 			if (*i1 > us->us_max_buf) {
3487 				*outlenp = 0;
3488 				return (ENOBUFS);
3489 			}
3490 			if (!checkonly) {
3491 				RD(q)->q_hiwat = *i1;
3492 				UDP_RD(q)->q_hiwat = *i1;
3493 				(void) mi_set_sth_hiwat(UDP_RD(q),
3494 				    udp_set_rcv_hiwat(udp, *i1));
3495 			}
3496 			break;
3497 		case SO_DGRAM_ERRIND:
3498 			if (!checkonly)
3499 				udp->udp_dgram_errind = onoff;
3500 			break;
3501 		case SO_RECVUCRED:
3502 			if (!checkonly)
3503 				udp->udp_recvucred = onoff;
3504 			break;
3505 		case SO_ALLZONES:
3506 			/*
3507 			 * "soft" error (negative)
3508 			 * option not handled at this level
3509 			 * Do not modify *outlenp.
3510 			 */
3511 			return (-EINVAL);
3512 		case SO_TIMESTAMP:
3513 			if (!checkonly)
3514 				udp->udp_timestamp = onoff;
3515 			break;
3516 		case SO_ANON_MLP:
3517 			if (!checkonly)
3518 				udp->udp_anon_mlp = onoff;
3519 			break;
3520 		case SO_MAC_EXEMPT:
3521 			if (secpolicy_net_mac_aware(cr) != 0 ||
3522 			    udp->udp_state != TS_UNBND)
3523 				return (EACCES);
3524 			if (!checkonly)
3525 				udp->udp_mac_exempt = onoff;
3526 			break;
3527 		case SCM_UCRED: {
3528 			struct ucred_s *ucr;
3529 			cred_t *cr, *newcr;
3530 			ts_label_t *tsl;
3531 
3532 			/*
3533 			 * Only sockets that have proper privileges and are
3534 			 * bound to MLPs will have any other value here, so
3535 			 * this implicitly tests for privilege to set label.
3536 			 */
3537 			if (connp->conn_mlp_type == mlptSingle)
3538 				break;
3539 			ucr = (struct ucred_s *)invalp;
3540 			if (inlen != ucredsize ||
3541 			    ucr->uc_labeloff < sizeof (*ucr) ||
3542 			    ucr->uc_labeloff + sizeof (bslabel_t) > inlen)
3543 				return (EINVAL);
3544 			if (!checkonly) {
3545 				mblk_t *mb;
3546 
3547 				if (attrs == NULL ||
3548 				    (mb = attrs->udpattr_mb) == NULL)
3549 					return (EINVAL);
3550 				if ((cr = DB_CRED(mb)) == NULL)
3551 					cr = udp->udp_connp->conn_cred;
3552 				ASSERT(cr != NULL);
3553 				if ((tsl = crgetlabel(cr)) == NULL)
3554 					return (EINVAL);
3555 				newcr = copycred_from_bslabel(cr, UCLABEL(ucr),
3556 				    tsl->tsl_doi, KM_NOSLEEP);
3557 				if (newcr == NULL)
3558 					return (ENOSR);
3559 				mblk_setcred(mb, newcr);
3560 				attrs->udpattr_credset = B_TRUE;
3561 				crfree(newcr);
3562 			}
3563 			break;
3564 		}
3565 		case SO_EXCLBIND:
3566 			if (!checkonly)
3567 				udp->udp_exclbind = onoff;
3568 			break;
3569 		default:
3570 			*outlenp = 0;
3571 			return (EINVAL);
3572 		}
3573 		break;
3574 	case IPPROTO_IP:
3575 		if (udp->udp_family != AF_INET) {
3576 			*outlenp = 0;
3577 			return (ENOPROTOOPT);
3578 		}
3579 		switch (name) {
3580 		case IP_OPTIONS:
3581 		case T_IP_OPTIONS:
3582 			/* Save options for use by IP. */
3583 			newlen = inlen + udp->udp_label_len;
3584 			if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) {
3585 				*outlenp = 0;
3586 				return (EINVAL);
3587 			}
3588 			if (checkonly)
3589 				break;
3590 
3591 			if (!tsol_option_set(&udp->udp_ip_snd_options,
3592 			    &udp->udp_ip_snd_options_len,
3593 			    udp->udp_label_len, invalp, inlen)) {
3594 				*outlenp = 0;
3595 				return (ENOMEM);
3596 			}
3597 
3598 			udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
3599 			    UDPH_SIZE + udp->udp_ip_snd_options_len;
3600 			(void) mi_set_sth_wroff(RD(q), udp->udp_max_hdr_len +
3601 			    us->us_wroff_extra);
3602 			break;
3603 
3604 		case IP_TTL:
3605 			if (!checkonly) {
3606 				udp->udp_ttl = (uchar_t)*i1;
3607 			}
3608 			break;
3609 		case IP_TOS:
3610 		case T_IP_TOS:
3611 			if (!checkonly) {
3612 				udp->udp_type_of_service = (uchar_t)*i1;
3613 			}
3614 			break;
3615 		case IP_MULTICAST_IF: {
3616 			/*
3617 			 * TODO should check OPTMGMT reply and undo this if
3618 			 * there is an error.
3619 			 */
3620 			struct in_addr *inap = (struct in_addr *)invalp;
3621 			if (!checkonly) {
3622 				udp->udp_multicast_if_addr =
3623 				    inap->s_addr;
3624 			}
3625 			break;
3626 		}
3627 		case IP_MULTICAST_TTL:
3628 			if (!checkonly)
3629 				udp->udp_multicast_ttl = *invalp;
3630 			break;
3631 		case IP_MULTICAST_LOOP:
3632 			if (!checkonly)
3633 				connp->conn_multicast_loop = *invalp;
3634 			break;
3635 		case IP_RECVOPTS:
3636 			if (!checkonly)
3637 				udp->udp_recvopts = onoff;
3638 			break;
3639 		case IP_RECVDSTADDR:
3640 			if (!checkonly)
3641 				udp->udp_recvdstaddr = onoff;
3642 			break;
3643 		case IP_RECVIF:
3644 			if (!checkonly)
3645 				udp->udp_recvif = onoff;
3646 			break;
3647 		case IP_RECVSLLA:
3648 			if (!checkonly)
3649 				udp->udp_recvslla = onoff;
3650 			break;
3651 		case IP_RECVTTL:
3652 			if (!checkonly)
3653 				udp->udp_recvttl = onoff;
3654 			break;
3655 		case IP_PKTINFO: {
3656 			/*
3657 			 * This also handles IP_RECVPKTINFO.
3658 			 * IP_PKTINFO and IP_RECVPKTINFO have same value.
3659 			 * Differentiation is based on the size of the
3660 			 * argument passed in.
3661 			 */
3662 			struct in_pktinfo *pktinfop;
3663 			ip4_pkt_t *attr_pktinfop;
3664 
3665 			if (checkonly)
3666 				break;
3667 
3668 			if (inlen == sizeof (int)) {
3669 				/*
3670 				 * This is IP_RECVPKTINFO option.
3671 				 * Keep a local copy of whether this option is
3672 				 * set or not and pass it down to IP for
3673 				 * processing.
3674 				 */
3675 
3676 				udp->udp_ip_recvpktinfo = onoff;
3677 				return (-EINVAL);
3678 			}
3679 
3680 			if (attrs == NULL ||
3681 			    (attr_pktinfop = attrs->udpattr_ipp4) == NULL) {
3682 				/*
3683 				 * sticky option or no buffer to return
3684 				 * the results.
3685 				 */
3686 				return (EINVAL);
3687 			}
3688 
3689 			if (inlen != sizeof (struct in_pktinfo))
3690 				return (EINVAL);
3691 
3692 			pktinfop = (struct in_pktinfo *)invalp;
3693 
3694 			/*
3695 			 * At least one of the values should be specified
3696 			 */
3697 			if (pktinfop->ipi_ifindex == 0 &&
3698 			    pktinfop->ipi_spec_dst.s_addr == INADDR_ANY) {
3699 				return (EINVAL);
3700 			}
3701 
3702 			attr_pktinfop->ip4_addr = pktinfop->ipi_spec_dst.s_addr;
3703 			attr_pktinfop->ip4_ill_index = pktinfop->ipi_ifindex;
3704 
3705 			break;
3706 		}
3707 		case IP_ADD_MEMBERSHIP:
3708 		case IP_DROP_MEMBERSHIP:
3709 		case IP_BLOCK_SOURCE:
3710 		case IP_UNBLOCK_SOURCE:
3711 		case IP_ADD_SOURCE_MEMBERSHIP:
3712 		case IP_DROP_SOURCE_MEMBERSHIP:
3713 		case MCAST_JOIN_GROUP:
3714 		case MCAST_LEAVE_GROUP:
3715 		case MCAST_BLOCK_SOURCE:
3716 		case MCAST_UNBLOCK_SOURCE:
3717 		case MCAST_JOIN_SOURCE_GROUP:
3718 		case MCAST_LEAVE_SOURCE_GROUP:
3719 		case IP_SEC_OPT:
3720 		case IP_NEXTHOP:
3721 			/*
3722 			 * "soft" error (negative)
3723 			 * option not handled at this level
3724 			 * Do not modify *outlenp.
3725 			 */
3726 			return (-EINVAL);
3727 		case IP_BOUND_IF:
3728 			if (!checkonly)
3729 				udp->udp_bound_if = *i1;
3730 			break;
3731 		case IP_UNSPEC_SRC:
3732 			if (!checkonly)
3733 				udp->udp_unspec_source = onoff;
3734 			break;
3735 		case IP_XMIT_IF:
3736 			if (!checkonly)
3737 				udp->udp_xmit_if = *i1;
3738 			break;
3739 		default:
3740 			*outlenp = 0;
3741 			return (EINVAL);
3742 		}
3743 		break;
3744 	case IPPROTO_IPV6: {
3745 		ip6_pkt_t		*ipp;
3746 		boolean_t		sticky;
3747 
3748 		if (udp->udp_family != AF_INET6) {
3749 			*outlenp = 0;
3750 			return (ENOPROTOOPT);
3751 		}
3752 		/*
3753 		 * Deal with both sticky options and ancillary data
3754 		 */
3755 		sticky = B_FALSE;
3756 		if (attrs == NULL || (ipp = attrs->udpattr_ipp6) ==
3757 		    NULL) {
3758 			/* sticky options, or none */
3759 			ipp = &udp->udp_sticky_ipp;
3760 			sticky = B_TRUE;
3761 		}
3762 
3763 		switch (name) {
3764 		case IPV6_MULTICAST_IF:
3765 			if (!checkonly)
3766 				udp->udp_multicast_if_index = *i1;
3767 			break;
3768 		case IPV6_UNICAST_HOPS:
3769 			/* -1 means use default */
3770 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
3771 				*outlenp = 0;
3772 				return (EINVAL);
3773 			}
3774 			if (!checkonly) {
3775 				if (*i1 == -1) {
3776 					udp->udp_ttl = ipp->ipp_unicast_hops =
3777 					    us->us_ipv6_hoplimit;
3778 					ipp->ipp_fields &= ~IPPF_UNICAST_HOPS;
3779 					/* Pass modified value to IP. */
3780 					*i1 = udp->udp_ttl;
3781 				} else {
3782 					udp->udp_ttl = ipp->ipp_unicast_hops =
3783 					    (uint8_t)*i1;
3784 					ipp->ipp_fields |= IPPF_UNICAST_HOPS;
3785 				}
3786 				/* Rebuild the header template */
3787 				error = udp_build_hdrs(q, udp);
3788 				if (error != 0) {
3789 					*outlenp = 0;
3790 					return (error);
3791 				}
3792 			}
3793 			break;
3794 		case IPV6_MULTICAST_HOPS:
3795 			/* -1 means use default */
3796 			if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) {
3797 				*outlenp = 0;
3798 				return (EINVAL);
3799 			}
3800 			if (!checkonly) {
3801 				if (*i1 == -1) {
3802 					udp->udp_multicast_ttl =
3803 					    ipp->ipp_multicast_hops =
3804 					    IP_DEFAULT_MULTICAST_TTL;
3805 					ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS;
3806 					/* Pass modified value to IP. */
3807 					*i1 = udp->udp_multicast_ttl;
3808 				} else {
3809 					udp->udp_multicast_ttl =
3810 					    ipp->ipp_multicast_hops =
3811 					    (uint8_t)*i1;
3812 					ipp->ipp_fields |= IPPF_MULTICAST_HOPS;
3813 				}
3814 			}
3815 			break;
3816 		case IPV6_MULTICAST_LOOP:
3817 			if (*i1 != 0 && *i1 != 1) {
3818 				*outlenp = 0;
3819 				return (EINVAL);
3820 			}
3821 			if (!checkonly)
3822 				connp->conn_multicast_loop = *i1;
3823 			break;
3824 		case IPV6_JOIN_GROUP:
3825 		case IPV6_LEAVE_GROUP:
3826 		case MCAST_JOIN_GROUP:
3827 		case MCAST_LEAVE_GROUP:
3828 		case MCAST_BLOCK_SOURCE:
3829 		case MCAST_UNBLOCK_SOURCE:
3830 		case MCAST_JOIN_SOURCE_GROUP:
3831 		case MCAST_LEAVE_SOURCE_GROUP:
3832 			/*
3833 			 * "soft" error (negative)
3834 			 * option not handled at this level
3835 			 * Note: Do not modify *outlenp
3836 			 */
3837 			return (-EINVAL);
3838 		case IPV6_BOUND_IF:
3839 			if (!checkonly)
3840 				udp->udp_bound_if = *i1;
3841 			break;
3842 		case IPV6_UNSPEC_SRC:
3843 			if (!checkonly)
3844 				udp->udp_unspec_source = onoff;
3845 			break;
3846 		/*
3847 		 * Set boolean switches for ancillary data delivery
3848 		 */
3849 		case IPV6_RECVPKTINFO:
3850 			if (!checkonly)
3851 				udp->udp_ip_recvpktinfo = onoff;
3852 			break;
3853 		case IPV6_RECVTCLASS:
3854 			if (!checkonly) {
3855 				udp->udp_ipv6_recvtclass = onoff;
3856 			}
3857 			break;
3858 		case IPV6_RECVPATHMTU:
3859 			if (!checkonly) {
3860 				udp->udp_ipv6_recvpathmtu = onoff;
3861 			}
3862 			break;
3863 		case IPV6_RECVHOPLIMIT:
3864 			if (!checkonly)
3865 				udp->udp_ipv6_recvhoplimit = onoff;
3866 			break;
3867 		case IPV6_RECVHOPOPTS:
3868 			if (!checkonly)
3869 				udp->udp_ipv6_recvhopopts = onoff;
3870 			break;
3871 		case IPV6_RECVDSTOPTS:
3872 			if (!checkonly)
3873 				udp->udp_ipv6_recvdstopts = onoff;
3874 			break;
3875 		case _OLD_IPV6_RECVDSTOPTS:
3876 			if (!checkonly)
3877 				udp->udp_old_ipv6_recvdstopts = onoff;
3878 			break;
3879 		case IPV6_RECVRTHDRDSTOPTS:
3880 			if (!checkonly)
3881 				udp->udp_ipv6_recvrthdrdstopts = onoff;
3882 			break;
3883 		case IPV6_RECVRTHDR:
3884 			if (!checkonly)
3885 				udp->udp_ipv6_recvrthdr = onoff;
3886 			break;
3887 		/*
3888 		 * Set sticky options or ancillary data.
3889 		 * If sticky options, (re)build any extension headers
3890 		 * that might be needed as a result.
3891 		 */
3892 		case IPV6_PKTINFO:
3893 			/*
3894 			 * The source address and ifindex are verified
3895 			 * in ip_opt_set(). For ancillary data the
3896 			 * source address is checked in ip_wput_v6.
3897 			 */
3898 			if (inlen != 0 && inlen != sizeof (struct in6_pktinfo))
3899 				return (EINVAL);
3900 			if (checkonly)
3901 				break;
3902 
3903 			if (inlen == 0) {
3904 				ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR);
3905 				ipp->ipp_sticky_ignored |=
3906 				    (IPPF_IFINDEX|IPPF_ADDR);
3907 			} else {
3908 				struct in6_pktinfo *pkti;
3909 
3910 				pkti = (struct in6_pktinfo *)invalp;
3911 				ipp->ipp_ifindex = pkti->ipi6_ifindex;
3912 				ipp->ipp_addr = pkti->ipi6_addr;
3913 				if (ipp->ipp_ifindex != 0)
3914 					ipp->ipp_fields |= IPPF_IFINDEX;
3915 				else
3916 					ipp->ipp_fields &= ~IPPF_IFINDEX;
3917 				if (!IN6_IS_ADDR_UNSPECIFIED(
3918 				    &ipp->ipp_addr))
3919 					ipp->ipp_fields |= IPPF_ADDR;
3920 				else
3921 					ipp->ipp_fields &= ~IPPF_ADDR;
3922 			}
3923 			if (sticky) {
3924 				error = udp_build_hdrs(q, udp);
3925 				if (error != 0)
3926 					return (error);
3927 			}
3928 			break;
3929 		case IPV6_HOPLIMIT:
3930 			if (sticky)
3931 				return (EINVAL);
3932 			if (inlen != 0 && inlen != sizeof (int))
3933 				return (EINVAL);
3934 			if (checkonly)
3935 				break;
3936 
3937 			if (inlen == 0) {
3938 				ipp->ipp_fields &= ~IPPF_HOPLIMIT;
3939 				ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT;
3940 			} else {
3941 				if (*i1 > 255 || *i1 < -1)
3942 					return (EINVAL);
3943 				if (*i1 == -1)
3944 					ipp->ipp_hoplimit =
3945 					    us->us_ipv6_hoplimit;
3946 				else
3947 					ipp->ipp_hoplimit = *i1;
3948 				ipp->ipp_fields |= IPPF_HOPLIMIT;
3949 			}
3950 			break;
3951 		case IPV6_TCLASS:
3952 			if (inlen != 0 && inlen != sizeof (int))
3953 				return (EINVAL);
3954 			if (checkonly)
3955 				break;
3956 
3957 			if (inlen == 0) {
3958 				ipp->ipp_fields &= ~IPPF_TCLASS;
3959 				ipp->ipp_sticky_ignored |= IPPF_TCLASS;
3960 			} else {
3961 				if (*i1 > 255 || *i1 < -1)
3962 					return (EINVAL);
3963 				if (*i1 == -1)
3964 					ipp->ipp_tclass = 0;
3965 				else
3966 					ipp->ipp_tclass = *i1;
3967 				ipp->ipp_fields |= IPPF_TCLASS;
3968 			}
3969 			if (sticky) {
3970 				error = udp_build_hdrs(q, udp);
3971 				if (error != 0)
3972 					return (error);
3973 			}
3974 			break;
3975 		case IPV6_NEXTHOP:
3976 			/*
3977 			 * IP will verify that the nexthop is reachable
3978 			 * and fail for sticky options.
3979 			 */
3980 			if (inlen != 0 && inlen != sizeof (sin6_t))
3981 				return (EINVAL);
3982 			if (checkonly)
3983 				break;
3984 
3985 			if (inlen == 0) {
3986 				ipp->ipp_fields &= ~IPPF_NEXTHOP;
3987 				ipp->ipp_sticky_ignored |= IPPF_NEXTHOP;
3988 			} else {
3989 				sin6_t *sin6 = (sin6_t *)invalp;
3990 
3991 				if (sin6->sin6_family != AF_INET6)
3992 					return (EAFNOSUPPORT);
3993 				if (IN6_IS_ADDR_V4MAPPED(
3994 				    &sin6->sin6_addr))
3995 					return (EADDRNOTAVAIL);
3996 				ipp->ipp_nexthop = sin6->sin6_addr;
3997 				if (!IN6_IS_ADDR_UNSPECIFIED(
3998 				    &ipp->ipp_nexthop))
3999 					ipp->ipp_fields |= IPPF_NEXTHOP;
4000 				else
4001 					ipp->ipp_fields &= ~IPPF_NEXTHOP;
4002 			}
4003 			if (sticky) {
4004 				error = udp_build_hdrs(q, udp);
4005 				if (error != 0)
4006 					return (error);
4007 			}
4008 			break;
4009 		case IPV6_HOPOPTS: {
4010 			ip6_hbh_t *hopts = (ip6_hbh_t *)invalp;
4011 			/*
4012 			 * Sanity checks - minimum size, size a multiple of
4013 			 * eight bytes, and matching size passed in.
4014 			 */
4015 			if (inlen != 0 &&
4016 			    inlen != (8 * (hopts->ip6h_len + 1)))
4017 				return (EINVAL);
4018 
4019 			if (checkonly)
4020 				break;
4021 
4022 			error = optcom_pkt_set(invalp, inlen, sticky,
4023 			    (uchar_t **)&ipp->ipp_hopopts,
4024 			    &ipp->ipp_hopoptslen,
4025 			    sticky ? udp->udp_label_len_v6 : 0);
4026 			if (error != 0)
4027 				return (error);
4028 			if (ipp->ipp_hopoptslen == 0) {
4029 				ipp->ipp_fields &= ~IPPF_HOPOPTS;
4030 				ipp->ipp_sticky_ignored |= IPPF_HOPOPTS;
4031 			} else {
4032 				ipp->ipp_fields |= IPPF_HOPOPTS;
4033 			}
4034 			if (sticky) {
4035 				error = udp_build_hdrs(q, udp);
4036 				if (error != 0)
4037 					return (error);
4038 			}
4039 			break;
4040 		}
4041 		case IPV6_RTHDRDSTOPTS: {
4042 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
4043 
4044 			/*
4045 			 * Sanity checks - minimum size, size a multiple of
4046 			 * eight bytes, and matching size passed in.
4047 			 */
4048 			if (inlen != 0 &&
4049 			    inlen != (8 * (dopts->ip6d_len + 1)))
4050 				return (EINVAL);
4051 
4052 			if (checkonly)
4053 				break;
4054 
4055 			if (inlen == 0) {
4056 				if (sticky &&
4057 				    (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) {
4058 					kmem_free(ipp->ipp_rtdstopts,
4059 					    ipp->ipp_rtdstoptslen);
4060 					ipp->ipp_rtdstopts = NULL;
4061 					ipp->ipp_rtdstoptslen = 0;
4062 				}
4063 
4064 				ipp->ipp_fields &= ~IPPF_RTDSTOPTS;
4065 				ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS;
4066 			} else {
4067 				error = optcom_pkt_set(invalp, inlen, sticky,
4068 				    (uchar_t **)&ipp->ipp_rtdstopts,
4069 				    &ipp->ipp_rtdstoptslen, 0);
4070 				if (error != 0)
4071 					return (error);
4072 				ipp->ipp_fields |= IPPF_RTDSTOPTS;
4073 			}
4074 			if (sticky) {
4075 				error = udp_build_hdrs(q, udp);
4076 				if (error != 0)
4077 					return (error);
4078 			}
4079 			break;
4080 		}
4081 		case IPV6_DSTOPTS: {
4082 			ip6_dest_t *dopts = (ip6_dest_t *)invalp;
4083 
4084 			/*
4085 			 * Sanity checks - minimum size, size a multiple of
4086 			 * eight bytes, and matching size passed in.
4087 			 */
4088 			if (inlen != 0 &&
4089 			    inlen != (8 * (dopts->ip6d_len + 1)))
4090 				return (EINVAL);
4091 
4092 			if (checkonly)
4093 				break;
4094 
4095 			if (inlen == 0) {
4096 				if (sticky &&
4097 				    (ipp->ipp_fields & IPPF_DSTOPTS) != 0) {
4098 					kmem_free(ipp->ipp_dstopts,
4099 					    ipp->ipp_dstoptslen);
4100 					ipp->ipp_dstopts = NULL;
4101 					ipp->ipp_dstoptslen = 0;
4102 				}
4103 				ipp->ipp_fields &= ~IPPF_DSTOPTS;
4104 				ipp->ipp_sticky_ignored |= IPPF_DSTOPTS;
4105 			} else {
4106 				error = optcom_pkt_set(invalp, inlen, sticky,
4107 				    (uchar_t **)&ipp->ipp_dstopts,
4108 				    &ipp->ipp_dstoptslen, 0);
4109 				if (error != 0)
4110 					return (error);
4111 				ipp->ipp_fields |= IPPF_DSTOPTS;
4112 			}
4113 			if (sticky) {
4114 				error = udp_build_hdrs(q, udp);
4115 				if (error != 0)
4116 					return (error);
4117 			}
4118 			break;
4119 		}
4120 		case IPV6_RTHDR: {
4121 			ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp;
4122 
4123 			/*
4124 			 * Sanity checks - minimum size, size a multiple of
4125 			 * eight bytes, and matching size passed in.
4126 			 */
4127 			if (inlen != 0 &&
4128 			    inlen != (8 * (rt->ip6r_len + 1)))
4129 				return (EINVAL);
4130 
4131 			if (checkonly)
4132 				break;
4133 
4134 			if (inlen == 0) {
4135 				if (sticky &&
4136 				    (ipp->ipp_fields & IPPF_RTHDR) != 0) {
4137 					kmem_free(ipp->ipp_rthdr,
4138 					    ipp->ipp_rthdrlen);
4139 					ipp->ipp_rthdr = NULL;
4140 					ipp->ipp_rthdrlen = 0;
4141 				}
4142 				ipp->ipp_fields &= ~IPPF_RTHDR;
4143 				ipp->ipp_sticky_ignored |= IPPF_RTHDR;
4144 			} else {
4145 				error = optcom_pkt_set(invalp, inlen, sticky,
4146 				    (uchar_t **)&ipp->ipp_rthdr,
4147 				    &ipp->ipp_rthdrlen, 0);
4148 				if (error != 0)
4149 					return (error);
4150 				ipp->ipp_fields |= IPPF_RTHDR;
4151 			}
4152 			if (sticky) {
4153 				error = udp_build_hdrs(q, udp);
4154 				if (error != 0)
4155 					return (error);
4156 			}
4157 			break;
4158 		}
4159 
4160 		case IPV6_DONTFRAG:
4161 			if (checkonly)
4162 				break;
4163 
4164 			if (onoff) {
4165 				ipp->ipp_fields |= IPPF_DONTFRAG;
4166 			} else {
4167 				ipp->ipp_fields &= ~IPPF_DONTFRAG;
4168 			}
4169 			break;
4170 
4171 		case IPV6_USE_MIN_MTU:
4172 			if (inlen != sizeof (int))
4173 				return (EINVAL);
4174 
4175 			if (*i1 < -1 || *i1 > 1)
4176 				return (EINVAL);
4177 
4178 			if (checkonly)
4179 				break;
4180 
4181 			ipp->ipp_fields |= IPPF_USE_MIN_MTU;
4182 			ipp->ipp_use_min_mtu = *i1;
4183 			break;
4184 
4185 		case IPV6_BOUND_PIF:
4186 		case IPV6_SEC_OPT:
4187 		case IPV6_DONTFAILOVER_IF:
4188 		case IPV6_SRC_PREFERENCES:
4189 		case IPV6_V6ONLY:
4190 			/* Handled at the IP level */
4191 			return (-EINVAL);
4192 		default:
4193 			*outlenp = 0;
4194 			return (EINVAL);
4195 		}
4196 		break;
4197 		}		/* end IPPROTO_IPV6 */
4198 	case IPPROTO_UDP:
4199 		switch (name) {
4200 		case UDP_ANONPRIVBIND:
4201 			if ((error = secpolicy_net_privaddr(cr, 0)) != 0) {
4202 				*outlenp = 0;
4203 				return (error);
4204 			}
4205 			if (!checkonly) {
4206 				udp->udp_anon_priv_bind = onoff;
4207 			}
4208 			break;
4209 		case UDP_EXCLBIND:
4210 			if (!checkonly)
4211 				udp->udp_exclbind = onoff;
4212 			break;
4213 		case UDP_RCVHDR:
4214 			if (!checkonly)
4215 				udp->udp_rcvhdr = onoff;
4216 			break;
4217 		case UDP_NAT_T_ENDPOINT:
4218 			if ((error = secpolicy_ip_config(cr, B_FALSE)) != 0) {
4219 				*outlenp = 0;
4220 				return (error);
4221 			}
4222 
4223 			/*
4224 			 * Use udp_family instead so we can avoid ambiguitites
4225 			 * with AF_INET6 sockets that may switch from IPv4
4226 			 * to IPv6.
4227 			 */
4228 			if (udp->udp_family != AF_INET) {
4229 				*outlenp = 0;
4230 				return (EAFNOSUPPORT);
4231 			}
4232 
4233 			if (!checkonly) {
4234 				udp->udp_nat_t_endpoint = onoff;
4235 
4236 				udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH +
4237 				    UDPH_SIZE + udp->udp_ip_snd_options_len;
4238 
4239 				/* Also, adjust wroff */
4240 				if (onoff) {
4241 					udp->udp_max_hdr_len +=
4242 					    sizeof (uint32_t);
4243 				}
4244 				(void) mi_set_sth_wroff(RD(q),
4245 				    udp->udp_max_hdr_len + us->us_wroff_extra);
4246 			}
4247 			break;
4248 		default:
4249 			*outlenp = 0;
4250 			return (EINVAL);
4251 		}
4252 		break;
4253 	default:
4254 		*outlenp = 0;
4255 		return (EINVAL);
4256 	}
4257 	/*
4258 	 * Common case of OK return with outval same as inval.
4259 	 */
4260 	if (invalp != outvalp) {
4261 		/* don't trust bcopy for identical src/dst */
4262 		(void) bcopy(invalp, outvalp, inlen);
4263 	}
4264 	*outlenp = inlen;
4265 	return (0);
4266 }
4267 
4268 /*
4269  * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl.
4270  * The headers include ip6i_t (if needed), ip6_t, any sticky extension
4271  * headers, and the udp header.
4272  * Returns failure if can't allocate memory.
4273  */
4274 static int
4275 udp_build_hdrs(queue_t *q, udp_t *udp)
4276 {
4277 	udp_stack_t *us = udp->udp_us;
4278 	uchar_t	*hdrs;
4279 	uint_t	hdrs_len;
4280 	ip6_t	*ip6h;
4281 	ip6i_t	*ip6i;
4282 	udpha_t	*udpha;
4283 	ip6_pkt_t *ipp = &udp->udp_sticky_ipp;
4284 
4285 	hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE;
4286 	ASSERT(hdrs_len != 0);
4287 	if (hdrs_len != udp->udp_sticky_hdrs_len) {
4288 		/* Need to reallocate */
4289 		hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP);
4290 		if (hdrs == NULL)
4291 			return (ENOMEM);
4292 
4293 		if (udp->udp_sticky_hdrs_len != 0) {
4294 			kmem_free(udp->udp_sticky_hdrs,
4295 			    udp->udp_sticky_hdrs_len);
4296 		}
4297 		udp->udp_sticky_hdrs = hdrs;
4298 		udp->udp_sticky_hdrs_len = hdrs_len;
4299 	}
4300 	ip_build_hdrs_v6(udp->udp_sticky_hdrs,
4301 	    udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP);
4302 
4303 	/* Set header fields not in ipp */
4304 	if (ipp->ipp_fields & IPPF_HAS_IP6I) {
4305 		ip6i = (ip6i_t *)udp->udp_sticky_hdrs;
4306 		ip6h = (ip6_t *)&ip6i[1];
4307 	} else {
4308 		ip6h = (ip6_t *)udp->udp_sticky_hdrs;
4309 	}
4310 
4311 	if (!(ipp->ipp_fields & IPPF_ADDR))
4312 		ip6h->ip6_src = udp->udp_v6src;
4313 
4314 	udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE);
4315 	udpha->uha_src_port = udp->udp_port;
4316 
4317 	/* Try to get everything in a single mblk */
4318 	if (hdrs_len > udp->udp_max_hdr_len) {
4319 		udp->udp_max_hdr_len = hdrs_len;
4320 		(void) mi_set_sth_wroff(RD(q), udp->udp_max_hdr_len +
4321 		    us->us_wroff_extra);
4322 	}
4323 	return (0);
4324 }
4325 
4326 /*
4327  * This routine retrieves the value of an ND variable in a udpparam_t
4328  * structure.  It is called through nd_getset when a user reads the
4329  * variable.
4330  */
4331 /* ARGSUSED */
4332 static int
4333 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
4334 {
4335 	udpparam_t *udppa = (udpparam_t *)cp;
4336 
4337 	(void) mi_mpprintf(mp, "%d", udppa->udp_param_value);
4338 	return (0);
4339 }
4340 
4341 /*
4342  * Walk through the param array specified registering each element with the
4343  * named dispatch (ND) handler.
4344  */
4345 static boolean_t
4346 udp_param_register(IDP *ndp, udpparam_t *udppa, int cnt)
4347 {
4348 	for (; cnt-- > 0; udppa++) {
4349 		if (udppa->udp_param_name && udppa->udp_param_name[0]) {
4350 			if (!nd_load(ndp, udppa->udp_param_name,
4351 			    udp_param_get, udp_param_set,
4352 			    (caddr_t)udppa)) {
4353 				nd_free(ndp);
4354 				return (B_FALSE);
4355 			}
4356 		}
4357 	}
4358 	if (!nd_load(ndp, "udp_extra_priv_ports",
4359 	    udp_extra_priv_ports_get, NULL, NULL)) {
4360 		nd_free(ndp);
4361 		return (B_FALSE);
4362 	}
4363 	if (!nd_load(ndp, "udp_extra_priv_ports_add",
4364 	    NULL, udp_extra_priv_ports_add, NULL)) {
4365 		nd_free(ndp);
4366 		return (B_FALSE);
4367 	}
4368 	if (!nd_load(ndp, "udp_extra_priv_ports_del",
4369 	    NULL, udp_extra_priv_ports_del, NULL)) {
4370 		nd_free(ndp);
4371 		return (B_FALSE);
4372 	}
4373 	if (!nd_load(ndp, "udp_status", udp_status_report, NULL,
4374 	    NULL)) {
4375 		nd_free(ndp);
4376 		return (B_FALSE);
4377 	}
4378 	if (!nd_load(ndp, "udp_bind_hash", udp_bind_hash_report, NULL,
4379 	    NULL)) {
4380 		nd_free(ndp);
4381 		return (B_FALSE);
4382 	}
4383 	return (B_TRUE);
4384 }
4385 
4386 /* This routine sets an ND variable in a udpparam_t structure. */
4387 /* ARGSUSED */
4388 static int
4389 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr)
4390 {
4391 	long		new_value;
4392 	udpparam_t	*udppa = (udpparam_t *)cp;
4393 
4394 	/*
4395 	 * Fail the request if the new value does not lie within the
4396 	 * required bounds.
4397 	 */
4398 	if (ddi_strtol(value, NULL, 10, &new_value) != 0 ||
4399 	    new_value < udppa->udp_param_min ||
4400 	    new_value > udppa->udp_param_max) {
4401 		return (EINVAL);
4402 	}
4403 
4404 	/* Set the new value */
4405 	udppa->udp_param_value = new_value;
4406 	return (0);
4407 }
4408 
4409 /*
4410  * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with
4411  * T_opthdr) and return the number of bytes copied.  'dbuf' may be NULL to
4412  * just count the length needed for allocation.  If 'dbuf' is non-NULL,
4413  * then it's assumed to be allocated to be large enough.
4414  *
4415  * Returns zero if trimming of the security option causes all options to go
4416  * away.
4417  */
4418 static size_t
4419 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf)
4420 {
4421 	struct T_opthdr *toh;
4422 	size_t hol = ipp->ipp_hopoptslen;
4423 	ip6_hbh_t *dstopt = NULL;
4424 	const ip6_hbh_t *srcopt = ipp->ipp_hopopts;
4425 	size_t tlen, olen, plen;
4426 	boolean_t deleting;
4427 	const struct ip6_opt *sopt, *lastpad;
4428 	struct ip6_opt *dopt;
4429 
4430 	if ((toh = (struct T_opthdr *)dbuf) != NULL) {
4431 		toh->level = IPPROTO_IPV6;
4432 		toh->name = IPV6_HOPOPTS;
4433 		toh->status = 0;
4434 		dstopt = (ip6_hbh_t *)(toh + 1);
4435 	}
4436 
4437 	/*
4438 	 * If labeling is enabled, then skip the label option
4439 	 * but get other options if there are any.
4440 	 */
4441 	if (is_system_labeled()) {
4442 		dopt = NULL;
4443 		if (dstopt != NULL) {
4444 			/* will fill in ip6h_len later */
4445 			dstopt->ip6h_nxt = srcopt->ip6h_nxt;
4446 			dopt = (struct ip6_opt *)(dstopt + 1);
4447 		}
4448 		sopt = (const struct ip6_opt *)(srcopt + 1);
4449 		hol -= sizeof (*srcopt);
4450 		tlen = sizeof (*dstopt);
4451 		lastpad = NULL;
4452 		deleting = B_FALSE;
4453 		/*
4454 		 * This loop finds the first (lastpad pointer) of any number of
4455 		 * pads that preceeds the security option, then treats the
4456 		 * security option as though it were a pad, and then finds the
4457 		 * next non-pad option (or end of list).
4458 		 *
4459 		 * It then treats the entire block as one big pad.  To preserve
4460 		 * alignment of any options that follow, or just the end of the
4461 		 * list, it computes a minimal new padding size that keeps the
4462 		 * same alignment for the next option.
4463 		 *
4464 		 * If it encounters just a sequence of pads with no security
4465 		 * option, those are copied as-is rather than collapsed.
4466 		 *
4467 		 * Note that to handle the end of list case, the code makes one
4468 		 * loop with 'hol' set to zero.
4469 		 */
4470 		for (;;) {
4471 			if (hol > 0) {
4472 				if (sopt->ip6o_type == IP6OPT_PAD1) {
4473 					if (lastpad == NULL)
4474 						lastpad = sopt;
4475 					sopt = (const struct ip6_opt *)
4476 					    &sopt->ip6o_len;
4477 					hol--;
4478 					continue;
4479 				}
4480 				olen = sopt->ip6o_len + sizeof (*sopt);
4481 				if (olen > hol)
4482 					olen = hol;
4483 				if (sopt->ip6o_type == IP6OPT_PADN ||
4484 				    sopt->ip6o_type == ip6opt_ls) {
4485 					if (sopt->ip6o_type == ip6opt_ls)
4486 						deleting = B_TRUE;
4487 					if (lastpad == NULL)
4488 						lastpad = sopt;
4489 					sopt = (const struct ip6_opt *)
4490 					    ((const char *)sopt + olen);
4491 					hol -= olen;
4492 					continue;
4493 				}
4494 			} else {
4495 				/* if nothing was copied at all, then delete */
4496 				if (tlen == sizeof (*dstopt))
4497 					return (0);
4498 				/* last pass; pick up any trailing padding */
4499 				olen = 0;
4500 			}
4501 			if (deleting) {
4502 				/*
4503 				 * compute aligning effect of deleted material
4504 				 * to reproduce with pad.
4505 				 */
4506 				plen = ((const char *)sopt -
4507 				    (const char *)lastpad) & 7;
4508 				tlen += plen;
4509 				if (dopt != NULL) {
4510 					if (plen == 1) {
4511 						dopt->ip6o_type = IP6OPT_PAD1;
4512 					} else if (plen > 1) {
4513 						plen -= sizeof (*dopt);
4514 						dopt->ip6o_type = IP6OPT_PADN;
4515 						dopt->ip6o_len = plen;
4516 						if (plen > 0)
4517 							bzero(dopt + 1, plen);
4518 					}
4519 					dopt = (struct ip6_opt *)
4520 					    ((char *)dopt + plen);
4521 				}
4522 				deleting = B_FALSE;
4523 				lastpad = NULL;
4524 			}
4525 			/* if there's uncopied padding, then copy that now */
4526 			if (lastpad != NULL) {
4527 				olen += (const char *)sopt -
4528 				    (const char *)lastpad;
4529 				sopt = lastpad;
4530 				lastpad = NULL;
4531 			}
4532 			if (dopt != NULL && olen > 0) {
4533 				bcopy(sopt, dopt, olen);
4534 				dopt = (struct ip6_opt *)((char *)dopt + olen);
4535 			}
4536 			if (hol == 0)
4537 				break;
4538 			tlen += olen;
4539 			sopt = (const struct ip6_opt *)
4540 			    ((const char *)sopt + olen);
4541 			hol -= olen;
4542 		}
4543 		/* go back and patch up the length value, rounded upward */
4544 		if (dstopt != NULL)
4545 			dstopt->ip6h_len = (tlen - 1) >> 3;
4546 	} else {
4547 		tlen = hol;
4548 		if (dstopt != NULL)
4549 			bcopy(srcopt, dstopt, hol);
4550 	}
4551 
4552 	tlen += sizeof (*toh);
4553 	if (toh != NULL)
4554 		toh->len = tlen;
4555 
4556 	return (tlen);
4557 }
4558 
4559 static void
4560 udp_input(conn_t *connp, mblk_t *mp)
4561 {
4562 	struct T_unitdata_ind	*tudi;
4563 	uchar_t			*rptr;		/* Pointer to IP header */
4564 	int			hdr_length;	/* Length of IP+UDP headers */
4565 	int			udi_size;	/* Size of T_unitdata_ind */
4566 	int			mp_len;
4567 	udp_t			*udp;
4568 	udpha_t			*udpha;
4569 	int			ipversion;
4570 	ip6_pkt_t		ipp;
4571 	ip6_t			*ip6h;
4572 	ip6i_t			*ip6i;
4573 	mblk_t			*mp1;
4574 	mblk_t			*options_mp = NULL;
4575 	ip_pktinfo_t		*pinfo = NULL;
4576 	cred_t			*cr = NULL;
4577 	queue_t			*q = connp->conn_rq;
4578 	pid_t			cpid;
4579 	cred_t			*rcr = connp->conn_cred;
4580 	udp_stack_t *us;
4581 
4582 	TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_START,
4583 	    "udp_rput_start: q %p mp %p", q, mp);
4584 
4585 	udp = connp->conn_udp;
4586 	us = udp->udp_us;
4587 	rptr = mp->b_rptr;
4588 	ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL);
4589 	ASSERT(OK_32PTR(rptr));
4590 
4591 	/*
4592 	 * IP should have prepended the options data in an M_CTL
4593 	 * Check M_CTL "type" to make sure are not here bcos of
4594 	 * a valid ICMP message
4595 	 */
4596 	if (DB_TYPE(mp) == M_CTL) {
4597 		if (MBLKL(mp) == sizeof (ip_pktinfo_t) &&
4598 		    ((ip_pktinfo_t *)mp->b_rptr)->ip_pkt_ulp_type ==
4599 		    IN_PKTINFO) {
4600 			/*
4601 			 * IP_RECVIF or IP_RECVSLLA or IPF_RECVADDR information
4602 			 * has been appended to the packet by IP. We need to
4603 			 * extract the mblk and adjust the rptr
4604 			 */
4605 			pinfo = (ip_pktinfo_t *)mp->b_rptr;
4606 			options_mp = mp;
4607 			mp = mp->b_cont;
4608 			rptr = mp->b_rptr;
4609 			UDP_STAT(us, udp_in_pktinfo);
4610 		} else {
4611 			/*
4612 			 * ICMP messages.
4613 			 */
4614 			udp_icmp_error(q, mp);
4615 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
4616 			    "udp_rput_end: q %p (%S)", q, "m_ctl");
4617 			return;
4618 		}
4619 	}
4620 
4621 	mp_len = msgdsize(mp);
4622 	/*
4623 	 * This is the inbound data path.
4624 	 * First, we check to make sure the IP version number is correct,
4625 	 * and then pull the IP and UDP headers into the first mblk.
4626 	 * Assume IP provides aligned packets - otherwise toss.
4627 	 * Also, check if we have a complete IP header.
4628 	 */
4629 
4630 	/* Initialize regardless if ipversion is IPv4 or IPv6 */
4631 	ipp.ipp_fields = 0;
4632 
4633 	ipversion = IPH_HDR_VERSION(rptr);
4634 	switch (ipversion) {
4635 	case IPV4_VERSION:
4636 		ASSERT(MBLKL(mp) >= sizeof (ipha_t));
4637 		ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP);
4638 		hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
4639 		if ((hdr_length > IP_SIMPLE_HDR_LENGTH + UDPH_SIZE) ||
4640 		    (udp->udp_ip_rcv_options_len)) {
4641 			/*
4642 			 * Handle IPv4 packets with options outside of the
4643 			 * main data path. Not needed for AF_INET6 sockets
4644 			 * since they don't support a getsockopt of IP_OPTIONS.
4645 			 */
4646 			if (udp->udp_family == AF_INET6)
4647 				break;
4648 			/*
4649 			 * UDP length check performed for IPv4 packets with
4650 			 * options to check whether UDP length specified in
4651 			 * the header is the same as the physical length of
4652 			 * the packet.
4653 			 */
4654 			udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
4655 			if (mp_len != (ntohs(udpha->uha_length) +
4656 			    hdr_length - UDPH_SIZE)) {
4657 				goto tossit;
4658 			}
4659 			/*
4660 			 * Handle the case where the packet has IP options
4661 			 * and the IP_RECVSLLA & IP_RECVIF are set
4662 			 */
4663 			if (pinfo != NULL)
4664 				mp = options_mp;
4665 			udp_become_writer(connp, mp, udp_rput_other_wrapper,
4666 			    SQTAG_UDP_INPUT);
4667 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
4668 			    "udp_rput_end: q %p (%S)", q, "end");
4669 			return;
4670 		}
4671 
4672 		/* Handle IPV6_RECVHOPLIMIT. */
4673 		if ((udp->udp_family == AF_INET6) && (pinfo != NULL) &&
4674 		    udp->udp_ip_recvpktinfo) {
4675 			if (pinfo->ip_pkt_flags & IPF_RECVIF) {
4676 				ipp.ipp_fields |= IPPF_IFINDEX;
4677 				ipp.ipp_ifindex = pinfo->ip_pkt_ifindex;
4678 			}
4679 		}
4680 		break;
4681 	case IPV6_VERSION:
4682 		/*
4683 		 * IPv6 packets can only be received by applications
4684 		 * that are prepared to receive IPv6 addresses.
4685 		 * The IP fanout must ensure this.
4686 		 */
4687 		ASSERT(udp->udp_family == AF_INET6);
4688 
4689 		ip6h = (ip6_t *)rptr;
4690 		ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr);
4691 
4692 		if (ip6h->ip6_nxt != IPPROTO_UDP) {
4693 			uint8_t nexthdrp;
4694 			/* Look for ifindex information */
4695 			if (ip6h->ip6_nxt == IPPROTO_RAW) {
4696 				ip6i = (ip6i_t *)ip6h;
4697 				if ((uchar_t *)&ip6i[1] > mp->b_wptr)
4698 					goto tossit;
4699 
4700 				if (ip6i->ip6i_flags & IP6I_IFINDEX) {
4701 					ASSERT(ip6i->ip6i_ifindex != 0);
4702 					ipp.ipp_fields |= IPPF_IFINDEX;
4703 					ipp.ipp_ifindex = ip6i->ip6i_ifindex;
4704 				}
4705 				rptr = (uchar_t *)&ip6i[1];
4706 				mp->b_rptr = rptr;
4707 				if (rptr == mp->b_wptr) {
4708 					mp1 = mp->b_cont;
4709 					freeb(mp);
4710 					mp = mp1;
4711 					rptr = mp->b_rptr;
4712 				}
4713 				if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE))
4714 					goto tossit;
4715 				ip6h = (ip6_t *)rptr;
4716 				mp_len = msgdsize(mp);
4717 			}
4718 			/*
4719 			 * Find any potentially interesting extension headers
4720 			 * as well as the length of the IPv6 + extension
4721 			 * headers.
4722 			 */
4723 			hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) +
4724 			    UDPH_SIZE;
4725 			ASSERT(nexthdrp == IPPROTO_UDP);
4726 		} else {
4727 			hdr_length = IPV6_HDR_LEN + UDPH_SIZE;
4728 			ip6i = NULL;
4729 		}
4730 		break;
4731 	default:
4732 		ASSERT(0);
4733 	}
4734 
4735 	/*
4736 	 * IP inspected the UDP header thus all of it must be in the mblk.
4737 	 * UDP length check is performed for IPv6 packets and IPv4 packets
4738 	 * without options to check if the size of the packet as specified
4739 	 * by the header is the same as the physical size of the packet.
4740 	 */
4741 	udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE));
4742 	if ((MBLKL(mp) < hdr_length) ||
4743 	    (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) {
4744 		goto tossit;
4745 	}
4746 
4747 	/* Walk past the headers. */
4748 	if (!udp->udp_rcvhdr) {
4749 		mp->b_rptr = rptr + hdr_length;
4750 		mp_len -= hdr_length;
4751 	}
4752 
4753 	/*
4754 	 * This is the inbound data path.  Packets are passed upstream as
4755 	 * T_UNITDATA_IND messages with full IP headers still attached.
4756 	 */
4757 	if (udp->udp_family == AF_INET) {
4758 		sin_t *sin;
4759 
4760 		ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
4761 
4762 		/*
4763 		 * Normally only send up the address.
4764 		 * If IP_RECVDSTADDR is set we include the destination IP
4765 		 * address as an option. With IP_RECVOPTS we include all
4766 		 * the IP options. Only ip_rput_other() handles packets
4767 		 * that contain IP options.
4768 		 */
4769 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
4770 		if (udp->udp_recvdstaddr) {
4771 			udi_size += sizeof (struct T_opthdr) +
4772 			    sizeof (struct in_addr);
4773 			UDP_STAT(us, udp_in_recvdstaddr);
4774 		}
4775 
4776 		if (udp->udp_ip_recvpktinfo && (pinfo != NULL) &&
4777 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
4778 			udi_size += sizeof (struct T_opthdr) +
4779 			    sizeof (struct in_pktinfo);
4780 			UDP_STAT(us, udp_ip_recvpktinfo);
4781 		}
4782 
4783 		/*
4784 		 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
4785 		 * space accordingly
4786 		 */
4787 		if (udp->udp_recvif && (pinfo != NULL) &&
4788 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4789 			udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
4790 			UDP_STAT(us, udp_in_recvif);
4791 		}
4792 
4793 		if (udp->udp_recvslla && (pinfo != NULL) &&
4794 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
4795 			udi_size += sizeof (struct T_opthdr) +
4796 			    sizeof (struct sockaddr_dl);
4797 			UDP_STAT(us, udp_in_recvslla);
4798 		}
4799 
4800 		if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) {
4801 			udi_size += sizeof (struct T_opthdr) + ucredsize;
4802 			cpid = DB_CPID(mp);
4803 			UDP_STAT(us, udp_in_recvucred);
4804 		}
4805 
4806 		/*
4807 		 * If SO_TIMESTAMP is set allocate the appropriate sized
4808 		 * buffer. Since gethrestime() expects a pointer aligned
4809 		 * argument, we allocate space necessary for extra
4810 		 * alignment (even though it might not be used).
4811 		 */
4812 		if (udp->udp_timestamp) {
4813 			udi_size += sizeof (struct T_opthdr) +
4814 			    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4815 			UDP_STAT(us, udp_in_timestamp);
4816 		}
4817 
4818 		/*
4819 		 * If IP_RECVTTL is set allocate the appropriate sized buffer
4820 		 */
4821 		if (udp->udp_recvttl) {
4822 			udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
4823 			UDP_STAT(us, udp_in_recvttl);
4824 		}
4825 		ASSERT(IPH_HDR_LENGTH((ipha_t *)rptr) == IP_SIMPLE_HDR_LENGTH);
4826 
4827 		/* Allocate a message block for the T_UNITDATA_IND structure. */
4828 		mp1 = allocb(udi_size, BPRI_MED);
4829 		if (mp1 == NULL) {
4830 			freemsg(mp);
4831 			if (options_mp != NULL)
4832 				freeb(options_mp);
4833 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
4834 			    "udp_rput_end: q %p (%S)", q, "allocbfail");
4835 			BUMP_MIB(&udp->udp_mib, udpInErrors);
4836 			return;
4837 		}
4838 		mp1->b_cont = mp;
4839 		mp = mp1;
4840 		mp->b_datap->db_type = M_PROTO;
4841 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
4842 		mp->b_wptr = (uchar_t *)tudi + udi_size;
4843 		tudi->PRIM_type = T_UNITDATA_IND;
4844 		tudi->SRC_length = sizeof (sin_t);
4845 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
4846 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
4847 		    sizeof (sin_t);
4848 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
4849 		tudi->OPT_length = udi_size;
4850 		sin = (sin_t *)&tudi[1];
4851 		sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
4852 		sin->sin_port =	udpha->uha_src_port;
4853 		sin->sin_family = udp->udp_family;
4854 		*(uint32_t *)&sin->sin_zero[0] = 0;
4855 		*(uint32_t *)&sin->sin_zero[4] = 0;
4856 
4857 		/*
4858 		 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
4859 		 * IP_RECVTTL has been set.
4860 		 */
4861 		if (udi_size != 0) {
4862 			/*
4863 			 * Copy in destination address before options to avoid
4864 			 * any padding issues.
4865 			 */
4866 			char *dstopt;
4867 
4868 			dstopt = (char *)&sin[1];
4869 			if (udp->udp_recvdstaddr) {
4870 				struct T_opthdr *toh;
4871 				ipaddr_t *dstptr;
4872 
4873 				toh = (struct T_opthdr *)dstopt;
4874 				toh->level = IPPROTO_IP;
4875 				toh->name = IP_RECVDSTADDR;
4876 				toh->len = sizeof (struct T_opthdr) +
4877 				    sizeof (ipaddr_t);
4878 				toh->status = 0;
4879 				dstopt += sizeof (struct T_opthdr);
4880 				dstptr = (ipaddr_t *)dstopt;
4881 				*dstptr = ((ipha_t *)rptr)->ipha_dst;
4882 				dstopt = (char *)toh + toh->len;
4883 				udi_size -= toh->len;
4884 			}
4885 
4886 			if (udp->udp_ip_recvpktinfo && (pinfo != NULL) &&
4887 			    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
4888 				struct T_opthdr *toh;
4889 				struct in_pktinfo *pktinfop;
4890 
4891 				toh = (struct T_opthdr *)dstopt;
4892 				toh->level = IPPROTO_IP;
4893 				toh->name = IP_PKTINFO;
4894 				toh->len = sizeof (struct T_opthdr) +
4895 				    sizeof (*pktinfop);
4896 				toh->status = 0;
4897 				dstopt += sizeof (struct T_opthdr);
4898 				pktinfop = (struct in_pktinfo *)dstopt;
4899 				pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
4900 				pktinfop->ipi_spec_dst =
4901 				    pinfo->ip_pkt_match_addr;
4902 				pktinfop->ipi_addr.s_addr =
4903 				    ((ipha_t *)rptr)->ipha_dst;
4904 
4905 				dstopt += sizeof (struct in_pktinfo);
4906 				udi_size -= toh->len;
4907 			}
4908 
4909 			if (udp->udp_recvslla && (pinfo != NULL) &&
4910 			    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
4911 
4912 				struct T_opthdr *toh;
4913 				struct sockaddr_dl	*dstptr;
4914 
4915 				toh = (struct T_opthdr *)dstopt;
4916 				toh->level = IPPROTO_IP;
4917 				toh->name = IP_RECVSLLA;
4918 				toh->len = sizeof (struct T_opthdr) +
4919 				    sizeof (struct sockaddr_dl);
4920 				toh->status = 0;
4921 				dstopt += sizeof (struct T_opthdr);
4922 				dstptr = (struct sockaddr_dl *)dstopt;
4923 				bcopy(&pinfo->ip_pkt_slla, dstptr,
4924 				    sizeof (struct sockaddr_dl));
4925 				dstopt = (char *)toh + toh->len;
4926 				udi_size -= toh->len;
4927 			}
4928 
4929 			if (udp->udp_recvif && (pinfo != NULL) &&
4930 			    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
4931 
4932 				struct T_opthdr *toh;
4933 				uint_t		*dstptr;
4934 
4935 				toh = (struct T_opthdr *)dstopt;
4936 				toh->level = IPPROTO_IP;
4937 				toh->name = IP_RECVIF;
4938 				toh->len = sizeof (struct T_opthdr) +
4939 				    sizeof (uint_t);
4940 				toh->status = 0;
4941 				dstopt += sizeof (struct T_opthdr);
4942 				dstptr = (uint_t *)dstopt;
4943 				*dstptr = pinfo->ip_pkt_ifindex;
4944 				dstopt = (char *)toh + toh->len;
4945 				udi_size -= toh->len;
4946 			}
4947 
4948 			if (cr != NULL) {
4949 				struct T_opthdr *toh;
4950 
4951 				toh = (struct T_opthdr *)dstopt;
4952 				toh->level = SOL_SOCKET;
4953 				toh->name = SCM_UCRED;
4954 				toh->len = sizeof (struct T_opthdr) + ucredsize;
4955 				toh->status = 0;
4956 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
4957 				dstopt = (char *)toh + toh->len;
4958 				udi_size -= toh->len;
4959 			}
4960 
4961 			if (udp->udp_timestamp) {
4962 				struct	T_opthdr *toh;
4963 
4964 				toh = (struct T_opthdr *)dstopt;
4965 				toh->level = SOL_SOCKET;
4966 				toh->name = SCM_TIMESTAMP;
4967 				toh->len = sizeof (struct T_opthdr) +
4968 				    sizeof (timestruc_t) + _POINTER_ALIGNMENT;
4969 				toh->status = 0;
4970 				dstopt += sizeof (struct T_opthdr);
4971 				/* Align for gethrestime() */
4972 				dstopt = (char *)P2ROUNDUP((intptr_t)dstopt,
4973 				    sizeof (intptr_t));
4974 				gethrestime((timestruc_t *)dstopt);
4975 				dstopt = (char *)toh + toh->len;
4976 				udi_size -= toh->len;
4977 			}
4978 
4979 			/*
4980 			 * CAUTION:
4981 			 * Due to aligment issues
4982 			 * Processing of IP_RECVTTL option
4983 			 * should always be the last. Adding
4984 			 * any option processing after this will
4985 			 * cause alignment panic.
4986 			 */
4987 			if (udp->udp_recvttl) {
4988 				struct	T_opthdr *toh;
4989 				uint8_t	*dstptr;
4990 
4991 				toh = (struct T_opthdr *)dstopt;
4992 				toh->level = IPPROTO_IP;
4993 				toh->name = IP_RECVTTL;
4994 				toh->len = sizeof (struct T_opthdr) +
4995 				    sizeof (uint8_t);
4996 				toh->status = 0;
4997 				dstopt += sizeof (struct T_opthdr);
4998 				dstptr = (uint8_t *)dstopt;
4999 				*dstptr = ((ipha_t *)rptr)->ipha_ttl;
5000 				dstopt = (char *)toh + toh->len;
5001 				udi_size -= toh->len;
5002 			}
5003 
5004 			/* Consumed all of allocated space */
5005 			ASSERT(udi_size == 0);
5006 		}
5007 	} else {
5008 		sin6_t *sin6;
5009 
5010 		/*
5011 		 * Handle both IPv4 and IPv6 packets for IPv6 sockets.
5012 		 *
5013 		 * Normally we only send up the address. If receiving of any
5014 		 * optional receive side information is enabled, we also send
5015 		 * that up as options.
5016 		 * [ Only udp_rput_other() handles packets that contain IP
5017 		 * options so code to account for does not appear immediately
5018 		 * below but elsewhere ]
5019 		 */
5020 		udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t);
5021 
5022 		if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS|
5023 		    IPPF_RTHDR|IPPF_IFINDEX)) {
5024 			if (udp->udp_ipv6_recvhopopts &&
5025 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
5026 				size_t hlen;
5027 
5028 				UDP_STAT(us, udp_in_recvhopopts);
5029 				hlen = copy_hop_opts(&ipp, NULL);
5030 				if (hlen == 0)
5031 					ipp.ipp_fields &= ~IPPF_HOPOPTS;
5032 				udi_size += hlen;
5033 			}
5034 			if ((udp->udp_ipv6_recvdstopts ||
5035 			    udp->udp_old_ipv6_recvdstopts) &&
5036 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
5037 				udi_size += sizeof (struct T_opthdr) +
5038 				    ipp.ipp_dstoptslen;
5039 				UDP_STAT(us, udp_in_recvdstopts);
5040 			}
5041 			if (((udp->udp_ipv6_recvdstopts &&
5042 			    udp->udp_ipv6_recvrthdr &&
5043 			    (ipp.ipp_fields & IPPF_RTHDR)) ||
5044 			    udp->udp_ipv6_recvrthdrdstopts) &&
5045 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
5046 				udi_size += sizeof (struct T_opthdr) +
5047 				    ipp.ipp_rtdstoptslen;
5048 				UDP_STAT(us, udp_in_recvrtdstopts);
5049 			}
5050 			if (udp->udp_ipv6_recvrthdr &&
5051 			    (ipp.ipp_fields & IPPF_RTHDR)) {
5052 				udi_size += sizeof (struct T_opthdr) +
5053 				    ipp.ipp_rthdrlen;
5054 				UDP_STAT(us, udp_in_recvrthdr);
5055 			}
5056 			if (udp->udp_ip_recvpktinfo &&
5057 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
5058 				udi_size += sizeof (struct T_opthdr) +
5059 				    sizeof (struct in6_pktinfo);
5060 				UDP_STAT(us, udp_in_recvpktinfo);
5061 			}
5062 
5063 		}
5064 		if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) {
5065 			udi_size += sizeof (struct T_opthdr) + ucredsize;
5066 			cpid = DB_CPID(mp);
5067 			UDP_STAT(us, udp_in_recvucred);
5068 		}
5069 
5070 		if (udp->udp_ipv6_recvhoplimit) {
5071 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
5072 			UDP_STAT(us, udp_in_recvhoplimit);
5073 		}
5074 
5075 		if (udp->udp_ipv6_recvtclass) {
5076 			udi_size += sizeof (struct T_opthdr) + sizeof (int);
5077 			UDP_STAT(us, udp_in_recvtclass);
5078 		}
5079 
5080 		mp1 = allocb(udi_size, BPRI_MED);
5081 		if (mp1 == NULL) {
5082 			freemsg(mp);
5083 			if (options_mp != NULL)
5084 				freeb(options_mp);
5085 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5086 			    "udp_rput_end: q %p (%S)", q, "allocbfail");
5087 			BUMP_MIB(&udp->udp_mib, udpInErrors);
5088 			return;
5089 		}
5090 		mp1->b_cont = mp;
5091 		mp = mp1;
5092 		mp->b_datap->db_type = M_PROTO;
5093 		tudi = (struct T_unitdata_ind *)mp->b_rptr;
5094 		mp->b_wptr = (uchar_t *)tudi + udi_size;
5095 		tudi->PRIM_type = T_UNITDATA_IND;
5096 		tudi->SRC_length = sizeof (sin6_t);
5097 		tudi->SRC_offset = sizeof (struct T_unitdata_ind);
5098 		tudi->OPT_offset = sizeof (struct T_unitdata_ind) +
5099 		    sizeof (sin6_t);
5100 		udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t));
5101 		tudi->OPT_length = udi_size;
5102 		sin6 = (sin6_t *)&tudi[1];
5103 		if (ipversion == IPV4_VERSION) {
5104 			in6_addr_t v6dst;
5105 
5106 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src,
5107 			    &sin6->sin6_addr);
5108 			IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst,
5109 			    &v6dst);
5110 			sin6->sin6_flowinfo = 0;
5111 			sin6->sin6_scope_id = 0;
5112 			sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst,
5113 			    connp->conn_zoneid, us->us_netstack);
5114 		} else {
5115 			sin6->sin6_addr = ip6h->ip6_src;
5116 			/* No sin6_flowinfo per API */
5117 			sin6->sin6_flowinfo = 0;
5118 			/* For link-scope source pass up scope id */
5119 			if ((ipp.ipp_fields & IPPF_IFINDEX) &&
5120 			    IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src))
5121 				sin6->sin6_scope_id = ipp.ipp_ifindex;
5122 			else
5123 				sin6->sin6_scope_id = 0;
5124 			sin6->__sin6_src_id = ip_srcid_find_addr(
5125 			    &ip6h->ip6_dst, connp->conn_zoneid,
5126 			    us->us_netstack);
5127 		}
5128 		sin6->sin6_port = udpha->uha_src_port;
5129 		sin6->sin6_family = udp->udp_family;
5130 
5131 		if (udi_size != 0) {
5132 			uchar_t *dstopt;
5133 
5134 			dstopt = (uchar_t *)&sin6[1];
5135 			if (udp->udp_ip_recvpktinfo &&
5136 			    (ipp.ipp_fields & IPPF_IFINDEX)) {
5137 				struct T_opthdr *toh;
5138 				struct in6_pktinfo *pkti;
5139 
5140 				toh = (struct T_opthdr *)dstopt;
5141 				toh->level = IPPROTO_IPV6;
5142 				toh->name = IPV6_PKTINFO;
5143 				toh->len = sizeof (struct T_opthdr) +
5144 				    sizeof (*pkti);
5145 				toh->status = 0;
5146 				dstopt += sizeof (struct T_opthdr);
5147 				pkti = (struct in6_pktinfo *)dstopt;
5148 				if (ipversion == IPV6_VERSION)
5149 					pkti->ipi6_addr = ip6h->ip6_dst;
5150 				else
5151 					IN6_IPADDR_TO_V4MAPPED(
5152 					    ((ipha_t *)rptr)->ipha_dst,
5153 					    &pkti->ipi6_addr);
5154 				pkti->ipi6_ifindex = ipp.ipp_ifindex;
5155 				dstopt += sizeof (*pkti);
5156 				udi_size -= toh->len;
5157 			}
5158 			if (udp->udp_ipv6_recvhoplimit) {
5159 				struct T_opthdr *toh;
5160 
5161 				toh = (struct T_opthdr *)dstopt;
5162 				toh->level = IPPROTO_IPV6;
5163 				toh->name = IPV6_HOPLIMIT;
5164 				toh->len = sizeof (struct T_opthdr) +
5165 				    sizeof (uint_t);
5166 				toh->status = 0;
5167 				dstopt += sizeof (struct T_opthdr);
5168 				if (ipversion == IPV6_VERSION)
5169 					*(uint_t *)dstopt = ip6h->ip6_hops;
5170 				else
5171 					*(uint_t *)dstopt =
5172 					    ((ipha_t *)rptr)->ipha_ttl;
5173 				dstopt += sizeof (uint_t);
5174 				udi_size -= toh->len;
5175 			}
5176 			if (udp->udp_ipv6_recvtclass) {
5177 				struct T_opthdr *toh;
5178 
5179 				toh = (struct T_opthdr *)dstopt;
5180 				toh->level = IPPROTO_IPV6;
5181 				toh->name = IPV6_TCLASS;
5182 				toh->len = sizeof (struct T_opthdr) +
5183 				    sizeof (uint_t);
5184 				toh->status = 0;
5185 				dstopt += sizeof (struct T_opthdr);
5186 				if (ipversion == IPV6_VERSION) {
5187 					*(uint_t *)dstopt =
5188 					    IPV6_FLOW_TCLASS(ip6h->ip6_flow);
5189 				} else {
5190 					ipha_t *ipha = (ipha_t *)rptr;
5191 					*(uint_t *)dstopt =
5192 					    ipha->ipha_type_of_service;
5193 				}
5194 				dstopt += sizeof (uint_t);
5195 				udi_size -= toh->len;
5196 			}
5197 			if (udp->udp_ipv6_recvhopopts &&
5198 			    (ipp.ipp_fields & IPPF_HOPOPTS)) {
5199 				size_t hlen;
5200 
5201 				hlen = copy_hop_opts(&ipp, dstopt);
5202 				dstopt += hlen;
5203 				udi_size -= hlen;
5204 			}
5205 			if (udp->udp_ipv6_recvdstopts &&
5206 			    udp->udp_ipv6_recvrthdr &&
5207 			    (ipp.ipp_fields & IPPF_RTHDR) &&
5208 			    (ipp.ipp_fields & IPPF_RTDSTOPTS)) {
5209 				struct T_opthdr *toh;
5210 
5211 				toh = (struct T_opthdr *)dstopt;
5212 				toh->level = IPPROTO_IPV6;
5213 				toh->name = IPV6_DSTOPTS;
5214 				toh->len = sizeof (struct T_opthdr) +
5215 				    ipp.ipp_rtdstoptslen;
5216 				toh->status = 0;
5217 				dstopt += sizeof (struct T_opthdr);
5218 				bcopy(ipp.ipp_rtdstopts, dstopt,
5219 				    ipp.ipp_rtdstoptslen);
5220 				dstopt += ipp.ipp_rtdstoptslen;
5221 				udi_size -= toh->len;
5222 			}
5223 			if (udp->udp_ipv6_recvrthdr &&
5224 			    (ipp.ipp_fields & IPPF_RTHDR)) {
5225 				struct T_opthdr *toh;
5226 
5227 				toh = (struct T_opthdr *)dstopt;
5228 				toh->level = IPPROTO_IPV6;
5229 				toh->name = IPV6_RTHDR;
5230 				toh->len = sizeof (struct T_opthdr) +
5231 				    ipp.ipp_rthdrlen;
5232 				toh->status = 0;
5233 				dstopt += sizeof (struct T_opthdr);
5234 				bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen);
5235 				dstopt += ipp.ipp_rthdrlen;
5236 				udi_size -= toh->len;
5237 			}
5238 			if (udp->udp_ipv6_recvdstopts &&
5239 			    (ipp.ipp_fields & IPPF_DSTOPTS)) {
5240 				struct T_opthdr *toh;
5241 
5242 				toh = (struct T_opthdr *)dstopt;
5243 				toh->level = IPPROTO_IPV6;
5244 				toh->name = IPV6_DSTOPTS;
5245 				toh->len = sizeof (struct T_opthdr) +
5246 				    ipp.ipp_dstoptslen;
5247 				toh->status = 0;
5248 				dstopt += sizeof (struct T_opthdr);
5249 				bcopy(ipp.ipp_dstopts, dstopt,
5250 				    ipp.ipp_dstoptslen);
5251 				dstopt += ipp.ipp_dstoptslen;
5252 				udi_size -= toh->len;
5253 			}
5254 
5255 			if (cr != NULL) {
5256 				struct T_opthdr *toh;
5257 
5258 				toh = (struct T_opthdr *)dstopt;
5259 				toh->level = SOL_SOCKET;
5260 				toh->name = SCM_UCRED;
5261 				toh->len = sizeof (struct T_opthdr) + ucredsize;
5262 				toh->status = 0;
5263 				(void) cred2ucred(cr, cpid, &toh[1], rcr);
5264 				dstopt += toh->len;
5265 				udi_size -= toh->len;
5266 			}
5267 			/* Consumed all of allocated space */
5268 			ASSERT(udi_size == 0);
5269 		}
5270 #undef	sin6
5271 		/* No IP_RECVDSTADDR for IPv6. */
5272 	}
5273 
5274 	BUMP_MIB(&udp->udp_mib, udpHCInDatagrams);
5275 	TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5276 	    "udp_rput_end: q %p (%S)", q, "end");
5277 	if (options_mp != NULL)
5278 		freeb(options_mp);
5279 
5280 	if (udp->udp_direct_sockfs) {
5281 		/*
5282 		 * There is nothing above us except for the stream head;
5283 		 * use the read-side synchronous stream interface in
5284 		 * order to reduce the time spent in interrupt thread.
5285 		 */
5286 		ASSERT(udp->udp_issocket);
5287 		udp_rcv_enqueue(UDP_RD(q), udp, mp, mp_len);
5288 	} else {
5289 		/*
5290 		 * Use regular STREAMS interface to pass data upstream
5291 		 * if this is not a socket endpoint, or if we have
5292 		 * switched over to the slow mode due to sockmod being
5293 		 * popped or a module being pushed on top of us.
5294 		 */
5295 		putnext(UDP_RD(q), mp);
5296 	}
5297 	return;
5298 
5299 tossit:
5300 	freemsg(mp);
5301 	if (options_mp != NULL)
5302 		freeb(options_mp);
5303 	BUMP_MIB(&udp->udp_mib, udpInErrors);
5304 }
5305 
5306 void
5307 udp_conn_recv(conn_t *connp, mblk_t *mp)
5308 {
5309 	_UDP_ENTER(connp, mp, udp_input_wrapper, SQTAG_UDP_FANOUT);
5310 }
5311 
5312 /* ARGSUSED */
5313 static void
5314 udp_input_wrapper(void *arg, mblk_t *mp, void *arg2)
5315 {
5316 	udp_input((conn_t *)arg, mp);
5317 	_UDP_EXIT((conn_t *)arg);
5318 }
5319 
5320 /*
5321  * Process non-M_DATA messages as well as M_DATA messages that requires
5322  * modifications to udp_ip_rcv_options i.e. IPv4 packets with IP options.
5323  */
5324 static void
5325 udp_rput_other(queue_t *q, mblk_t *mp)
5326 {
5327 	struct T_unitdata_ind	*tudi;
5328 	mblk_t			*mp1;
5329 	uchar_t			*rptr;
5330 	uchar_t			*new_rptr;
5331 	int			hdr_length;
5332 	int			udi_size;	/* Size of T_unitdata_ind */
5333 	int			opt_len;	/* Length of IP options */
5334 	sin_t			*sin;
5335 	struct T_error_ack	*tea;
5336 	mblk_t			*options_mp = NULL;
5337 	ip_pktinfo_t		*pinfo;
5338 	boolean_t		recv_on = B_FALSE;
5339 	cred_t			*cr = NULL;
5340 	udp_t			*udp = Q_TO_UDP(q);
5341 	pid_t			cpid;
5342 	cred_t			*rcr = udp->udp_connp->conn_cred;
5343 	udp_stack_t		*us = udp->udp_us;
5344 
5345 	TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_START,
5346 	    "udp_rput_other: q %p mp %p", q, mp);
5347 
5348 	ASSERT(OK_32PTR(mp->b_rptr));
5349 	rptr = mp->b_rptr;
5350 
5351 	switch (mp->b_datap->db_type) {
5352 	case M_CTL:
5353 		/*
5354 		 * We are here only if IP_RECVSLLA and/or IP_RECVIF are set
5355 		 */
5356 		recv_on = B_TRUE;
5357 		options_mp = mp;
5358 		pinfo = (ip_pktinfo_t *)options_mp->b_rptr;
5359 
5360 		/*
5361 		 * The actual data is in mp->b_cont
5362 		 */
5363 		mp = mp->b_cont;
5364 		ASSERT(OK_32PTR(mp->b_rptr));
5365 		rptr = mp->b_rptr;
5366 		break;
5367 	case M_DATA:
5368 		/*
5369 		 * M_DATA messages contain IPv4 datagrams.  They are handled
5370 		 * after this switch.
5371 		 */
5372 		break;
5373 	case M_PROTO:
5374 	case M_PCPROTO:
5375 		/* M_PROTO messages contain some type of TPI message. */
5376 		ASSERT((uintptr_t)(mp->b_wptr - rptr) <= (uintptr_t)INT_MAX);
5377 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
5378 			freemsg(mp);
5379 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5380 			    "udp_rput_other_end: q %p (%S)", q, "protoshort");
5381 			return;
5382 		}
5383 		tea = (struct T_error_ack *)rptr;
5384 
5385 		switch (tea->PRIM_type) {
5386 		case T_ERROR_ACK:
5387 			switch (tea->ERROR_prim) {
5388 			case O_T_BIND_REQ:
5389 			case T_BIND_REQ: {
5390 				/*
5391 				 * If our O_T_BIND_REQ/T_BIND_REQ fails,
5392 				 * clear out the associated port and source
5393 				 * address before passing the message
5394 				 * upstream. If this was caused by a T_CONN_REQ
5395 				 * revert back to bound state.
5396 				 */
5397 				udp_fanout_t	*udpf;
5398 
5399 				udpf = &us->us_bind_fanout[UDP_BIND_HASH(
5400 				    udp->udp_port, us->us_bind_fanout_size)];
5401 				mutex_enter(&udpf->uf_lock);
5402 				if (udp->udp_state == TS_DATA_XFER) {
5403 					/* Connect failed */
5404 					tea->ERROR_prim = T_CONN_REQ;
5405 					/* Revert back to the bound source */
5406 					udp->udp_v6src = udp->udp_bound_v6src;
5407 					udp->udp_state = TS_IDLE;
5408 					mutex_exit(&udpf->uf_lock);
5409 					if (udp->udp_family == AF_INET6)
5410 						(void) udp_build_hdrs(q, udp);
5411 					break;
5412 				}
5413 
5414 				if (udp->udp_discon_pending) {
5415 					tea->ERROR_prim = T_DISCON_REQ;
5416 					udp->udp_discon_pending = 0;
5417 				}
5418 					V6_SET_ZERO(udp->udp_v6src);
5419 					V6_SET_ZERO(udp->udp_bound_v6src);
5420 					udp->udp_state = TS_UNBND;
5421 					udp_bind_hash_remove(udp, B_TRUE);
5422 					udp->udp_port = 0;
5423 					mutex_exit(&udpf->uf_lock);
5424 					if (udp->udp_family == AF_INET6)
5425 						(void) udp_build_hdrs(q, udp);
5426 					break;
5427 				}
5428 			default:
5429 				break;
5430 			}
5431 			break;
5432 		case T_BIND_ACK:
5433 			udp_rput_bind_ack(q, mp);
5434 			return;
5435 
5436 		case T_OPTMGMT_ACK:
5437 		case T_OK_ACK:
5438 			break;
5439 		default:
5440 			freemsg(mp);
5441 			return;
5442 		}
5443 		putnext(UDP_RD(q), mp);
5444 		return;
5445 	}
5446 
5447 	/*
5448 	 * This is the inbound data path.
5449 	 * First, we make sure the data contains both IP and UDP headers.
5450 	 *
5451 	 * This handle IPv4 packets for only AF_INET sockets.
5452 	 * AF_INET6 sockets can never access udp_ip_rcv_options thus there
5453 	 * is no need saving the options.
5454 	 */
5455 	ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION);
5456 	hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE;
5457 	if (mp->b_wptr - rptr < hdr_length) {
5458 		if (!pullupmsg(mp, hdr_length)) {
5459 			freemsg(mp);
5460 			if (options_mp != NULL)
5461 				freeb(options_mp);
5462 			BUMP_MIB(&udp->udp_mib, udpInErrors);
5463 			TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5464 			    "udp_rput_other_end: q %p (%S)", q, "hdrshort");
5465 			return;
5466 		}
5467 		rptr = mp->b_rptr;
5468 	}
5469 	/* Walk past the headers. */
5470 	new_rptr = rptr + hdr_length;
5471 	if (!udp->udp_rcvhdr)
5472 		mp->b_rptr = new_rptr;
5473 
5474 	/* Save the options if any */
5475 	opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE);
5476 	if (opt_len > 0) {
5477 		if (opt_len > udp->udp_ip_rcv_options_len) {
5478 			if (udp->udp_ip_rcv_options_len)
5479 				mi_free((char *)udp->udp_ip_rcv_options);
5480 			udp->udp_ip_rcv_options_len = 0;
5481 			udp->udp_ip_rcv_options =
5482 			    (uchar_t *)mi_alloc(opt_len, BPRI_HI);
5483 			if (udp->udp_ip_rcv_options)
5484 				udp->udp_ip_rcv_options_len = opt_len;
5485 		}
5486 		if (udp->udp_ip_rcv_options_len) {
5487 			bcopy(rptr + IP_SIMPLE_HDR_LENGTH,
5488 			    udp->udp_ip_rcv_options, opt_len);
5489 			/* Adjust length if we are resusing the space */
5490 			udp->udp_ip_rcv_options_len = opt_len;
5491 		}
5492 	} else if (udp->udp_ip_rcv_options_len) {
5493 		mi_free((char *)udp->udp_ip_rcv_options);
5494 		udp->udp_ip_rcv_options = NULL;
5495 		udp->udp_ip_rcv_options_len = 0;
5496 	}
5497 
5498 	/*
5499 	 * Normally only send up the address.
5500 	 * If IP_RECVDSTADDR is set we include the destination IP
5501 	 * address as an option. With IP_RECVOPTS we include all
5502 	 * the IP options.
5503 	 */
5504 	udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
5505 	if (udp->udp_recvdstaddr) {
5506 		udi_size += sizeof (struct T_opthdr) + sizeof (struct in_addr);
5507 		UDP_STAT(us, udp_in_recvdstaddr);
5508 	}
5509 
5510 	if (udp->udp_ip_recvpktinfo && recv_on &&
5511 	    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
5512 		udi_size += sizeof (struct T_opthdr) +
5513 		    sizeof (struct in_pktinfo);
5514 		UDP_STAT(us, udp_ip_recvpktinfo);
5515 	}
5516 
5517 	if (udp->udp_recvopts && opt_len > 0) {
5518 		udi_size += sizeof (struct T_opthdr) + opt_len;
5519 		UDP_STAT(us, udp_in_recvopts);
5520 	}
5521 
5522 	/*
5523 	 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate
5524 	 * space accordingly
5525 	 */
5526 	if (udp->udp_recvif && recv_on &&
5527 	    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
5528 		udi_size += sizeof (struct T_opthdr) + sizeof (uint_t);
5529 		UDP_STAT(us, udp_in_recvif);
5530 	}
5531 
5532 	if (udp->udp_recvslla && recv_on &&
5533 	    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
5534 		udi_size += sizeof (struct T_opthdr) +
5535 		    sizeof (struct sockaddr_dl);
5536 		UDP_STAT(us, udp_in_recvslla);
5537 	}
5538 
5539 	if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) {
5540 		udi_size += sizeof (struct T_opthdr) + ucredsize;
5541 		cpid = DB_CPID(mp);
5542 		UDP_STAT(us, udp_in_recvucred);
5543 	}
5544 	/*
5545 	 * If IP_RECVTTL is set allocate the appropriate sized buffer
5546 	 */
5547 	if (udp->udp_recvttl) {
5548 		udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t);
5549 		UDP_STAT(us, udp_in_recvttl);
5550 	}
5551 
5552 	/* Allocate a message block for the T_UNITDATA_IND structure. */
5553 	mp1 = allocb(udi_size, BPRI_MED);
5554 	if (mp1 == NULL) {
5555 		freemsg(mp);
5556 		if (options_mp != NULL)
5557 			freeb(options_mp);
5558 		TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5559 		    "udp_rput_other_end: q %p (%S)", q, "allocbfail");
5560 		BUMP_MIB(&udp->udp_mib, udpInErrors);
5561 		return;
5562 	}
5563 	mp1->b_cont = mp;
5564 	mp = mp1;
5565 	mp->b_datap->db_type = M_PROTO;
5566 	tudi = (struct T_unitdata_ind *)mp->b_rptr;
5567 	mp->b_wptr = (uchar_t *)tudi + udi_size;
5568 	tudi->PRIM_type = T_UNITDATA_IND;
5569 	tudi->SRC_length = sizeof (sin_t);
5570 	tudi->SRC_offset = sizeof (struct T_unitdata_ind);
5571 	tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin_t);
5572 	udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t));
5573 	tudi->OPT_length = udi_size;
5574 
5575 	sin = (sin_t *)&tudi[1];
5576 	sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src;
5577 	sin->sin_port =	((in_port_t *)
5578 	    new_rptr)[-(UDPH_SIZE/sizeof (in_port_t))];
5579 	sin->sin_family = AF_INET;
5580 	*(uint32_t *)&sin->sin_zero[0] = 0;
5581 	*(uint32_t *)&sin->sin_zero[4] = 0;
5582 
5583 	/*
5584 	 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or
5585 	 * IP_RECVTTL has been set.
5586 	 */
5587 	if (udi_size != 0) {
5588 		/*
5589 		 * Copy in destination address before options to avoid any
5590 		 * padding issues.
5591 		 */
5592 		char *dstopt;
5593 
5594 		dstopt = (char *)&sin[1];
5595 		if (udp->udp_recvdstaddr) {
5596 			struct T_opthdr *toh;
5597 			ipaddr_t *dstptr;
5598 
5599 			toh = (struct T_opthdr *)dstopt;
5600 			toh->level = IPPROTO_IP;
5601 			toh->name = IP_RECVDSTADDR;
5602 			toh->len = sizeof (struct T_opthdr) + sizeof (ipaddr_t);
5603 			toh->status = 0;
5604 			dstopt += sizeof (struct T_opthdr);
5605 			dstptr = (ipaddr_t *)dstopt;
5606 			*dstptr = (((ipaddr_t *)rptr)[4]);
5607 			dstopt += sizeof (ipaddr_t);
5608 			udi_size -= toh->len;
5609 		}
5610 		if (udp->udp_recvopts && udi_size != 0) {
5611 			struct T_opthdr *toh;
5612 
5613 			toh = (struct T_opthdr *)dstopt;
5614 			toh->level = IPPROTO_IP;
5615 			toh->name = IP_RECVOPTS;
5616 			toh->len = sizeof (struct T_opthdr) + opt_len;
5617 			toh->status = 0;
5618 			dstopt += sizeof (struct T_opthdr);
5619 			bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt, opt_len);
5620 			dstopt += opt_len;
5621 			udi_size -= toh->len;
5622 		}
5623 		if (udp->udp_ip_recvpktinfo && recv_on &&
5624 		    (pinfo->ip_pkt_flags & IPF_RECVADDR)) {
5625 
5626 			struct T_opthdr *toh;
5627 			struct in_pktinfo *pktinfop;
5628 
5629 			toh = (struct T_opthdr *)dstopt;
5630 			toh->level = IPPROTO_IP;
5631 			toh->name = IP_PKTINFO;
5632 			toh->len = sizeof (struct T_opthdr) +
5633 			    sizeof (*pktinfop);
5634 			toh->status = 0;
5635 			dstopt += sizeof (struct T_opthdr);
5636 			pktinfop = (struct in_pktinfo *)dstopt;
5637 			pktinfop->ipi_ifindex = pinfo->ip_pkt_ifindex;
5638 			pktinfop->ipi_spec_dst = pinfo->ip_pkt_match_addr;
5639 
5640 			pktinfop->ipi_addr.s_addr = ((ipha_t *)rptr)->ipha_dst;
5641 
5642 			dstopt += sizeof (struct in_pktinfo);
5643 			udi_size -= toh->len;
5644 		}
5645 
5646 		if (udp->udp_recvslla && recv_on &&
5647 		    (pinfo->ip_pkt_flags & IPF_RECVSLLA)) {
5648 
5649 			struct T_opthdr *toh;
5650 			struct sockaddr_dl	*dstptr;
5651 
5652 			toh = (struct T_opthdr *)dstopt;
5653 			toh->level = IPPROTO_IP;
5654 			toh->name = IP_RECVSLLA;
5655 			toh->len = sizeof (struct T_opthdr) +
5656 			    sizeof (struct sockaddr_dl);
5657 			toh->status = 0;
5658 			dstopt += sizeof (struct T_opthdr);
5659 			dstptr = (struct sockaddr_dl *)dstopt;
5660 			bcopy(&pinfo->ip_pkt_slla, dstptr,
5661 			    sizeof (struct sockaddr_dl));
5662 			dstopt += sizeof (struct sockaddr_dl);
5663 			udi_size -= toh->len;
5664 		}
5665 
5666 		if (udp->udp_recvif && recv_on &&
5667 		    (pinfo->ip_pkt_flags & IPF_RECVIF)) {
5668 
5669 			struct T_opthdr *toh;
5670 			uint_t		*dstptr;
5671 
5672 			toh = (struct T_opthdr *)dstopt;
5673 			toh->level = IPPROTO_IP;
5674 			toh->name = IP_RECVIF;
5675 			toh->len = sizeof (struct T_opthdr) +
5676 			    sizeof (uint_t);
5677 			toh->status = 0;
5678 			dstopt += sizeof (struct T_opthdr);
5679 			dstptr = (uint_t *)dstopt;
5680 			*dstptr = pinfo->ip_pkt_ifindex;
5681 			dstopt += sizeof (uint_t);
5682 			udi_size -= toh->len;
5683 		}
5684 
5685 		if (cr != NULL) {
5686 			struct T_opthdr *toh;
5687 
5688 			toh = (struct T_opthdr *)dstopt;
5689 			toh->level = SOL_SOCKET;
5690 			toh->name = SCM_UCRED;
5691 			toh->len = sizeof (struct T_opthdr) + ucredsize;
5692 			toh->status = 0;
5693 			(void) cred2ucred(cr, cpid, &toh[1], rcr);
5694 			dstopt += toh->len;
5695 			udi_size -= toh->len;
5696 		}
5697 
5698 		if (udp->udp_recvttl) {
5699 			struct	T_opthdr *toh;
5700 			uint8_t	*dstptr;
5701 
5702 			toh = (struct T_opthdr *)dstopt;
5703 			toh->level = IPPROTO_IP;
5704 			toh->name = IP_RECVTTL;
5705 			toh->len = sizeof (struct T_opthdr) +
5706 			    sizeof (uint8_t);
5707 			toh->status = 0;
5708 			dstopt += sizeof (struct T_opthdr);
5709 			dstptr = (uint8_t *)dstopt;
5710 			*dstptr = ((ipha_t *)rptr)->ipha_ttl;
5711 			dstopt += sizeof (uint8_t);
5712 			udi_size -= toh->len;
5713 		}
5714 
5715 		ASSERT(udi_size == 0);	/* "Consumed" all of allocated space */
5716 	}
5717 	BUMP_MIB(&udp->udp_mib, udpHCInDatagrams);
5718 	TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END,
5719 	    "udp_rput_other_end: q %p (%S)", q, "end");
5720 	if (options_mp != NULL)
5721 		freeb(options_mp);
5722 
5723 	if (udp->udp_direct_sockfs) {
5724 		/*
5725 		 * There is nothing above us except for the stream head;
5726 		 * use the read-side synchronous stream interface in
5727 		 * order to reduce the time spent in interrupt thread.
5728 		 */
5729 		ASSERT(udp->udp_issocket);
5730 		udp_rcv_enqueue(UDP_RD(q), udp, mp, msgdsize(mp));
5731 	} else {
5732 		/*
5733 		 * Use regular STREAMS interface to pass data upstream
5734 		 * if this is not a socket endpoint, or if we have
5735 		 * switched over to the slow mode due to sockmod being
5736 		 * popped or a module being pushed on top of us.
5737 		 */
5738 		putnext(UDP_RD(q), mp);
5739 	}
5740 }
5741 
5742 /* ARGSUSED */
5743 static void
5744 udp_rput_other_wrapper(void *arg, mblk_t *mp, void *arg2)
5745 {
5746 	conn_t *connp = arg;
5747 
5748 	udp_rput_other(connp->conn_rq, mp);
5749 	udp_exit(connp);
5750 }
5751 
5752 /*
5753  * Process a T_BIND_ACK
5754  */
5755 static void
5756 udp_rput_bind_ack(queue_t *q, mblk_t *mp)
5757 {
5758 	udp_t	*udp = Q_TO_UDP(q);
5759 	mblk_t	*mp1;
5760 	ire_t	*ire;
5761 	struct T_bind_ack *tba;
5762 	uchar_t *addrp;
5763 	ipa_conn_t	*ac;
5764 	ipa6_conn_t	*ac6;
5765 
5766 	if (udp->udp_discon_pending)
5767 		udp->udp_discon_pending = 0;
5768 
5769 	/*
5770 	 * If a broadcast/multicast address was bound set
5771 	 * the source address to 0.
5772 	 * This ensures no datagrams with broadcast address
5773 	 * as source address are emitted (which would violate
5774 	 * RFC1122 - Hosts requirements)
5775 	 *
5776 	 * Note that when connecting the returned IRE is
5777 	 * for the destination address and we only perform
5778 	 * the broadcast check for the source address (it
5779 	 * is OK to connect to a broadcast/multicast address.)
5780 	 */
5781 	mp1 = mp->b_cont;
5782 	if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) {
5783 		ire = (ire_t *)mp1->b_rptr;
5784 
5785 		/*
5786 		 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast
5787 		 * local address.
5788 		 */
5789 		if (ire->ire_type == IRE_BROADCAST &&
5790 		    udp->udp_state != TS_DATA_XFER) {
5791 			/* This was just a local bind to a broadcast addr */
5792 			V6_SET_ZERO(udp->udp_v6src);
5793 			if (udp->udp_family == AF_INET6)
5794 				(void) udp_build_hdrs(q, udp);
5795 		} else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) {
5796 			/*
5797 			 * Local address not yet set - pick it from the
5798 			 * T_bind_ack
5799 			 */
5800 			tba = (struct T_bind_ack *)mp->b_rptr;
5801 			addrp = &mp->b_rptr[tba->ADDR_offset];
5802 			switch (udp->udp_family) {
5803 			case AF_INET:
5804 				if (tba->ADDR_length == sizeof (ipa_conn_t)) {
5805 					ac = (ipa_conn_t *)addrp;
5806 				} else {
5807 					ASSERT(tba->ADDR_length ==
5808 					    sizeof (ipa_conn_x_t));
5809 					ac = &((ipa_conn_x_t *)addrp)->acx_conn;
5810 				}
5811 				IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr,
5812 				    &udp->udp_v6src);
5813 				break;
5814 			case AF_INET6:
5815 				if (tba->ADDR_length == sizeof (ipa6_conn_t)) {
5816 					ac6 = (ipa6_conn_t *)addrp;
5817 				} else {
5818 					ASSERT(tba->ADDR_length ==
5819 					    sizeof (ipa6_conn_x_t));
5820 					ac6 = &((ipa6_conn_x_t *)
5821 					    addrp)->ac6x_conn;
5822 				}
5823 				udp->udp_v6src = ac6->ac6_laddr;
5824 				(void) udp_build_hdrs(q, udp);
5825 				break;
5826 			}
5827 		}
5828 		mp1 = mp1->b_cont;
5829 	}
5830 	/*
5831 	 * Look for one or more appended ACK message added by
5832 	 * udp_connect or udp_disconnect.
5833 	 * If none found just send up the T_BIND_ACK.
5834 	 * udp_connect has appended a T_OK_ACK and a T_CONN_CON.
5835 	 * udp_disconnect has appended a T_OK_ACK.
5836 	 */
5837 	if (mp1 != NULL) {
5838 		if (mp->b_cont == mp1)
5839 			mp->b_cont = NULL;
5840 		else {
5841 			ASSERT(mp->b_cont->b_cont == mp1);
5842 			mp->b_cont->b_cont = NULL;
5843 		}
5844 		freemsg(mp);
5845 		mp = mp1;
5846 		while (mp != NULL) {
5847 			mp1 = mp->b_cont;
5848 			mp->b_cont = NULL;
5849 			putnext(UDP_RD(q), mp);
5850 			mp = mp1;
5851 		}
5852 		return;
5853 	}
5854 	freemsg(mp->b_cont);
5855 	mp->b_cont = NULL;
5856 	putnext(UDP_RD(q), mp);
5857 }
5858 
5859 /*
5860  * return SNMP stuff in buffer in mpdata
5861  */
5862 int
5863 udp_snmp_get(queue_t *q, mblk_t *mpctl)
5864 {
5865 	mblk_t			*mpdata;
5866 	mblk_t			*mp_conn_ctl;
5867 	mblk_t			*mp_attr_ctl;
5868 	mblk_t			*mp6_conn_ctl;
5869 	mblk_t			*mp6_attr_ctl;
5870 	mblk_t			*mp_conn_tail;
5871 	mblk_t			*mp_attr_tail;
5872 	mblk_t			*mp6_conn_tail;
5873 	mblk_t			*mp6_attr_tail;
5874 	struct opthdr		*optp;
5875 	mib2_udpEntry_t		ude;
5876 	mib2_udp6Entry_t	ude6;
5877 	mib2_transportMLPEntry_t mlp;
5878 	int			state;
5879 	zoneid_t		zoneid;
5880 	int			i;
5881 	connf_t			*connfp;
5882 	conn_t			*connp = Q_TO_CONN(q);
5883 	udp_t			*udp = connp->conn_udp;
5884 	int			v4_conn_idx;
5885 	int			v6_conn_idx;
5886 	boolean_t		needattr;
5887 	ip_stack_t		*ipst = connp->conn_netstack->netstack_ip;
5888 
5889 	mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL;
5890 	if (mpctl == NULL ||
5891 	    (mpdata = mpctl->b_cont) == NULL ||
5892 	    (mp_conn_ctl = copymsg(mpctl)) == NULL ||
5893 	    (mp_attr_ctl = copymsg(mpctl)) == NULL ||
5894 	    (mp6_conn_ctl = copymsg(mpctl)) == NULL ||
5895 	    (mp6_attr_ctl = copymsg(mpctl)) == NULL) {
5896 		freemsg(mp_conn_ctl);
5897 		freemsg(mp_attr_ctl);
5898 		freemsg(mp6_conn_ctl);
5899 		return (0);
5900 	}
5901 
5902 	zoneid = connp->conn_zoneid;
5903 
5904 	/* fixed length structure for IPv4 and IPv6 counters */
5905 	SET_MIB(udp->udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t));
5906 	SET_MIB(udp->udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t));
5907 	/* synchronize 64- and 32-bit counters */
5908 	SYNC32_MIB(&udp->udp_mib, udpInDatagrams, udpHCInDatagrams);
5909 	SYNC32_MIB(&udp->udp_mib, udpOutDatagrams, udpHCOutDatagrams);
5910 
5911 	optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)];
5912 	optp->level = MIB2_UDP;
5913 	optp->name = 0;
5914 	(void) snmp_append_data(mpdata, (char *)&udp->udp_mib,
5915 	    sizeof (udp->udp_mib));
5916 	optp->len = msgdsize(mpdata);
5917 	qreply(q, mpctl);
5918 
5919 	mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL;
5920 	v4_conn_idx = v6_conn_idx = 0;
5921 
5922 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
5923 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
5924 		connp = NULL;
5925 
5926 		while ((connp = ipcl_get_next_conn(connfp, connp,
5927 		    IPCL_UDP))) {
5928 			udp = connp->conn_udp;
5929 			if (zoneid != connp->conn_zoneid)
5930 				continue;
5931 
5932 			/*
5933 			 * Note that the port numbers are sent in
5934 			 * host byte order
5935 			 */
5936 
5937 			if (udp->udp_state == TS_UNBND)
5938 				state = MIB2_UDP_unbound;
5939 			else if (udp->udp_state == TS_IDLE)
5940 				state = MIB2_UDP_idle;
5941 			else if (udp->udp_state == TS_DATA_XFER)
5942 				state = MIB2_UDP_connected;
5943 			else
5944 				state = MIB2_UDP_unknown;
5945 
5946 			needattr = B_FALSE;
5947 			bzero(&mlp, sizeof (mlp));
5948 			if (connp->conn_mlp_type != mlptSingle) {
5949 				if (connp->conn_mlp_type == mlptShared ||
5950 				    connp->conn_mlp_type == mlptBoth)
5951 					mlp.tme_flags |= MIB2_TMEF_SHARED;
5952 				if (connp->conn_mlp_type == mlptPrivate ||
5953 				    connp->conn_mlp_type == mlptBoth)
5954 					mlp.tme_flags |= MIB2_TMEF_PRIVATE;
5955 				needattr = B_TRUE;
5956 			}
5957 
5958 			/*
5959 			 * Create an IPv4 table entry for IPv4 entries and also
5960 			 * any IPv6 entries which are bound to in6addr_any
5961 			 * (i.e. anything a IPv4 peer could connect/send to).
5962 			 */
5963 			if (udp->udp_ipversion == IPV4_VERSION ||
5964 			    (udp->udp_state <= TS_IDLE &&
5965 			    IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) {
5966 				ude.udpEntryInfo.ue_state = state;
5967 				/*
5968 				 * If in6addr_any this will set it to
5969 				 * INADDR_ANY
5970 				 */
5971 				ude.udpLocalAddress =
5972 				    V4_PART_OF_V6(udp->udp_v6src);
5973 				ude.udpLocalPort = ntohs(udp->udp_port);
5974 				if (udp->udp_state == TS_DATA_XFER) {
5975 					/*
5976 					 * Can potentially get here for
5977 					 * v6 socket if another process
5978 					 * (say, ping) has just done a
5979 					 * sendto(), changing the state
5980 					 * from the TS_IDLE above to
5981 					 * TS_DATA_XFER by the time we hit
5982 					 * this part of the code.
5983 					 */
5984 					ude.udpEntryInfo.ue_RemoteAddress =
5985 					    V4_PART_OF_V6(udp->udp_v6dst);
5986 					ude.udpEntryInfo.ue_RemotePort =
5987 					    ntohs(udp->udp_dstport);
5988 				} else {
5989 					ude.udpEntryInfo.ue_RemoteAddress = 0;
5990 					ude.udpEntryInfo.ue_RemotePort = 0;
5991 				}
5992 
5993 				/*
5994 				 * We make the assumption that all udp_t
5995 				 * structs will be created within an address
5996 				 * region no larger than 32-bits.
5997 				 */
5998 				ude.udpInstance = (uint32_t)(uintptr_t)udp;
5999 				ude.udpCreationProcess =
6000 				    (udp->udp_open_pid < 0) ?
6001 				    MIB2_UNKNOWN_PROCESS :
6002 				    udp->udp_open_pid;
6003 				ude.udpCreationTime = udp->udp_open_time;
6004 
6005 				(void) snmp_append_data2(mp_conn_ctl->b_cont,
6006 				    &mp_conn_tail, (char *)&ude, sizeof (ude));
6007 				mlp.tme_connidx = v4_conn_idx++;
6008 				if (needattr)
6009 					(void) snmp_append_data2(
6010 					    mp_attr_ctl->b_cont, &mp_attr_tail,
6011 					    (char *)&mlp, sizeof (mlp));
6012 			}
6013 			if (udp->udp_ipversion == IPV6_VERSION) {
6014 				ude6.udp6EntryInfo.ue_state  = state;
6015 				ude6.udp6LocalAddress = udp->udp_v6src;
6016 				ude6.udp6LocalPort = ntohs(udp->udp_port);
6017 				ude6.udp6IfIndex = udp->udp_bound_if;
6018 				if (udp->udp_state == TS_DATA_XFER) {
6019 					ude6.udp6EntryInfo.ue_RemoteAddress =
6020 					    udp->udp_v6dst;
6021 					ude6.udp6EntryInfo.ue_RemotePort =
6022 					    ntohs(udp->udp_dstport);
6023 				} else {
6024 					ude6.udp6EntryInfo.ue_RemoteAddress =
6025 					    sin6_null.sin6_addr;
6026 					ude6.udp6EntryInfo.ue_RemotePort = 0;
6027 				}
6028 				/*
6029 				 * We make the assumption that all udp_t
6030 				 * structs will be created within an address
6031 				 * region no larger than 32-bits.
6032 				 */
6033 				ude6.udp6Instance = (uint32_t)(uintptr_t)udp;
6034 				ude6.udp6CreationProcess =
6035 				    (udp->udp_open_pid < 0) ?
6036 				    MIB2_UNKNOWN_PROCESS :
6037 				    udp->udp_open_pid;
6038 				ude6.udp6CreationTime = udp->udp_open_time;
6039 
6040 				(void) snmp_append_data2(mp6_conn_ctl->b_cont,
6041 				    &mp6_conn_tail, (char *)&ude6,
6042 				    sizeof (ude6));
6043 				mlp.tme_connidx = v6_conn_idx++;
6044 				if (needattr)
6045 					(void) snmp_append_data2(
6046 					    mp6_attr_ctl->b_cont,
6047 					    &mp6_attr_tail, (char *)&mlp,
6048 					    sizeof (mlp));
6049 			}
6050 		}
6051 	}
6052 
6053 	/* IPv4 UDP endpoints */
6054 	optp = (struct opthdr *)&mp_conn_ctl->b_rptr[
6055 	    sizeof (struct T_optmgmt_ack)];
6056 	optp->level = MIB2_UDP;
6057 	optp->name = MIB2_UDP_ENTRY;
6058 	optp->len = msgdsize(mp_conn_ctl->b_cont);
6059 	qreply(q, mp_conn_ctl);
6060 
6061 	/* table of MLP attributes... */
6062 	optp = (struct opthdr *)&mp_attr_ctl->b_rptr[
6063 	    sizeof (struct T_optmgmt_ack)];
6064 	optp->level = MIB2_UDP;
6065 	optp->name = EXPER_XPORT_MLP;
6066 	optp->len = msgdsize(mp_attr_ctl->b_cont);
6067 	if (optp->len == 0)
6068 		freemsg(mp_attr_ctl);
6069 	else
6070 		qreply(q, mp_attr_ctl);
6071 
6072 	/* IPv6 UDP endpoints */
6073 	optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[
6074 	    sizeof (struct T_optmgmt_ack)];
6075 	optp->level = MIB2_UDP6;
6076 	optp->name = MIB2_UDP6_ENTRY;
6077 	optp->len = msgdsize(mp6_conn_ctl->b_cont);
6078 	qreply(q, mp6_conn_ctl);
6079 
6080 	/* table of MLP attributes... */
6081 	optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[
6082 	    sizeof (struct T_optmgmt_ack)];
6083 	optp->level = MIB2_UDP6;
6084 	optp->name = EXPER_XPORT_MLP;
6085 	optp->len = msgdsize(mp6_attr_ctl->b_cont);
6086 	if (optp->len == 0)
6087 		freemsg(mp6_attr_ctl);
6088 	else
6089 		qreply(q, mp6_attr_ctl);
6090 
6091 	return (1);
6092 }
6093 
6094 /*
6095  * Return 0 if invalid set request, 1 otherwise, including non-udp requests.
6096  * NOTE: Per MIB-II, UDP has no writable data.
6097  * TODO:  If this ever actually tries to set anything, it needs to be
6098  * to do the appropriate locking.
6099  */
6100 /* ARGSUSED */
6101 int
6102 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name,
6103     uchar_t *ptr, int len)
6104 {
6105 	switch (level) {
6106 	case MIB2_UDP:
6107 		return (0);
6108 	default:
6109 		return (1);
6110 	}
6111 }
6112 
6113 static void
6114 udp_report_item(mblk_t *mp, udp_t *udp)
6115 {
6116 	char *state;
6117 	char addrbuf1[INET6_ADDRSTRLEN];
6118 	char addrbuf2[INET6_ADDRSTRLEN];
6119 	uint_t print_len, buf_len;
6120 
6121 	buf_len = mp->b_datap->db_lim - mp->b_wptr;
6122 	ASSERT(buf_len >= 0);
6123 	if (buf_len == 0)
6124 		return;
6125 
6126 	if (udp->udp_state == TS_UNBND)
6127 		state = "UNBOUND";
6128 	else if (udp->udp_state == TS_IDLE)
6129 		state = "IDLE";
6130 	else if (udp->udp_state == TS_DATA_XFER)
6131 		state = "CONNECTED";
6132 	else
6133 		state = "UnkState";
6134 	print_len = snprintf((char *)mp->b_wptr, buf_len,
6135 	    MI_COL_PTRFMT_STR "%4d %5u %s %s %5u %s\n",
6136 	    (void *)udp, udp->udp_connp->conn_zoneid, ntohs(udp->udp_port),
6137 	    inet_ntop(AF_INET6, &udp->udp_v6src, addrbuf1, sizeof (addrbuf1)),
6138 	    inet_ntop(AF_INET6, &udp->udp_v6dst, addrbuf2, sizeof (addrbuf2)),
6139 	    ntohs(udp->udp_dstport), state);
6140 	if (print_len < buf_len) {
6141 		mp->b_wptr += print_len;
6142 	} else {
6143 		mp->b_wptr += buf_len;
6144 	}
6145 }
6146 
6147 /* Report for ndd "udp_status" */
6148 /* ARGSUSED */
6149 static int
6150 udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr)
6151 {
6152 	zoneid_t zoneid;
6153 	connf_t	*connfp;
6154 	conn_t	*connp = Q_TO_CONN(q);
6155 	udp_t	*udp = connp->conn_udp;
6156 	int	i;
6157 	udp_stack_t *us = udp->udp_us;
6158 	ip_stack_t *ipst = connp->conn_netstack->netstack_ip;
6159 
6160 	/*
6161 	 * Because of the ndd constraint, at most we can have 64K buffer
6162 	 * to put in all UDP info.  So to be more efficient, just
6163 	 * allocate a 64K buffer here, assuming we need that large buffer.
6164 	 * This may be a problem as any user can read udp_status.  Therefore
6165 	 * we limit the rate of doing this using us_ndd_get_info_interval.
6166 	 * This should be OK as normal users should not do this too often.
6167 	 */
6168 	if (cr == NULL || secpolicy_ip_config(cr, B_TRUE) != 0) {
6169 		if (ddi_get_lbolt() - us->us_last_ndd_get_info_time <
6170 		    drv_usectohz(us->us_ndd_get_info_interval * 1000)) {
6171 			(void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG);
6172 			return (0);
6173 		}
6174 	}
6175 	if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) {
6176 		/* The following may work even if we cannot get a large buf. */
6177 		(void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG);
6178 		return (0);
6179 	}
6180 	(void) mi_mpprintf(mp,
6181 	    "UDP     " MI_COL_HDRPAD_STR
6182 	/*   12345678[89ABCDEF] */
6183 	    " zone lport src addr        dest addr       port  state");
6184 	/*    1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */
6185 
6186 	zoneid = connp->conn_zoneid;
6187 
6188 	for (i = 0; i < CONN_G_HASH_SIZE; i++) {
6189 		connfp = &ipst->ips_ipcl_globalhash_fanout[i];
6190 		connp = NULL;
6191 
6192 		while ((connp = ipcl_get_next_conn(connfp, connp,
6193 		    IPCL_UDP))) {
6194 			udp = connp->conn_udp;
6195 			if (zoneid != GLOBAL_ZONEID &&
6196 			    zoneid != connp->conn_zoneid)
6197 				continue;
6198 
6199 			udp_report_item(mp->b_cont, udp);
6200 		}
6201 	}
6202 	us->us_last_ndd_get_info_time = ddi_get_lbolt();
6203 	return (0);
6204 }
6205 
6206 /*
6207  * This routine creates a T_UDERROR_IND message and passes it upstream.
6208  * The address and options are copied from the T_UNITDATA_REQ message
6209  * passed in mp.  This message is freed.
6210  */
6211 static void
6212 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen,
6213     t_scalar_t err)
6214 {
6215 	struct T_unitdata_req *tudr;
6216 	mblk_t	*mp1;
6217 	uchar_t	*optaddr;
6218 	t_scalar_t optlen;
6219 
6220 	if (DB_TYPE(mp) == M_DATA) {
6221 		ASSERT(destaddr != NULL && destlen != 0);
6222 		optaddr = NULL;
6223 		optlen = 0;
6224 	} else {
6225 		if ((mp->b_wptr < mp->b_rptr) ||
6226 		    (MBLKL(mp)) < sizeof (struct T_unitdata_req)) {
6227 			goto done;
6228 		}
6229 		tudr = (struct T_unitdata_req *)mp->b_rptr;
6230 		destaddr = mp->b_rptr + tudr->DEST_offset;
6231 		if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr ||
6232 		    destaddr + tudr->DEST_length < mp->b_rptr ||
6233 		    destaddr + tudr->DEST_length > mp->b_wptr) {
6234 			goto done;
6235 		}
6236 		optaddr = mp->b_rptr + tudr->OPT_offset;
6237 		if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr ||
6238 		    optaddr + tudr->OPT_length < mp->b_rptr ||
6239 		    optaddr + tudr->OPT_length > mp->b_wptr) {
6240 			goto done;
6241 		}
6242 		destlen = tudr->DEST_length;
6243 		optlen = tudr->OPT_length;
6244 	}
6245 
6246 	mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen,
6247 	    (char *)optaddr, optlen, err);
6248 	if (mp1 != NULL)
6249 		putnext(UDP_RD(q), mp1);
6250 
6251 done:
6252 	freemsg(mp);
6253 }
6254 
6255 /*
6256  * This routine removes a port number association from a stream.  It
6257  * is called by udp_wput to handle T_UNBIND_REQ messages.
6258  */
6259 static void
6260 udp_unbind(queue_t *q, mblk_t *mp)
6261 {
6262 	udp_t *udp = Q_TO_UDP(q);
6263 
6264 	/* If a bind has not been done, we can't unbind. */
6265 	if (udp->udp_state == TS_UNBND) {
6266 		udp_err_ack(q, mp, TOUTSTATE, 0);
6267 		return;
6268 	}
6269 	if (cl_inet_unbind != NULL) {
6270 		/*
6271 		 * Running in cluster mode - register unbind information
6272 		 */
6273 		if (udp->udp_ipversion == IPV4_VERSION) {
6274 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET,
6275 			    (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)),
6276 			    (in_port_t)udp->udp_port);
6277 		} else {
6278 			(*cl_inet_unbind)(IPPROTO_UDP, AF_INET6,
6279 			    (uint8_t *)&(udp->udp_v6src),
6280 			    (in_port_t)udp->udp_port);
6281 		}
6282 	}
6283 
6284 	udp_bind_hash_remove(udp, B_FALSE);
6285 	V6_SET_ZERO(udp->udp_v6src);
6286 	V6_SET_ZERO(udp->udp_bound_v6src);
6287 	udp->udp_port = 0;
6288 	udp->udp_state = TS_UNBND;
6289 
6290 	if (udp->udp_family == AF_INET6) {
6291 		int error;
6292 
6293 		/* Rebuild the header template */
6294 		error = udp_build_hdrs(q, udp);
6295 		if (error != 0) {
6296 			udp_err_ack(q, mp, TSYSERR, error);
6297 			return;
6298 		}
6299 	}
6300 	/*
6301 	 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK
6302 	 * and therefore ip_unbind must never return NULL.
6303 	 */
6304 	mp = ip_unbind(q, mp);
6305 	ASSERT(mp != NULL);
6306 	putnext(UDP_RD(q), mp);
6307 }
6308 
6309 /*
6310  * Don't let port fall into the privileged range.
6311  * Since the extra privileged ports can be arbitrary we also
6312  * ensure that we exclude those from consideration.
6313  * us->us_epriv_ports is not sorted thus we loop over it until
6314  * there are no changes.
6315  */
6316 static in_port_t
6317 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random)
6318 {
6319 	int i;
6320 	in_port_t nextport;
6321 	boolean_t restart = B_FALSE;
6322 	udp_stack_t *us = udp->udp_us;
6323 
6324 	if (random && udp_random_anon_port != 0) {
6325 		(void) random_get_pseudo_bytes((uint8_t *)&port,
6326 		    sizeof (in_port_t));
6327 		/*
6328 		 * Unless changed by a sys admin, the smallest anon port
6329 		 * is 32768 and the largest anon port is 65535.  It is
6330 		 * very likely (50%) for the random port to be smaller
6331 		 * than the smallest anon port.  When that happens,
6332 		 * add port % (anon port range) to the smallest anon
6333 		 * port to get the random port.  It should fall into the
6334 		 * valid anon port range.
6335 		 */
6336 		if (port < us->us_smallest_anon_port) {
6337 			port = us->us_smallest_anon_port +
6338 			    port % (us->us_largest_anon_port -
6339 			    us->us_smallest_anon_port);
6340 		}
6341 	}
6342 
6343 retry:
6344 	if (port < us->us_smallest_anon_port)
6345 		port = us->us_smallest_anon_port;
6346 
6347 	if (port > us->us_largest_anon_port) {
6348 		port = us->us_smallest_anon_port;
6349 		if (restart)
6350 			return (0);
6351 		restart = B_TRUE;
6352 	}
6353 
6354 	if (port < us->us_smallest_nonpriv_port)
6355 		port = us->us_smallest_nonpriv_port;
6356 
6357 	for (i = 0; i < us->us_num_epriv_ports; i++) {
6358 		if (port == us->us_epriv_ports[i]) {
6359 			port++;
6360 			/*
6361 			 * Make sure that the port is in the
6362 			 * valid range.
6363 			 */
6364 			goto retry;
6365 		}
6366 	}
6367 
6368 	if (is_system_labeled() &&
6369 	    (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred),
6370 	    port, IPPROTO_UDP, B_TRUE)) != 0) {
6371 		port = nextport;
6372 		goto retry;
6373 	}
6374 
6375 	return (port);
6376 }
6377 
6378 static int
6379 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst)
6380 {
6381 	int err;
6382 	uchar_t opt_storage[IP_MAX_OPT_LENGTH];
6383 	udp_t *udp = Q_TO_UDP(wq);
6384 
6385 	err = tsol_compute_label(DB_CREDDEF(mp, udp->udp_connp->conn_cred), dst,
6386 	    opt_storage, udp->udp_mac_exempt,
6387 	    udp->udp_us->us_netstack->netstack_ip);
6388 	if (err == 0) {
6389 		err = tsol_update_options(&udp->udp_ip_snd_options,
6390 		    &udp->udp_ip_snd_options_len, &udp->udp_label_len,
6391 		    opt_storage);
6392 	}
6393 	if (err != 0) {
6394 		DTRACE_PROBE4(
6395 		    tx__ip__log__info__updatelabel__udp,
6396 		    char *, "queue(1) failed to update options(2) on mp(3)",
6397 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
6398 	} else {
6399 		IN6_IPADDR_TO_V4MAPPED(dst, &udp->udp_v6lastdst);
6400 	}
6401 	return (err);
6402 }
6403 
6404 static mblk_t *
6405 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port,
6406     uint_t srcid, int *error, boolean_t insert_spi)
6407 {
6408 	udp_t	*udp = connp->conn_udp;
6409 	queue_t	*q = connp->conn_wq;
6410 	mblk_t	*mp1 = mp;
6411 	mblk_t	*mp2;
6412 	ipha_t	*ipha;
6413 	int	ip_hdr_length;
6414 	uint32_t ip_len;
6415 	udpha_t	*udpha;
6416 	udpattrs_t	attrs;
6417 	uchar_t	ip_snd_opt[IP_MAX_OPT_LENGTH];
6418 	uint32_t	ip_snd_opt_len = 0;
6419 	ip4_pkt_t  pktinfo;
6420 	ip4_pkt_t  *pktinfop = &pktinfo;
6421 	ip_opt_info_t optinfo;
6422 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
6423 	udp_stack_t	*us = udp->udp_us;
6424 	ipsec_stack_t	*ipss = ipst->ips_netstack->netstack_ipsec;
6425 
6426 
6427 	*error = 0;
6428 	pktinfop->ip4_ill_index = 0;
6429 	pktinfop->ip4_addr = INADDR_ANY;
6430 	optinfo.ip_opt_flags = 0;
6431 	optinfo.ip_opt_ill_index = 0;
6432 
6433 	if (v4dst == INADDR_ANY)
6434 		v4dst = htonl(INADDR_LOOPBACK);
6435 
6436 	/*
6437 	 * If options passed in, feed it for verification and handling
6438 	 */
6439 	attrs.udpattr_credset = B_FALSE;
6440 	if (DB_TYPE(mp) != M_DATA) {
6441 		mp1 = mp->b_cont;
6442 		if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) {
6443 			attrs.udpattr_ipp4 = pktinfop;
6444 			attrs.udpattr_mb = mp;
6445 			if (udp_unitdata_opt_process(q, mp, error, &attrs) < 0)
6446 				goto done;
6447 			/*
6448 			 * Note: success in processing options.
6449 			 * mp option buffer represented by
6450 			 * OPT_length/offset now potentially modified
6451 			 * and contain option setting results
6452 			 */
6453 			ASSERT(*error == 0);
6454 		}
6455 	}
6456 
6457 	/* mp1 points to the M_DATA mblk carrying the packet */
6458 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
6459 
6460 	/*
6461 	 * Check if our saved options are valid; update if not.
6462 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
6463 	 * to different destination may require different labels,
6464 	 * or worse, UDP packets to same IP address may require
6465 	 * different labels due to use of shared all-zones address.
6466 	 * We use conn_lock to ensure that lastdst, ip_snd_options,
6467 	 * and ip_snd_options_len are consistent for the current
6468 	 * destination and are updated atomically.
6469 	 */
6470 	mutex_enter(&connp->conn_lock);
6471 	if (is_system_labeled()) {
6472 		/* Using UDP MLP requires SCM_UCRED from user */
6473 		if (connp->conn_mlp_type != mlptSingle &&
6474 		    !attrs.udpattr_credset) {
6475 			mutex_exit(&connp->conn_lock);
6476 			DTRACE_PROBE4(
6477 			    tx__ip__log__info__output__udp,
6478 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
6479 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
6480 			*error = ECONNREFUSED;
6481 			goto done;
6482 		}
6483 		/*
6484 		 * update label option for this UDP socket if
6485 		 * - the destination has changed, or
6486 		 * - the UDP socket is MLP
6487 		 */
6488 		if ((!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) ||
6489 		    V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst ||
6490 		    connp->conn_mlp_type != mlptSingle) &&
6491 		    (*error = udp_update_label(q, mp, v4dst)) != 0) {
6492 			mutex_exit(&connp->conn_lock);
6493 			goto done;
6494 		}
6495 	}
6496 	if (udp->udp_ip_snd_options_len > 0) {
6497 		ip_snd_opt_len = udp->udp_ip_snd_options_len;
6498 		bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len);
6499 	}
6500 	mutex_exit(&connp->conn_lock);
6501 
6502 	/* Add an IP header */
6503 	ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len +
6504 	    (insert_spi ? sizeof (uint32_t) : 0);
6505 	ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length];
6506 	if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) ||
6507 	    !OK_32PTR(ipha)) {
6508 		mp2 = allocb(ip_hdr_length + us->us_wroff_extra, BPRI_LO);
6509 		if (mp2 == NULL) {
6510 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6511 			    "udp_wput_end: q %p (%S)", q, "allocbfail2");
6512 			*error = ENOMEM;
6513 			goto done;
6514 		}
6515 		mp2->b_wptr = DB_LIM(mp2);
6516 		mp2->b_cont = mp1;
6517 		mp1 = mp2;
6518 		if (DB_TYPE(mp) != M_DATA)
6519 			mp->b_cont = mp1;
6520 		else
6521 			mp = mp1;
6522 
6523 		ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length);
6524 	}
6525 	ip_hdr_length -= (UDPH_SIZE + (insert_spi ? sizeof (uint32_t) : 0));
6526 #ifdef	_BIG_ENDIAN
6527 	/* Set version, header length, and tos */
6528 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
6529 	    ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) |
6530 	    udp->udp_type_of_service);
6531 	/* Set ttl and protocol */
6532 	*(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP;
6533 #else
6534 	/* Set version, header length, and tos */
6535 	*(uint16_t *)&ipha->ipha_version_and_hdr_length =
6536 	    ((udp->udp_type_of_service << 8) |
6537 	    ((IP_VERSION << 4) | (ip_hdr_length>>2)));
6538 	/* Set ttl and protocol */
6539 	*(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl;
6540 #endif
6541 	if (pktinfop->ip4_addr != INADDR_ANY) {
6542 		ipha->ipha_src = pktinfop->ip4_addr;
6543 		optinfo.ip_opt_flags = IP_VERIFY_SRC;
6544 	} else {
6545 		/*
6546 		 * Copy our address into the packet.  If this is zero,
6547 		 * first look at __sin6_src_id for a hint. If we leave the
6548 		 * source as INADDR_ANY then ip will fill in the real source
6549 		 * address.
6550 		 */
6551 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src);
6552 		if (srcid != 0 && ipha->ipha_src == INADDR_ANY) {
6553 			in6_addr_t v6src;
6554 
6555 			ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid,
6556 			    us->us_netstack);
6557 			IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src);
6558 		}
6559 	}
6560 
6561 	if (pktinfop->ip4_ill_index != 0) {
6562 		optinfo.ip_opt_ill_index = pktinfop->ip4_ill_index;
6563 	}
6564 
6565 	ipha->ipha_fragment_offset_and_flags = 0;
6566 	ipha->ipha_ident = 0;
6567 
6568 	mp1->b_rptr = (uchar_t *)ipha;
6569 
6570 	ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <=
6571 	    (uintptr_t)UINT_MAX);
6572 
6573 	/* Determine length of packet */
6574 	ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha);
6575 	if ((mp2 = mp1->b_cont) != NULL) {
6576 		do {
6577 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
6578 			ip_len += (uint32_t)MBLKL(mp2);
6579 		} while ((mp2 = mp2->b_cont) != NULL);
6580 	}
6581 	/*
6582 	 * If the size of the packet is greater than the maximum allowed by
6583 	 * ip, return an error. Passing this down could cause panics because
6584 	 * the size will have wrapped and be inconsistent with the msg size.
6585 	 */
6586 	if (ip_len > IP_MAXPACKET) {
6587 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6588 		    "udp_wput_end: q %p (%S)", q, "IP length exceeded");
6589 		*error = EMSGSIZE;
6590 		goto done;
6591 	}
6592 	ipha->ipha_length = htons((uint16_t)ip_len);
6593 	ip_len -= ip_hdr_length;
6594 	ip_len = htons((uint16_t)ip_len);
6595 	udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length);
6596 
6597 	/* Insert all-0s SPI now. */
6598 	if (insert_spi)
6599 		*((uint32_t *)(udpha + 1)) = 0;
6600 
6601 	/*
6602 	 * Copy in the destination address
6603 	 */
6604 	ipha->ipha_dst = v4dst;
6605 
6606 	/*
6607 	 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic.
6608 	 */
6609 	if (CLASSD(v4dst))
6610 		ipha->ipha_ttl = udp->udp_multicast_ttl;
6611 
6612 	udpha->uha_dst_port = port;
6613 	udpha->uha_src_port = udp->udp_port;
6614 
6615 	if (ip_snd_opt_len > 0) {
6616 		uint32_t	cksum;
6617 
6618 		bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len);
6619 		/*
6620 		 * Massage source route putting first source route in ipha_dst.
6621 		 * Ignore the destination in T_unitdata_req.
6622 		 * Create a checksum adjustment for a source route, if any.
6623 		 */
6624 		cksum = ip_massage_options(ipha, us->us_netstack);
6625 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
6626 		cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) +
6627 		    (ipha->ipha_dst & 0xFFFF);
6628 		if ((int)cksum < 0)
6629 			cksum--;
6630 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
6631 		/*
6632 		 * IP does the checksum if uha_checksum is non-zero,
6633 		 * We make it easy for IP to include our pseudo header
6634 		 * by putting our length in uha_checksum.
6635 		 */
6636 		cksum += ip_len;
6637 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
6638 		/* There might be a carry. */
6639 		cksum = (cksum & 0xFFFF) + (cksum >> 16);
6640 #ifdef _LITTLE_ENDIAN
6641 		if (us->us_do_checksum)
6642 			ip_len = (cksum << 16) | ip_len;
6643 #else
6644 		if (us->us_do_checksum)
6645 			ip_len = (ip_len << 16) | cksum;
6646 		else
6647 			ip_len <<= 16;
6648 #endif
6649 	} else {
6650 		/*
6651 		 * IP does the checksum if uha_checksum is non-zero,
6652 		 * We make it easy for IP to include our pseudo header
6653 		 * by putting our length in uha_checksum.
6654 		 */
6655 		if (us->us_do_checksum)
6656 			ip_len |= (ip_len << 16);
6657 #ifndef _LITTLE_ENDIAN
6658 		else
6659 			ip_len <<= 16;
6660 #endif
6661 	}
6662 
6663 	/* Set UDP length and checksum */
6664 	*((uint32_t *)&udpha->uha_length) = ip_len;
6665 	if (DB_CRED(mp) != NULL)
6666 		mblk_setcred(mp1, DB_CRED(mp));
6667 
6668 	if (DB_TYPE(mp) != M_DATA) {
6669 		ASSERT(mp != mp1);
6670 		freeb(mp);
6671 	}
6672 
6673 	/* mp has been consumed and we'll return success */
6674 	ASSERT(*error == 0);
6675 	mp = NULL;
6676 
6677 	/* We're done.  Pass the packet to ip. */
6678 	BUMP_MIB(&udp->udp_mib, udpHCOutDatagrams);
6679 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
6680 	    "udp_wput_end: q %p (%S)", q, "end");
6681 
6682 	if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 ||
6683 	    CONN_OUTBOUND_POLICY_PRESENT(connp, ipss) ||
6684 	    connp->conn_dontroute || connp->conn_xmit_if_ill != NULL ||
6685 	    connp->conn_nofailover_ill != NULL ||
6686 	    connp->conn_outgoing_ill != NULL || optinfo.ip_opt_flags != 0 ||
6687 	    optinfo.ip_opt_ill_index != 0 ||
6688 	    ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION ||
6689 	    IPP_ENABLED(IPP_LOCAL_OUT, ipst) ||
6690 	    ipst->ips_ip_g_mrouter != NULL) {
6691 		UDP_STAT(us, udp_ip_send);
6692 		ip_output_options(connp, mp1, connp->conn_wq, IP_WPUT,
6693 		    &optinfo);
6694 	} else {
6695 		udp_send_data(udp, connp->conn_wq, mp1, ipha);
6696 	}
6697 
6698 done:
6699 	if (*error != 0) {
6700 		ASSERT(mp != NULL);
6701 		BUMP_MIB(&udp->udp_mib, udpOutErrors);
6702 	}
6703 	return (mp);
6704 }
6705 
6706 static void
6707 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha)
6708 {
6709 	conn_t	*connp = udp->udp_connp;
6710 	ipaddr_t src, dst;
6711 	ill_t	*ill;
6712 	ire_t	*ire;
6713 	ipif_t	*ipif = NULL;
6714 	mblk_t	*ire_fp_mp;
6715 	uint_t	ire_fp_mp_len;
6716 	uint16_t *up;
6717 	uint32_t cksum, hcksum_txflags;
6718 	queue_t	*dev_q;
6719 	boolean_t retry_caching;
6720 	udp_stack_t *us = udp->udp_us;
6721 	ip_stack_t	*ipst = connp->conn_netstack->netstack_ip;
6722 
6723 	dst = ipha->ipha_dst;
6724 	src = ipha->ipha_src;
6725 	ASSERT(ipha->ipha_ident == 0);
6726 
6727 	if (CLASSD(dst)) {
6728 		int err;
6729 
6730 		ipif = conn_get_held_ipif(connp,
6731 		    &connp->conn_multicast_ipif, &err);
6732 
6733 		if (ipif == NULL || ipif->ipif_isv6 ||
6734 		    (ipif->ipif_ill->ill_phyint->phyint_flags &
6735 		    PHYI_LOOPBACK)) {
6736 			if (ipif != NULL)
6737 				ipif_refrele(ipif);
6738 			UDP_STAT(us, udp_ip_send);
6739 			ip_output(connp, mp, q, IP_WPUT);
6740 			return;
6741 		}
6742 	}
6743 
6744 	retry_caching = B_FALSE;
6745 	mutex_enter(&connp->conn_lock);
6746 	ire = connp->conn_ire_cache;
6747 	ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT));
6748 
6749 	if (ire == NULL || ire->ire_addr != dst ||
6750 	    (ire->ire_marks & IRE_MARK_CONDEMNED)) {
6751 		retry_caching = B_TRUE;
6752 	} else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) {
6753 		ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr;
6754 
6755 		ASSERT(ipif != NULL);
6756 		if (stq_ill != ipif->ipif_ill && (stq_ill->ill_group == NULL ||
6757 		    stq_ill->ill_group != ipif->ipif_ill->ill_group))
6758 			retry_caching = B_TRUE;
6759 	}
6760 
6761 	if (!retry_caching) {
6762 		ASSERT(ire != NULL);
6763 		IRE_REFHOLD(ire);
6764 		mutex_exit(&connp->conn_lock);
6765 	} else {
6766 		boolean_t cached = B_FALSE;
6767 
6768 		connp->conn_ire_cache = NULL;
6769 		mutex_exit(&connp->conn_lock);
6770 
6771 		/* Release the old ire */
6772 		if (ire != NULL) {
6773 			IRE_REFRELE_NOTR(ire);
6774 			ire = NULL;
6775 		}
6776 
6777 		if (CLASSD(dst)) {
6778 			ASSERT(ipif != NULL);
6779 			ire = ire_ctable_lookup(dst, 0, 0, ipif,
6780 			    connp->conn_zoneid, MBLK_GETLABEL(mp),
6781 			    MATCH_IRE_ILL_GROUP, ipst);
6782 		} else {
6783 			ASSERT(ipif == NULL);
6784 			ire = ire_cache_lookup(dst, connp->conn_zoneid,
6785 			    MBLK_GETLABEL(mp), ipst);
6786 		}
6787 
6788 		if (ire == NULL) {
6789 			if (ipif != NULL)
6790 				ipif_refrele(ipif);
6791 			UDP_STAT(us, udp_ire_null);
6792 			ip_output(connp, mp, q, IP_WPUT);
6793 			return;
6794 		}
6795 		IRE_REFHOLD_NOTR(ire);
6796 
6797 		mutex_enter(&connp->conn_lock);
6798 		if (CONN_CACHE_IRE(connp) && connp->conn_ire_cache == NULL) {
6799 			rw_enter(&ire->ire_bucket->irb_lock, RW_READER);
6800 			if (!(ire->ire_marks & IRE_MARK_CONDEMNED)) {
6801 				connp->conn_ire_cache = ire;
6802 				cached = B_TRUE;
6803 			}
6804 			rw_exit(&ire->ire_bucket->irb_lock);
6805 		}
6806 		mutex_exit(&connp->conn_lock);
6807 
6808 		/*
6809 		 * We can continue to use the ire but since it was not
6810 		 * cached, we should drop the extra reference.
6811 		 */
6812 		if (!cached)
6813 			IRE_REFRELE_NOTR(ire);
6814 	}
6815 	ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION);
6816 	ASSERT(!CLASSD(dst) || ipif != NULL);
6817 
6818 	/*
6819 	 * Check if we can take the fast-path.
6820 	 * Note that "incomplete" ire's (where the link-layer for next hop
6821 	 * is not resolved, or where the fast-path header in nce_fp_mp is not
6822 	 * available yet) are sent down the legacy (slow) path
6823 	 */
6824 	if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) ||
6825 	    (ire->ire_flags & RTF_MULTIRT) || (ire->ire_stq == NULL) ||
6826 	    (ire->ire_max_frag < ntohs(ipha->ipha_length)) ||
6827 	    (connp->conn_nexthop_set) ||
6828 	    (ire->ire_nce == NULL) ||
6829 	    ((ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL) ||
6830 	    ((ire_fp_mp_len = MBLKL(ire_fp_mp)) > MBLKHEAD(mp))) {
6831 		if (ipif != NULL)
6832 			ipif_refrele(ipif);
6833 		UDP_STAT(us, udp_ip_ire_send);
6834 		IRE_REFRELE(ire);
6835 		ip_output(connp, mp, q, IP_WPUT);
6836 		return;
6837 	}
6838 
6839 	ill = ire_to_ill(ire);
6840 	ASSERT(ill != NULL);
6841 
6842 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutRequests);
6843 
6844 	dev_q = ire->ire_stq->q_next;
6845 	ASSERT(dev_q != NULL);
6846 	/*
6847 	 * If the service thread is already running, or if the driver
6848 	 * queue is currently flow-controlled, queue this packet.
6849 	 */
6850 	if ((q->q_first != NULL || connp->conn_draining) ||
6851 	    ((dev_q->q_next || dev_q->q_first) && !canput(dev_q))) {
6852 		if (ipst->ips_ip_output_queue) {
6853 			(void) putq(q, mp);
6854 		} else {
6855 			BUMP_MIB(ill->ill_ip_mib, ipIfStatsOutDiscards);
6856 			freemsg(mp);
6857 		}
6858 		if (ipif != NULL)
6859 			ipif_refrele(ipif);
6860 		IRE_REFRELE(ire);
6861 		return;
6862 	}
6863 
6864 	ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1);
6865 #ifndef _BIG_ENDIAN
6866 	ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8);
6867 #endif
6868 
6869 	if (src == INADDR_ANY && !connp->conn_unspec_src) {
6870 		if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC))
6871 			src = ipha->ipha_src = ipif->ipif_src_addr;
6872 		else
6873 			src = ipha->ipha_src = ire->ire_src_addr;
6874 	}
6875 
6876 	if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) {
6877 		ASSERT(ill->ill_hcksum_capab != NULL);
6878 		hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags;
6879 	} else {
6880 		hcksum_txflags = 0;
6881 	}
6882 
6883 	/* pseudo-header checksum (do it in parts for IP header checksum) */
6884 	cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF);
6885 
6886 	ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION);
6887 	up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH);
6888 	if (*up != 0) {
6889 		IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags,
6890 		    mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH,
6891 		    ntohs(ipha->ipha_length), cksum);
6892 
6893 		/* Software checksum? */
6894 		if (DB_CKSUMFLAGS(mp) == 0) {
6895 			UDP_STAT(us, udp_out_sw_cksum);
6896 			UDP_STAT_UPDATE(us, udp_out_sw_cksum_bytes,
6897 			    ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH);
6898 		}
6899 	}
6900 
6901 	if (!CLASSD(dst)) {
6902 		ipha->ipha_fragment_offset_and_flags |=
6903 		    (uint32_t)htons(ire->ire_frag_flag);
6904 	}
6905 
6906 	/* Calculate IP header checksum if hardware isn't capable */
6907 	if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) {
6908 		IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0],
6909 		    ((uint16_t *)ipha)[4]);
6910 	}
6911 
6912 	if (CLASSD(dst)) {
6913 		ilm_t *ilm;
6914 
6915 		ILM_WALKER_HOLD(ill);
6916 		ilm = ilm_lookup_ill(ill, dst, ALL_ZONES);
6917 		ILM_WALKER_RELE(ill);
6918 		if (ilm != NULL) {
6919 			ip_multicast_loopback(q, ill, mp,
6920 			    connp->conn_multicast_loop ? 0 :
6921 			    IP_FF_NO_MCAST_LOOP, connp->conn_zoneid);
6922 		}
6923 
6924 		/* If multicast TTL is 0 then we are done */
6925 		if (ipha->ipha_ttl == 0) {
6926 			if (ipif != NULL)
6927 				ipif_refrele(ipif);
6928 			freemsg(mp);
6929 			IRE_REFRELE(ire);
6930 			return;
6931 		}
6932 	}
6933 
6934 	ASSERT(DB_TYPE(ire_fp_mp) == M_DATA);
6935 	mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len;
6936 	bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len);
6937 
6938 	UPDATE_OB_PKT_COUNT(ire);
6939 	ire->ire_last_used_time = lbolt;
6940 
6941 	BUMP_MIB(ill->ill_ip_mib, ipIfStatsHCOutTransmits);
6942 	UPDATE_MIB(ill->ill_ip_mib, ipIfStatsHCOutOctets,
6943 	    ntohs(ipha->ipha_length));
6944 
6945 	if (ILL_DLS_CAPABLE(ill)) {
6946 		/*
6947 		 * Send the packet directly to DLD, where it may be queued
6948 		 * depending on the availability of transmit resources at
6949 		 * the media layer.
6950 		 */
6951 		IP_DLS_ILL_TX(ill, ipha, mp, ipst);
6952 	} else {
6953 		DTRACE_PROBE4(ip4__physical__out__start,
6954 		    ill_t *, NULL, ill_t *, ill,
6955 		    ipha_t *, ipha, mblk_t *, mp);
6956 		FW_HOOKS(ipst->ips_ip4_physical_out_event,
6957 		    ipst->ips_ipv4firewall_physical_out,
6958 		    NULL, ill, ipha, mp, mp, ipst);
6959 		DTRACE_PROBE1(ip4__physical__out__end, mblk_t *, mp);
6960 		if (mp != NULL)
6961 			putnext(ire->ire_stq, mp);
6962 	}
6963 
6964 	if (ipif != NULL)
6965 		ipif_refrele(ipif);
6966 	IRE_REFRELE(ire);
6967 }
6968 
6969 static boolean_t
6970 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst)
6971 {
6972 	udp_t *udp = Q_TO_UDP(wq);
6973 	int err;
6974 	uchar_t opt_storage[TSOL_MAX_IPV6_OPTION];
6975 
6976 	err = tsol_compute_label_v6(DB_CREDDEF(mp, udp->udp_connp->conn_cred),
6977 	    dst, opt_storage, udp->udp_mac_exempt,
6978 	    udp->udp_us->us_netstack->netstack_ip);
6979 	if (err == 0) {
6980 		err = tsol_update_sticky(&udp->udp_sticky_ipp,
6981 		    &udp->udp_label_len_v6, opt_storage);
6982 	}
6983 	if (err != 0) {
6984 		DTRACE_PROBE4(
6985 		    tx__ip__log__drop__updatelabel__udp6,
6986 		    char *, "queue(1) failed to update options(2) on mp(3)",
6987 		    queue_t *, wq, char *, opt_storage, mblk_t *, mp);
6988 	} else {
6989 		udp->udp_v6lastdst = *dst;
6990 	}
6991 	return (err);
6992 }
6993 
6994 /*
6995  * This routine handles all messages passed downstream.  It either
6996  * consumes the message or passes it downstream; it never queues a
6997  * a message.
6998  */
6999 static void
7000 udp_output(conn_t *connp, mblk_t *mp, struct sockaddr *addr, socklen_t addrlen)
7001 {
7002 	sin6_t		*sin6;
7003 	sin_t		*sin;
7004 	ipaddr_t	v4dst;
7005 	uint16_t	port;
7006 	uint_t		srcid;
7007 	queue_t		*q = connp->conn_wq;
7008 	udp_t		*udp = connp->conn_udp;
7009 	int		error = 0;
7010 	struct sockaddr_storage ss;
7011 	udp_stack_t *us = udp->udp_us;
7012 	boolean_t	insert_spi = udp->udp_nat_t_endpoint;
7013 
7014 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START,
7015 	    "udp_wput_start: connp %p mp %p", connp, mp);
7016 
7017 	/*
7018 	 * We directly handle several cases here: T_UNITDATA_REQ message
7019 	 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for both
7020 	 * connected and non-connected socket.  The latter carries the
7021 	 * address structure along when this routine gets called.
7022 	 */
7023 	switch (DB_TYPE(mp)) {
7024 	case M_DATA:
7025 		if (!udp->udp_direct_sockfs || udp->udp_state != TS_DATA_XFER) {
7026 			if (!udp->udp_direct_sockfs ||
7027 			    addr == NULL || addrlen == 0) {
7028 				/* Not connected; address is required */
7029 				BUMP_MIB(&udp->udp_mib, udpOutErrors);
7030 				UDP_STAT(us, udp_out_err_notconn);
7031 				freemsg(mp);
7032 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7033 				    "udp_wput_end: connp %p (%S)", connp,
7034 				    "not-connected; address required");
7035 				return;
7036 			}
7037 			ASSERT(udp->udp_issocket);
7038 			UDP_DBGSTAT(us, udp_data_notconn);
7039 			/* Not connected; do some more checks below */
7040 			break;
7041 		}
7042 		/* M_DATA for connected socket */
7043 		UDP_DBGSTAT(us, udp_data_conn);
7044 		IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst);
7045 
7046 		/* Initialize addr and addrlen as if they're passed in */
7047 		if (udp->udp_family == AF_INET) {
7048 			sin = (sin_t *)&ss;
7049 			sin->sin_family = AF_INET;
7050 			sin->sin_port = udp->udp_dstport;
7051 			sin->sin_addr.s_addr = v4dst;
7052 			addr = (struct sockaddr *)sin;
7053 			addrlen = sizeof (*sin);
7054 		} else {
7055 			sin6 = (sin6_t *)&ss;
7056 			sin6->sin6_family = AF_INET6;
7057 			sin6->sin6_port = udp->udp_dstport;
7058 			sin6->sin6_flowinfo = udp->udp_flowinfo;
7059 			sin6->sin6_addr = udp->udp_v6dst;
7060 			sin6->sin6_scope_id = 0;
7061 			sin6->__sin6_src_id = 0;
7062 			addr = (struct sockaddr *)sin6;
7063 			addrlen = sizeof (*sin6);
7064 		}
7065 
7066 		if (udp->udp_family == AF_INET ||
7067 		    IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst)) {
7068 			/*
7069 			 * Handle both AF_INET and AF_INET6; the latter
7070 			 * for IPV4 mapped destination addresses.  Note
7071 			 * here that both addr and addrlen point to the
7072 			 * corresponding struct depending on the address
7073 			 * family of the socket.
7074 			 */
7075 			mp = udp_output_v4(connp, mp, v4dst,
7076 			    udp->udp_dstport, 0, &error, insert_spi);
7077 		} else {
7078 			mp = udp_output_v6(connp, mp, sin6, &error);
7079 		}
7080 		if (error != 0) {
7081 			ASSERT(addr != NULL && addrlen != 0);
7082 			goto ud_error;
7083 		}
7084 		return;
7085 	case M_PROTO:
7086 	case M_PCPROTO: {
7087 		struct T_unitdata_req *tudr;
7088 
7089 		ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX);
7090 		tudr = (struct T_unitdata_req *)mp->b_rptr;
7091 
7092 		/* Handle valid T_UNITDATA_REQ here */
7093 		if (MBLKL(mp) >= sizeof (*tudr) &&
7094 		    ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) {
7095 			if (mp->b_cont == NULL) {
7096 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7097 				    "udp_wput_end: q %p (%S)", q, "badaddr");
7098 				error = EPROTO;
7099 				goto ud_error;
7100 			}
7101 
7102 			if (!MBLKIN(mp, 0, tudr->DEST_offset +
7103 			    tudr->DEST_length)) {
7104 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7105 				    "udp_wput_end: q %p (%S)", q, "badaddr");
7106 				error = EADDRNOTAVAIL;
7107 				goto ud_error;
7108 			}
7109 			/*
7110 			 * If a port has not been bound to the stream, fail.
7111 			 * This is not a problem when sockfs is directly
7112 			 * above us, because it will ensure that the socket
7113 			 * is first bound before allowing data to be sent.
7114 			 */
7115 			if (udp->udp_state == TS_UNBND) {
7116 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7117 				    "udp_wput_end: q %p (%S)", q, "outstate");
7118 				error = EPROTO;
7119 				goto ud_error;
7120 			}
7121 			addr = (struct sockaddr *)
7122 			    &mp->b_rptr[tudr->DEST_offset];
7123 			addrlen = tudr->DEST_length;
7124 			if (tudr->OPT_length != 0)
7125 				UDP_STAT(us, udp_out_opt);
7126 			break;
7127 		}
7128 		/* FALLTHRU */
7129 	}
7130 	default:
7131 		udp_become_writer(connp, mp, udp_wput_other_wrapper,
7132 		    SQTAG_UDP_OUTPUT);
7133 		return;
7134 	}
7135 	ASSERT(addr != NULL);
7136 
7137 	switch (udp->udp_family) {
7138 	case AF_INET6:
7139 		sin6 = (sin6_t *)addr;
7140 		if (!OK_32PTR((char *)sin6) || addrlen != sizeof (sin6_t) ||
7141 		    sin6->sin6_family != AF_INET6) {
7142 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7143 			    "udp_wput_end: q %p (%S)", q, "badaddr");
7144 			error = EADDRNOTAVAIL;
7145 			goto ud_error;
7146 		}
7147 
7148 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
7149 			/*
7150 			 * Destination is a non-IPv4-compatible IPv6 address.
7151 			 * Send out an IPv6 format packet.
7152 			 */
7153 			mp = udp_output_v6(connp, mp, sin6, &error);
7154 			if (error != 0)
7155 				goto ud_error;
7156 
7157 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7158 			    "udp_wput_end: q %p (%S)", q, "udp_output_v6");
7159 			return;
7160 		}
7161 		/*
7162 		 * If the local address is not zero or a mapped address
7163 		 * return an error.  It would be possible to send an IPv4
7164 		 * packet but the response would never make it back to the
7165 		 * application since it is bound to a non-mapped address.
7166 		 */
7167 		if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) &&
7168 		    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
7169 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7170 			    "udp_wput_end: q %p (%S)", q, "badaddr");
7171 			error = EADDRNOTAVAIL;
7172 			goto ud_error;
7173 		}
7174 		/* Send IPv4 packet without modifying udp_ipversion */
7175 		/* Extract port and ipaddr */
7176 		port = sin6->sin6_port;
7177 		IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst);
7178 		srcid = sin6->__sin6_src_id;
7179 		break;
7180 
7181 	case AF_INET:
7182 		sin = (sin_t *)addr;
7183 		if (!OK_32PTR((char *)sin) || addrlen != sizeof (sin_t) ||
7184 		    sin->sin_family != AF_INET) {
7185 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END,
7186 			    "udp_wput_end: q %p (%S)", q, "badaddr");
7187 			error = EADDRNOTAVAIL;
7188 			goto ud_error;
7189 		}
7190 		/* Extract port and ipaddr */
7191 		port = sin->sin_port;
7192 		v4dst = sin->sin_addr.s_addr;
7193 		srcid = 0;
7194 		break;
7195 	}
7196 
7197 	mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error, insert_spi);
7198 	if (error != 0) {
7199 ud_error:
7200 		UDP_STAT(us, udp_out_err_output);
7201 		ASSERT(mp != NULL);
7202 		/* mp is freed by the following routine */
7203 		udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen,
7204 		    (t_scalar_t)error);
7205 	}
7206 }
7207 
7208 /* ARGSUSED */
7209 static void
7210 udp_output_wrapper(void *arg, mblk_t *mp, void *arg2)
7211 {
7212 	udp_output((conn_t *)arg, mp, NULL, 0);
7213 	_UDP_EXIT((conn_t *)arg);
7214 }
7215 
7216 static void
7217 udp_wput(queue_t *q, mblk_t *mp)
7218 {
7219 	_UDP_ENTER(Q_TO_CONN(UDP_WR(q)), mp, udp_output_wrapper,
7220 	    SQTAG_UDP_WPUT);
7221 }
7222 
7223 /*
7224  * Allocate and prepare a T_UNITDATA_REQ message.
7225  */
7226 static mblk_t *
7227 udp_tudr_alloc(struct sockaddr *addr, socklen_t addrlen)
7228 {
7229 	struct T_unitdata_req *tudr;
7230 	mblk_t *mp;
7231 
7232 	mp = allocb(sizeof (*tudr) + addrlen, BPRI_MED);
7233 	if (mp != NULL) {
7234 		mp->b_wptr += sizeof (*tudr) + addrlen;
7235 		DB_TYPE(mp) = M_PROTO;
7236 
7237 		tudr = (struct T_unitdata_req *)mp->b_rptr;
7238 		tudr->PRIM_type = T_UNITDATA_REQ;
7239 		tudr->DEST_length = addrlen;
7240 		tudr->DEST_offset = (t_scalar_t)sizeof (*tudr);
7241 		tudr->OPT_length = 0;
7242 		tudr->OPT_offset = 0;
7243 		bcopy(addr, tudr+1, addrlen);
7244 	}
7245 	return (mp);
7246 }
7247 
7248 /*
7249  * Entry point for sockfs when udp is in "direct sockfs" mode.  This mode
7250  * is valid when we are directly beneath the stream head, and thus sockfs
7251  * is able to bypass STREAMS and directly call us, passing along the sockaddr
7252  * structure without the cumbersome T_UNITDATA_REQ interface.  Note that
7253  * this is done for both connected and non-connected endpoint.
7254  */
7255 void
7256 udp_wput_data(queue_t *q, mblk_t *mp, struct sockaddr *addr, socklen_t addrlen)
7257 {
7258 	conn_t	*connp;
7259 	udp_t	*udp;
7260 	udp_stack_t *us;
7261 
7262 	q = UDP_WR(q);
7263 	connp = Q_TO_CONN(q);
7264 	udp = connp->conn_udp;
7265 	us = udp->udp_us;
7266 
7267 	/* udpsockfs should only send down M_DATA for this entry point */
7268 	ASSERT(DB_TYPE(mp) == M_DATA);
7269 
7270 	mutex_enter(&connp->conn_lock);
7271 	UDP_MODE_ASSERTIONS(udp, UDP_ENTER);
7272 
7273 	if (udp->udp_mode != UDP_MT_HOT) {
7274 		/*
7275 		 * We can't enter this conn right away because another
7276 		 * thread is currently executing as writer; therefore we
7277 		 * need to deposit the message into the squeue to be
7278 		 * drained later.  If a socket address is present, we
7279 		 * need to create a T_UNITDATA_REQ message as placeholder.
7280 		 */
7281 		if (addr != NULL && addrlen != 0) {
7282 			mblk_t *tudr_mp = udp_tudr_alloc(addr, addrlen);
7283 
7284 			if (tudr_mp == NULL) {
7285 				mutex_exit(&connp->conn_lock);
7286 				BUMP_MIB(&udp->udp_mib, udpOutErrors);
7287 				UDP_STAT(us, udp_out_err_tudr);
7288 				freemsg(mp);
7289 				return;
7290 			}
7291 			/* Tag the packet with T_UNITDATA_REQ */
7292 			tudr_mp->b_cont = mp;
7293 			mp = tudr_mp;
7294 		}
7295 		mutex_exit(&connp->conn_lock);
7296 		udp_enter(connp, mp, udp_output_wrapper, SQTAG_UDP_WPUT);
7297 		return;
7298 	}
7299 
7300 	/* We can execute as reader right away. */
7301 	UDP_READERS_INCREF(udp);
7302 	mutex_exit(&connp->conn_lock);
7303 
7304 	udp_output(connp, mp, addr, addrlen);
7305 
7306 	udp_exit(connp);
7307 }
7308 
7309 /*
7310  * udp_output_v6():
7311  * Assumes that udp_wput did some sanity checking on the destination
7312  * address.
7313  */
7314 static mblk_t *
7315 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error)
7316 {
7317 	ip6_t		*ip6h;
7318 	ip6i_t		*ip6i;	/* mp1->b_rptr even if no ip6i_t */
7319 	mblk_t		*mp1 = mp;
7320 	mblk_t		*mp2;
7321 	int		udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE;
7322 	size_t		ip_len;
7323 	udpha_t		*udph;
7324 	udp_t		*udp = connp->conn_udp;
7325 	queue_t		*q = connp->conn_wq;
7326 	ip6_pkt_t	ipp_s;	/* For ancillary data options */
7327 	ip6_pkt_t	*ipp = &ipp_s;
7328 	ip6_pkt_t	*tipp;	/* temporary ipp */
7329 	uint32_t	csum = 0;
7330 	uint_t		ignore = 0;
7331 	uint_t		option_exists = 0, is_sticky = 0;
7332 	uint8_t		*cp;
7333 	uint8_t		*nxthdr_ptr;
7334 	in6_addr_t	ip6_dst;
7335 	udpattrs_t	attrs;
7336 	boolean_t	opt_present;
7337 	ip6_hbh_t	*hopoptsptr = NULL;
7338 	uint_t		hopoptslen = 0;
7339 	boolean_t	is_ancillary = B_FALSE;
7340 	udp_stack_t	*us = udp->udp_us;
7341 
7342 	*error = 0;
7343 
7344 	/*
7345 	 * If the local address is a mapped address return
7346 	 * an error.
7347 	 * It would be possible to send an IPv6 packet but the
7348 	 * response would never make it back to the application
7349 	 * since it is bound to a mapped address.
7350 	 */
7351 	if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) {
7352 		*error = EADDRNOTAVAIL;
7353 		goto done;
7354 	}
7355 
7356 	ipp->ipp_fields = 0;
7357 	ipp->ipp_sticky_ignored = 0;
7358 
7359 	/*
7360 	 * If TPI options passed in, feed it for verification and handling
7361 	 */
7362 	attrs.udpattr_credset = B_FALSE;
7363 	opt_present = B_FALSE;
7364 	if (DB_TYPE(mp) != M_DATA) {
7365 		mp1 = mp->b_cont;
7366 		if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) {
7367 			attrs.udpattr_ipp6 = ipp;
7368 			attrs.udpattr_mb = mp;
7369 			if (udp_unitdata_opt_process(q, mp, error, &attrs) < 0)
7370 				goto done;
7371 			ASSERT(*error == 0);
7372 			opt_present = B_TRUE;
7373 		}
7374 	}
7375 	ignore = ipp->ipp_sticky_ignored;
7376 
7377 	/* mp1 points to the M_DATA mblk carrying the packet */
7378 	ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA);
7379 
7380 	if (sin6->sin6_scope_id != 0 &&
7381 	    IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
7382 		/*
7383 		 * IPPF_SCOPE_ID is special.  It's neither a sticky
7384 		 * option nor ancillary data.  It needs to be
7385 		 * explicitly set in options_exists.
7386 		 */
7387 		option_exists |= IPPF_SCOPE_ID;
7388 	}
7389 
7390 	/*
7391 	 * Compute the destination address
7392 	 */
7393 	ip6_dst = sin6->sin6_addr;
7394 	if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
7395 		ip6_dst = ipv6_loopback;
7396 
7397 	/*
7398 	 * If we're not going to the same destination as last time, then
7399 	 * recompute the label required.  This is done in a separate routine to
7400 	 * avoid blowing up our stack here.
7401 	 *
7402 	 * TSOL Note: Since we are not in WRITER mode, UDP packets
7403 	 * to different destination may require different labels,
7404 	 * or worse, UDP packets to same IP address may require
7405 	 * different labels due to use of shared all-zones address.
7406 	 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts,
7407 	 * and sticky ipp_hopoptslen are consistent for the current
7408 	 * destination and are updated atomically.
7409 	 */
7410 	mutex_enter(&connp->conn_lock);
7411 	if (is_system_labeled()) {
7412 		/* Using UDP MLP requires SCM_UCRED from user */
7413 		if (connp->conn_mlp_type != mlptSingle &&
7414 		    !attrs.udpattr_credset) {
7415 			DTRACE_PROBE4(
7416 			    tx__ip__log__info__output__udp6,
7417 			    char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)",
7418 			    mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q);
7419 			*error = ECONNREFUSED;
7420 			mutex_exit(&connp->conn_lock);
7421 			goto done;
7422 		}
7423 		/*
7424 		 * update label option for this UDP socket if
7425 		 * - the destination has changed, or
7426 		 * - the UDP socket is MLP
7427 		 */
7428 		if ((opt_present ||
7429 		    !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst) ||
7430 		    connp->conn_mlp_type != mlptSingle) &&
7431 		    (*error = udp_update_label_v6(q, mp, &ip6_dst)) != 0) {
7432 			mutex_exit(&connp->conn_lock);
7433 			goto done;
7434 		}
7435 	}
7436 
7437 	/*
7438 	 * If there's a security label here, then we ignore any options the
7439 	 * user may try to set.  We keep the peer's label as a hidden sticky
7440 	 * option. We make a private copy of this label before releasing the
7441 	 * lock so that label is kept consistent with the destination addr.
7442 	 */
7443 	if (udp->udp_label_len_v6 > 0) {
7444 		ignore &= ~IPPF_HOPOPTS;
7445 		ipp->ipp_fields &= ~IPPF_HOPOPTS;
7446 	}
7447 
7448 	if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) {
7449 		/* No sticky options nor ancillary data. */
7450 		mutex_exit(&connp->conn_lock);
7451 		goto no_options;
7452 	}
7453 
7454 	/*
7455 	 * Go through the options figuring out where each is going to
7456 	 * come from and build two masks.  The first mask indicates if
7457 	 * the option exists at all.  The second mask indicates if the
7458 	 * option is sticky or ancillary.
7459 	 */
7460 	if (!(ignore & IPPF_HOPOPTS)) {
7461 		if (ipp->ipp_fields & IPPF_HOPOPTS) {
7462 			option_exists |= IPPF_HOPOPTS;
7463 			udp_ip_hdr_len += ipp->ipp_hopoptslen;
7464 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) {
7465 			option_exists |= IPPF_HOPOPTS;
7466 			is_sticky |= IPPF_HOPOPTS;
7467 			ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0);
7468 			hopoptsptr = kmem_alloc(
7469 			    udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP);
7470 			if (hopoptsptr == NULL) {
7471 				*error = ENOMEM;
7472 				mutex_exit(&connp->conn_lock);
7473 				goto done;
7474 			}
7475 			hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen;
7476 			bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr,
7477 			    hopoptslen);
7478 			udp_ip_hdr_len += hopoptslen;
7479 		}
7480 	}
7481 	mutex_exit(&connp->conn_lock);
7482 
7483 	if (!(ignore & IPPF_RTHDR)) {
7484 		if (ipp->ipp_fields & IPPF_RTHDR) {
7485 			option_exists |= IPPF_RTHDR;
7486 			udp_ip_hdr_len += ipp->ipp_rthdrlen;
7487 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) {
7488 			option_exists |= IPPF_RTHDR;
7489 			is_sticky |= IPPF_RTHDR;
7490 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen;
7491 		}
7492 	}
7493 
7494 	if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) {
7495 		if (ipp->ipp_fields & IPPF_RTDSTOPTS) {
7496 			option_exists |= IPPF_RTDSTOPTS;
7497 			udp_ip_hdr_len += ipp->ipp_rtdstoptslen;
7498 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) {
7499 			option_exists |= IPPF_RTDSTOPTS;
7500 			is_sticky |= IPPF_RTDSTOPTS;
7501 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen;
7502 		}
7503 	}
7504 
7505 	if (!(ignore & IPPF_DSTOPTS)) {
7506 		if (ipp->ipp_fields & IPPF_DSTOPTS) {
7507 			option_exists |= IPPF_DSTOPTS;
7508 			udp_ip_hdr_len += ipp->ipp_dstoptslen;
7509 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) {
7510 			option_exists |= IPPF_DSTOPTS;
7511 			is_sticky |= IPPF_DSTOPTS;
7512 			udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen;
7513 		}
7514 	}
7515 
7516 	if (!(ignore & IPPF_IFINDEX)) {
7517 		if (ipp->ipp_fields & IPPF_IFINDEX) {
7518 			option_exists |= IPPF_IFINDEX;
7519 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) {
7520 			option_exists |= IPPF_IFINDEX;
7521 			is_sticky |= IPPF_IFINDEX;
7522 		}
7523 	}
7524 
7525 	if (!(ignore & IPPF_ADDR)) {
7526 		if (ipp->ipp_fields & IPPF_ADDR) {
7527 			option_exists |= IPPF_ADDR;
7528 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) {
7529 			option_exists |= IPPF_ADDR;
7530 			is_sticky |= IPPF_ADDR;
7531 		}
7532 	}
7533 
7534 	if (!(ignore & IPPF_DONTFRAG)) {
7535 		if (ipp->ipp_fields & IPPF_DONTFRAG) {
7536 			option_exists |= IPPF_DONTFRAG;
7537 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) {
7538 			option_exists |= IPPF_DONTFRAG;
7539 			is_sticky |= IPPF_DONTFRAG;
7540 		}
7541 	}
7542 
7543 	if (!(ignore & IPPF_USE_MIN_MTU)) {
7544 		if (ipp->ipp_fields & IPPF_USE_MIN_MTU) {
7545 			option_exists |= IPPF_USE_MIN_MTU;
7546 		} else if (udp->udp_sticky_ipp.ipp_fields &
7547 		    IPPF_USE_MIN_MTU) {
7548 			option_exists |= IPPF_USE_MIN_MTU;
7549 			is_sticky |= IPPF_USE_MIN_MTU;
7550 		}
7551 	}
7552 
7553 	if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT))
7554 		option_exists |= IPPF_HOPLIMIT;
7555 	/* IPV6_HOPLIMIT can never be sticky */
7556 	ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT));
7557 
7558 	if (!(ignore & IPPF_UNICAST_HOPS) &&
7559 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) {
7560 		option_exists |= IPPF_UNICAST_HOPS;
7561 		is_sticky |= IPPF_UNICAST_HOPS;
7562 	}
7563 
7564 	if (!(ignore & IPPF_MULTICAST_HOPS) &&
7565 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) {
7566 		option_exists |= IPPF_MULTICAST_HOPS;
7567 		is_sticky |= IPPF_MULTICAST_HOPS;
7568 	}
7569 
7570 	if (!(ignore & IPPF_TCLASS)) {
7571 		if (ipp->ipp_fields & IPPF_TCLASS) {
7572 			option_exists |= IPPF_TCLASS;
7573 		} else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) {
7574 			option_exists |= IPPF_TCLASS;
7575 			is_sticky |= IPPF_TCLASS;
7576 		}
7577 	}
7578 
7579 	if (!(ignore & IPPF_NEXTHOP) &&
7580 	    (udp->udp_sticky_ipp.ipp_fields & IPPF_NEXTHOP)) {
7581 		option_exists |= IPPF_NEXTHOP;
7582 		is_sticky |= IPPF_NEXTHOP;
7583 	}
7584 
7585 no_options:
7586 
7587 	/*
7588 	 * If any options carried in the ip6i_t were specified, we
7589 	 * need to account for the ip6i_t in the data we'll be sending
7590 	 * down.
7591 	 */
7592 	if (option_exists & IPPF_HAS_IP6I)
7593 		udp_ip_hdr_len += sizeof (ip6i_t);
7594 
7595 	/* check/fix buffer config, setup pointers into it */
7596 	ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len];
7597 	if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) ||
7598 	    !OK_32PTR(ip6h)) {
7599 		/* Try to get everything in a single mblk next time */
7600 		if (udp_ip_hdr_len > udp->udp_max_hdr_len) {
7601 			udp->udp_max_hdr_len = udp_ip_hdr_len;
7602 			(void) mi_set_sth_wroff(UDP_RD(q),
7603 			    udp->udp_max_hdr_len + us->us_wroff_extra);
7604 		}
7605 		mp2 = allocb(udp_ip_hdr_len + us->us_wroff_extra, BPRI_LO);
7606 		if (mp2 == NULL) {
7607 			*error = ENOMEM;
7608 			goto done;
7609 		}
7610 		mp2->b_wptr = DB_LIM(mp2);
7611 		mp2->b_cont = mp1;
7612 		mp1 = mp2;
7613 		if (DB_TYPE(mp) != M_DATA)
7614 			mp->b_cont = mp1;
7615 		else
7616 			mp = mp1;
7617 
7618 		ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len);
7619 	}
7620 	mp1->b_rptr = (unsigned char *)ip6h;
7621 	ip6i = (ip6i_t *)ip6h;
7622 
7623 #define	ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp)
7624 	if (option_exists & IPPF_HAS_IP6I) {
7625 		ip6h = (ip6_t *)&ip6i[1];
7626 		ip6i->ip6i_flags = 0;
7627 		ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
7628 
7629 		/* sin6_scope_id takes precendence over IPPF_IFINDEX */
7630 		if (option_exists & IPPF_SCOPE_ID) {
7631 			ip6i->ip6i_flags |= IP6I_IFINDEX;
7632 			ip6i->ip6i_ifindex = sin6->sin6_scope_id;
7633 		} else if (option_exists & IPPF_IFINDEX) {
7634 			tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX);
7635 			ASSERT(tipp->ipp_ifindex != 0);
7636 			ip6i->ip6i_flags |= IP6I_IFINDEX;
7637 			ip6i->ip6i_ifindex = tipp->ipp_ifindex;
7638 		}
7639 
7640 		if (option_exists & IPPF_ADDR) {
7641 			/*
7642 			 * Enable per-packet source address verification if
7643 			 * IPV6_PKTINFO specified the source address.
7644 			 * ip6_src is set in the transport's _wput function.
7645 			 */
7646 			ip6i->ip6i_flags |= IP6I_VERIFY_SRC;
7647 		}
7648 
7649 		if (option_exists & IPPF_DONTFRAG) {
7650 			ip6i->ip6i_flags |= IP6I_DONTFRAG;
7651 		}
7652 
7653 		if (option_exists & IPPF_USE_MIN_MTU) {
7654 			ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU(
7655 			    ip6i->ip6i_flags, ipp->ipp_use_min_mtu);
7656 		}
7657 
7658 		if (option_exists & IPPF_NEXTHOP) {
7659 			tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP);
7660 			ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop));
7661 			ip6i->ip6i_flags |= IP6I_NEXTHOP;
7662 			ip6i->ip6i_nexthop = tipp->ipp_nexthop;
7663 		}
7664 
7665 		/*
7666 		 * tell IP this is an ip6i_t private header
7667 		 */
7668 		ip6i->ip6i_nxt = IPPROTO_RAW;
7669 	}
7670 
7671 	/* Initialize IPv6 header */
7672 	ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW;
7673 	bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src));
7674 
7675 	/* Set the hoplimit of the outgoing packet. */
7676 	if (option_exists & IPPF_HOPLIMIT) {
7677 		/* IPV6_HOPLIMIT ancillary data overrides all other settings. */
7678 		ip6h->ip6_hops = ipp->ipp_hoplimit;
7679 		ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7680 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
7681 		ip6h->ip6_hops = udp->udp_multicast_ttl;
7682 		if (option_exists & IPPF_MULTICAST_HOPS)
7683 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7684 	} else {
7685 		ip6h->ip6_hops = udp->udp_ttl;
7686 		if (option_exists & IPPF_UNICAST_HOPS)
7687 			ip6i->ip6i_flags |= IP6I_HOPLIMIT;
7688 	}
7689 
7690 	if (option_exists & IPPF_ADDR) {
7691 		tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR);
7692 		ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr));
7693 		ip6h->ip6_src = tipp->ipp_addr;
7694 	} else {
7695 		/*
7696 		 * The source address was not set using IPV6_PKTINFO.
7697 		 * First look at the bound source.
7698 		 * If unspecified fallback to __sin6_src_id.
7699 		 */
7700 		ip6h->ip6_src = udp->udp_v6src;
7701 		if (sin6->__sin6_src_id != 0 &&
7702 		    IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) {
7703 			ip_srcid_find_id(sin6->__sin6_src_id,
7704 			    &ip6h->ip6_src, connp->conn_zoneid,
7705 			    us->us_netstack);
7706 		}
7707 	}
7708 
7709 	nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt;
7710 	cp = (uint8_t *)&ip6h[1];
7711 
7712 	/*
7713 	 * Here's where we have to start stringing together
7714 	 * any extension headers in the right order:
7715 	 * Hop-by-hop, destination, routing, and final destination opts.
7716 	 */
7717 	if (option_exists & IPPF_HOPOPTS) {
7718 		/* Hop-by-hop options */
7719 		ip6_hbh_t *hbh = (ip6_hbh_t *)cp;
7720 		tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS);
7721 		if (hopoptslen == 0) {
7722 			hopoptsptr = tipp->ipp_hopopts;
7723 			hopoptslen = tipp->ipp_hopoptslen;
7724 			is_ancillary = B_TRUE;
7725 		}
7726 
7727 		*nxthdr_ptr = IPPROTO_HOPOPTS;
7728 		nxthdr_ptr = &hbh->ip6h_nxt;
7729 
7730 		bcopy(hopoptsptr, cp, hopoptslen);
7731 		cp += hopoptslen;
7732 
7733 		if (hopoptsptr != NULL && !is_ancillary) {
7734 			kmem_free(hopoptsptr, hopoptslen);
7735 			hopoptsptr = NULL;
7736 			hopoptslen = 0;
7737 		}
7738 	}
7739 	/*
7740 	 * En-route destination options
7741 	 * Only do them if there's a routing header as well
7742 	 */
7743 	if (option_exists & IPPF_RTDSTOPTS) {
7744 		ip6_dest_t *dst = (ip6_dest_t *)cp;
7745 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS);
7746 
7747 		*nxthdr_ptr = IPPROTO_DSTOPTS;
7748 		nxthdr_ptr = &dst->ip6d_nxt;
7749 
7750 		bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen);
7751 		cp += tipp->ipp_rtdstoptslen;
7752 	}
7753 	/*
7754 	 * Routing header next
7755 	 */
7756 	if (option_exists & IPPF_RTHDR) {
7757 		ip6_rthdr_t *rt = (ip6_rthdr_t *)cp;
7758 		tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR);
7759 
7760 		*nxthdr_ptr = IPPROTO_ROUTING;
7761 		nxthdr_ptr = &rt->ip6r_nxt;
7762 
7763 		bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen);
7764 		cp += tipp->ipp_rthdrlen;
7765 	}
7766 	/*
7767 	 * Do ultimate destination options
7768 	 */
7769 	if (option_exists & IPPF_DSTOPTS) {
7770 		ip6_dest_t *dest = (ip6_dest_t *)cp;
7771 		tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS);
7772 
7773 		*nxthdr_ptr = IPPROTO_DSTOPTS;
7774 		nxthdr_ptr = &dest->ip6d_nxt;
7775 
7776 		bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen);
7777 		cp += tipp->ipp_dstoptslen;
7778 	}
7779 	/*
7780 	 * Now set the last header pointer to the proto passed in
7781 	 */
7782 	ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE));
7783 	*nxthdr_ptr = IPPROTO_UDP;
7784 
7785 	/* Update UDP header */
7786 	udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE);
7787 	udph->uha_dst_port = sin6->sin6_port;
7788 	udph->uha_src_port = udp->udp_port;
7789 
7790 	/*
7791 	 * Copy in the destination address
7792 	 */
7793 	ip6h->ip6_dst = ip6_dst;
7794 
7795 	ip6h->ip6_vcf =
7796 	    (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) |
7797 	    (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK);
7798 
7799 	if (option_exists & IPPF_TCLASS) {
7800 		tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS);
7801 		ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf,
7802 		    tipp->ipp_tclass);
7803 	}
7804 
7805 	if (option_exists & IPPF_RTHDR) {
7806 		ip6_rthdr_t	*rth;
7807 
7808 		/*
7809 		 * Perform any processing needed for source routing.
7810 		 * We know that all extension headers will be in the same mblk
7811 		 * as the IPv6 header.
7812 		 */
7813 		rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr);
7814 		if (rth != NULL && rth->ip6r_segleft != 0) {
7815 			if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) {
7816 				/*
7817 				 * Drop packet - only support Type 0 routing.
7818 				 * Notify the application as well.
7819 				 */
7820 				*error = EPROTO;
7821 				goto done;
7822 			}
7823 
7824 			/*
7825 			 * rth->ip6r_len is twice the number of
7826 			 * addresses in the header. Thus it must be even.
7827 			 */
7828 			if (rth->ip6r_len & 0x1) {
7829 				*error = EPROTO;
7830 				goto done;
7831 			}
7832 			/*
7833 			 * Shuffle the routing header and ip6_dst
7834 			 * addresses, and get the checksum difference
7835 			 * between the first hop (in ip6_dst) and
7836 			 * the destination (in the last routing hdr entry).
7837 			 */
7838 			csum = ip_massage_options_v6(ip6h, rth,
7839 			    us->us_netstack);
7840 			/*
7841 			 * Verify that the first hop isn't a mapped address.
7842 			 * Routers along the path need to do this verification
7843 			 * for subsequent hops.
7844 			 */
7845 			if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) {
7846 				*error = EADDRNOTAVAIL;
7847 				goto done;
7848 			}
7849 
7850 			cp += (rth->ip6r_len + 1)*8;
7851 		}
7852 	}
7853 
7854 	/* count up length of UDP packet */
7855 	ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN;
7856 	if ((mp2 = mp1->b_cont) != NULL) {
7857 		do {
7858 			ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX);
7859 			ip_len += (uint32_t)MBLKL(mp2);
7860 		} while ((mp2 = mp2->b_cont) != NULL);
7861 	}
7862 
7863 	/*
7864 	 * If the size of the packet is greater than the maximum allowed by
7865 	 * ip, return an error. Passing this down could cause panics because
7866 	 * the size will have wrapped and be inconsistent with the msg size.
7867 	 */
7868 	if (ip_len > IP_MAXPACKET) {
7869 		*error = EMSGSIZE;
7870 		goto done;
7871 	}
7872 
7873 	/* Store the UDP length. Subtract length of extension hdrs */
7874 	udph->uha_length = htons(ip_len + IPV6_HDR_LEN -
7875 	    (int)((uchar_t *)udph - (uchar_t *)ip6h));
7876 
7877 	/*
7878 	 * We make it easy for IP to include our pseudo header
7879 	 * by putting our length in uh_checksum, modified (if
7880 	 * we have a routing header) by the checksum difference
7881 	 * between the ultimate destination and first hop addresses.
7882 	 * Note: UDP over IPv6 must always checksum the packet.
7883 	 */
7884 	csum += udph->uha_length;
7885 	csum = (csum & 0xFFFF) + (csum >> 16);
7886 	udph->uha_checksum = (uint16_t)csum;
7887 
7888 #ifdef _LITTLE_ENDIAN
7889 	ip_len = htons(ip_len);
7890 #endif
7891 	ip6h->ip6_plen = ip_len;
7892 	if (DB_CRED(mp) != NULL)
7893 		mblk_setcred(mp1, DB_CRED(mp));
7894 
7895 	if (DB_TYPE(mp) != M_DATA) {
7896 		ASSERT(mp != mp1);
7897 		freeb(mp);
7898 	}
7899 
7900 	/* mp has been consumed and we'll return success */
7901 	ASSERT(*error == 0);
7902 	mp = NULL;
7903 
7904 	/* We're done. Pass the packet to IP */
7905 	BUMP_MIB(&udp->udp_mib, udpHCOutDatagrams);
7906 	ip_output_v6(connp, mp1, q, IP_WPUT);
7907 
7908 done:
7909 	if (hopoptsptr != NULL && !is_ancillary) {
7910 		kmem_free(hopoptsptr, hopoptslen);
7911 		hopoptsptr = NULL;
7912 	}
7913 	if (*error != 0) {
7914 		ASSERT(mp != NULL);
7915 		BUMP_MIB(&udp->udp_mib, udpOutErrors);
7916 	}
7917 	return (mp);
7918 }
7919 
7920 static void
7921 udp_wput_other(queue_t *q, mblk_t *mp)
7922 {
7923 	uchar_t	*rptr = mp->b_rptr;
7924 	struct datab *db;
7925 	struct iocblk *iocp;
7926 	cred_t	*cr;
7927 	conn_t	*connp = Q_TO_CONN(q);
7928 	udp_t	*udp = connp->conn_udp;
7929 	udp_stack_t *us;
7930 
7931 	TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START,
7932 	    "udp_wput_other_start: q %p", q);
7933 
7934 	us = udp->udp_us;
7935 	db = mp->b_datap;
7936 
7937 	cr = DB_CREDDEF(mp, connp->conn_cred);
7938 
7939 	switch (db->db_type) {
7940 	case M_PROTO:
7941 	case M_PCPROTO:
7942 		if (mp->b_wptr - rptr < sizeof (t_scalar_t)) {
7943 			freemsg(mp);
7944 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7945 			    "udp_wput_other_end: q %p (%S)", q, "protoshort");
7946 			return;
7947 		}
7948 		switch (((t_primp_t)rptr)->type) {
7949 		case T_ADDR_REQ:
7950 			udp_addr_req(q, mp);
7951 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7952 			    "udp_wput_other_end: q %p (%S)", q, "addrreq");
7953 			return;
7954 		case O_T_BIND_REQ:
7955 		case T_BIND_REQ:
7956 			udp_bind(q, mp);
7957 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7958 			    "udp_wput_other_end: q %p (%S)", q, "bindreq");
7959 			return;
7960 		case T_CONN_REQ:
7961 			udp_connect(q, mp);
7962 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7963 			    "udp_wput_other_end: q %p (%S)", q, "connreq");
7964 			return;
7965 		case T_CAPABILITY_REQ:
7966 			udp_capability_req(q, mp);
7967 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7968 			    "udp_wput_other_end: q %p (%S)", q, "capabreq");
7969 			return;
7970 		case T_INFO_REQ:
7971 			udp_info_req(q, mp);
7972 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7973 			    "udp_wput_other_end: q %p (%S)", q, "inforeq");
7974 			return;
7975 		case T_UNITDATA_REQ:
7976 			/*
7977 			 * If a T_UNITDATA_REQ gets here, the address must
7978 			 * be bad.  Valid T_UNITDATA_REQs are handled
7979 			 * in udp_wput.
7980 			 */
7981 			udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL);
7982 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7983 			    "udp_wput_other_end: q %p (%S)", q, "unitdatareq");
7984 			return;
7985 		case T_UNBIND_REQ:
7986 			udp_unbind(q, mp);
7987 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
7988 			    "udp_wput_other_end: q %p (%S)", q, "unbindreq");
7989 			return;
7990 		case T_SVR4_OPTMGMT_REQ:
7991 			if (!snmpcom_req(q, mp, udp_snmp_set, udp_snmp_get, cr))
7992 				/*
7993 				 * Use upper queue for option processing in
7994 				 * case the request is not handled at this
7995 				 * level and needs to be passed down to IP.
7996 				 */
7997 				(void) svr4_optcom_req(_WR(UDP_RD(q)),
7998 				    mp, cr, &udp_opt_obj);
7999 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8000 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
8001 			return;
8002 
8003 		case T_OPTMGMT_REQ:
8004 			/*
8005 			 * Use upper queue for option processing in
8006 			 * case the request is not handled at this
8007 			 * level and needs to be passed down to IP.
8008 			 */
8009 			(void) tpi_optcom_req(_WR(UDP_RD(q)),
8010 			    mp, cr, &udp_opt_obj);
8011 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8012 			    "udp_wput_other_end: q %p (%S)", q, "optmgmtreq");
8013 			return;
8014 
8015 		case T_DISCON_REQ:
8016 			udp_disconnect(q, mp);
8017 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8018 			    "udp_wput_other_end: q %p (%S)", q, "disconreq");
8019 			return;
8020 
8021 		/* The following TPI message is not supported by udp. */
8022 		case O_T_CONN_RES:
8023 		case T_CONN_RES:
8024 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
8025 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8026 			    "udp_wput_other_end: q %p (%S)", q,
8027 			    "connres/disconreq");
8028 			return;
8029 
8030 		/* The following 3 TPI messages are illegal for udp. */
8031 		case T_DATA_REQ:
8032 		case T_EXDATA_REQ:
8033 		case T_ORDREL_REQ:
8034 			udp_err_ack(q, mp, TNOTSUPPORT, 0);
8035 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8036 			    "udp_wput_other_end: q %p (%S)", q,
8037 			    "data/exdata/ordrel");
8038 			return;
8039 		default:
8040 			break;
8041 		}
8042 		break;
8043 	case M_FLUSH:
8044 		if (*rptr & FLUSHW)
8045 			flushq(q, FLUSHDATA);
8046 		break;
8047 	case M_IOCTL:
8048 		iocp = (struct iocblk *)mp->b_rptr;
8049 		switch (iocp->ioc_cmd) {
8050 		case TI_GETPEERNAME:
8051 			if (udp->udp_state != TS_DATA_XFER) {
8052 				/*
8053 				 * If a default destination address has not
8054 				 * been associated with the stream, then we
8055 				 * don't know the peer's name.
8056 				 */
8057 				iocp->ioc_error = ENOTCONN;
8058 				iocp->ioc_count = 0;
8059 				mp->b_datap->db_type = M_IOCACK;
8060 				putnext(UDP_RD(q), mp);
8061 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8062 				    "udp_wput_other_end: q %p (%S)", q,
8063 				    "getpeername");
8064 				return;
8065 			}
8066 			/* FALLTHRU */
8067 		case TI_GETMYNAME: {
8068 			/*
8069 			 * For TI_GETPEERNAME and TI_GETMYNAME, we first
8070 			 * need to copyin the user's strbuf structure.
8071 			 * Processing will continue in the M_IOCDATA case
8072 			 * below.
8073 			 */
8074 			mi_copyin(q, mp, NULL,
8075 			    SIZEOF_STRUCT(strbuf, iocp->ioc_flag));
8076 			TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8077 			    "udp_wput_other_end: q %p (%S)", q, "getmyname");
8078 			return;
8079 			}
8080 		case ND_SET:
8081 			/* nd_getset performs the necessary checking */
8082 		case ND_GET:
8083 			if (nd_getset(q, us->us_nd, mp)) {
8084 				putnext(UDP_RD(q), mp);
8085 				TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8086 				    "udp_wput_other_end: q %p (%S)", q, "get");
8087 				return;
8088 			}
8089 			break;
8090 		case _SIOCSOCKFALLBACK:
8091 			/*
8092 			 * Either sockmod is about to be popped and the
8093 			 * socket would now be treated as a plain stream,
8094 			 * or a module is about to be pushed so we could
8095 			 * no longer use read-side synchronous stream.
8096 			 * Drain any queued data and disable direct sockfs
8097 			 * interface from now on.
8098 			 */
8099 			if (!udp->udp_issocket) {
8100 				DB_TYPE(mp) = M_IOCNAK;
8101 				iocp->ioc_error = EINVAL;
8102 			} else {
8103 				udp->udp_issocket = B_FALSE;
8104 				if (udp->udp_direct_sockfs) {
8105 					/*
8106 					 * Disable read-side synchronous
8107 					 * stream interface and drain any
8108 					 * queued data.
8109 					 */
8110 					udp_rcv_drain(UDP_RD(q), udp,
8111 					    B_FALSE);
8112 					ASSERT(!udp->udp_direct_sockfs);
8113 					UDP_STAT(us, udp_sock_fallback);
8114 				}
8115 				DB_TYPE(mp) = M_IOCACK;
8116 				iocp->ioc_error = 0;
8117 			}
8118 			iocp->ioc_count = 0;
8119 			iocp->ioc_rval = 0;
8120 			putnext(UDP_RD(q), mp);
8121 			return;
8122 		default:
8123 			break;
8124 		}
8125 		break;
8126 	case M_IOCDATA:
8127 		udp_wput_iocdata(q, mp);
8128 		TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8129 		    "udp_wput_other_end: q %p (%S)", q, "iocdata");
8130 		return;
8131 	default:
8132 		/* Unrecognized messages are passed through without change. */
8133 		break;
8134 	}
8135 	TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END,
8136 	    "udp_wput_other_end: q %p (%S)", q, "end");
8137 	ip_output(connp, mp, q, IP_WPUT);
8138 }
8139 
8140 /* ARGSUSED */
8141 static void
8142 udp_wput_other_wrapper(void *arg, mblk_t *mp, void *arg2)
8143 {
8144 	udp_wput_other(((conn_t *)arg)->conn_wq, mp);
8145 	udp_exit((conn_t *)arg);
8146 }
8147 
8148 /*
8149  * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA
8150  * messages.
8151  */
8152 static void
8153 udp_wput_iocdata(queue_t *q, mblk_t *mp)
8154 {
8155 	mblk_t	*mp1;
8156 	STRUCT_HANDLE(strbuf, sb);
8157 	uint16_t port;
8158 	in6_addr_t	v6addr;
8159 	ipaddr_t	v4addr;
8160 	uint32_t	flowinfo = 0;
8161 	int		addrlen;
8162 	udp_t		*udp = Q_TO_UDP(q);
8163 
8164 	/* Make sure it is one of ours. */
8165 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
8166 	case TI_GETMYNAME:
8167 	case TI_GETPEERNAME:
8168 		break;
8169 	default:
8170 		ip_output(udp->udp_connp, mp, q, IP_WPUT);
8171 		return;
8172 	}
8173 
8174 	q = WR(UDP_RD(q));
8175 	switch (mi_copy_state(q, mp, &mp1)) {
8176 	case -1:
8177 		return;
8178 	case MI_COPY_CASE(MI_COPY_IN, 1):
8179 		break;
8180 	case MI_COPY_CASE(MI_COPY_OUT, 1):
8181 		/*
8182 		 * The address has been copied out, so now
8183 		 * copyout the strbuf.
8184 		 */
8185 		mi_copyout(q, mp);
8186 		return;
8187 	case MI_COPY_CASE(MI_COPY_OUT, 2):
8188 		/*
8189 		 * The address and strbuf have been copied out.
8190 		 * We're done, so just acknowledge the original
8191 		 * M_IOCTL.
8192 		 */
8193 		mi_copy_done(q, mp, 0);
8194 		return;
8195 	default:
8196 		/*
8197 		 * Something strange has happened, so acknowledge
8198 		 * the original M_IOCTL with an EPROTO error.
8199 		 */
8200 		mi_copy_done(q, mp, EPROTO);
8201 		return;
8202 	}
8203 
8204 	/*
8205 	 * Now we have the strbuf structure for TI_GETMYNAME
8206 	 * and TI_GETPEERNAME.  Next we copyout the requested
8207 	 * address and then we'll copyout the strbuf.
8208 	 */
8209 	STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag,
8210 	    (void *)mp1->b_rptr);
8211 	if (udp->udp_family == AF_INET)
8212 		addrlen = sizeof (sin_t);
8213 	else
8214 		addrlen = sizeof (sin6_t);
8215 
8216 	if (STRUCT_FGET(sb, maxlen) < addrlen) {
8217 		mi_copy_done(q, mp, EINVAL);
8218 		return;
8219 	}
8220 	switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) {
8221 	case TI_GETMYNAME:
8222 		if (udp->udp_family == AF_INET) {
8223 			ASSERT(udp->udp_ipversion == IPV4_VERSION);
8224 			if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) &&
8225 			    !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
8226 				v4addr = V4_PART_OF_V6(udp->udp_v6src);
8227 			} else {
8228 				/*
8229 				 * INADDR_ANY
8230 				 * udp_v6src is not set, we might be bound to
8231 				 * broadcast/multicast. Use udp_bound_v6src as
8232 				 * local address instead (that could
8233 				 * also still be INADDR_ANY)
8234 				 */
8235 				v4addr = V4_PART_OF_V6(udp->udp_bound_v6src);
8236 			}
8237 		} else {
8238 			/* udp->udp_family == AF_INET6 */
8239 			if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) {
8240 				v6addr = udp->udp_v6src;
8241 			} else {
8242 				/*
8243 				 * UNSPECIFIED
8244 				 * udp_v6src is not set, we might be bound to
8245 				 * broadcast/multicast. Use udp_bound_v6src as
8246 				 * local address instead (that could
8247 				 * also still be UNSPECIFIED)
8248 				 */
8249 				v6addr = udp->udp_bound_v6src;
8250 			}
8251 		}
8252 		port = udp->udp_port;
8253 		break;
8254 	case TI_GETPEERNAME:
8255 		if (udp->udp_state != TS_DATA_XFER) {
8256 			mi_copy_done(q, mp, ENOTCONN);
8257 			return;
8258 		}
8259 		if (udp->udp_family == AF_INET) {
8260 			ASSERT(udp->udp_ipversion == IPV4_VERSION);
8261 			v4addr = V4_PART_OF_V6(udp->udp_v6dst);
8262 		} else {
8263 			/* udp->udp_family == AF_INET6) */
8264 			v6addr = udp->udp_v6dst;
8265 			flowinfo = udp->udp_flowinfo;
8266 		}
8267 		port = udp->udp_dstport;
8268 		break;
8269 	default:
8270 		mi_copy_done(q, mp, EPROTO);
8271 		return;
8272 	}
8273 	mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE);
8274 	if (!mp1)
8275 		return;
8276 
8277 	if (udp->udp_family == AF_INET) {
8278 		sin_t *sin;
8279 
8280 		STRUCT_FSET(sb, len, (int)sizeof (sin_t));
8281 		sin = (sin_t *)mp1->b_rptr;
8282 		mp1->b_wptr = (uchar_t *)&sin[1];
8283 		*sin = sin_null;
8284 		sin->sin_family = AF_INET;
8285 		sin->sin_addr.s_addr = v4addr;
8286 		sin->sin_port = port;
8287 	} else {
8288 		/* udp->udp_family == AF_INET6 */
8289 		sin6_t *sin6;
8290 
8291 		STRUCT_FSET(sb, len, (int)sizeof (sin6_t));
8292 		sin6 = (sin6_t *)mp1->b_rptr;
8293 		mp1->b_wptr = (uchar_t *)&sin6[1];
8294 		*sin6 = sin6_null;
8295 		sin6->sin6_family = AF_INET6;
8296 		sin6->sin6_flowinfo = flowinfo;
8297 		sin6->sin6_addr = v6addr;
8298 		sin6->sin6_port = port;
8299 	}
8300 	/* Copy out the address */
8301 	mi_copyout(q, mp);
8302 }
8303 
8304 
8305 static int
8306 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp,
8307     udpattrs_t *udpattrs)
8308 {
8309 	struct T_unitdata_req *udreqp;
8310 	int is_absreq_failure;
8311 	cred_t *cr;
8312 	conn_t	*connp = Q_TO_CONN(q);
8313 
8314 	ASSERT(((t_primp_t)mp->b_rptr)->type);
8315 
8316 	cr = DB_CREDDEF(mp, connp->conn_cred);
8317 
8318 	udreqp = (struct T_unitdata_req *)mp->b_rptr;
8319 
8320 	/*
8321 	 * Use upper queue for option processing since the callback
8322 	 * routines expect to be called in UDP instance instead of IP.
8323 	 */
8324 	*errorp = tpi_optcom_buf(_WR(UDP_RD(q)), mp, &udreqp->OPT_length,
8325 	    udreqp->OPT_offset, cr, &udp_opt_obj,
8326 	    udpattrs, &is_absreq_failure);
8327 
8328 	if (*errorp != 0) {
8329 		/*
8330 		 * Note: No special action needed in this
8331 		 * module for "is_absreq_failure"
8332 		 */
8333 		return (-1);		/* failure */
8334 	}
8335 	ASSERT(is_absreq_failure == 0);
8336 	return (0);	/* success */
8337 }
8338 
8339 void
8340 udp_ddi_init(void)
8341 {
8342 	UDP6_MAJ = ddi_name_to_major(UDP6);
8343 	udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr,
8344 	    udp_opt_obj.odb_opt_arr_cnt);
8345 
8346 	udp_cache = kmem_cache_create("udp_cache", sizeof (udp_t),
8347 	    CACHE_ALIGN_SIZE, NULL, NULL, NULL, NULL, NULL, 0);
8348 
8349 	/*
8350 	 * We want to be informed each time a stack is created or
8351 	 * destroyed in the kernel, so we can maintain the
8352 	 * set of udp_stack_t's.
8353 	 */
8354 	netstack_register(NS_UDP, udp_stack_init, NULL, udp_stack_fini);
8355 }
8356 
8357 void
8358 udp_ddi_destroy(void)
8359 {
8360 	netstack_unregister(NS_UDP);
8361 
8362 	kmem_cache_destroy(udp_cache);
8363 }
8364 
8365 /*
8366  * Initialize the UDP stack instance.
8367  */
8368 static void *
8369 udp_stack_init(netstackid_t stackid, netstack_t *ns)
8370 {
8371 	udp_stack_t	*us;
8372 	udpparam_t	*pa;
8373 	int		i;
8374 
8375 	us = (udp_stack_t *)kmem_zalloc(sizeof (*us), KM_SLEEP);
8376 	us->us_netstack = ns;
8377 
8378 	us->us_num_epriv_ports = UDP_NUM_EPRIV_PORTS;
8379 	us->us_epriv_ports[0] = 2049;
8380 	us->us_epriv_ports[1] = 4045;
8381 
8382 	/*
8383 	 * The smallest anonymous port in the priviledged port range which UDP
8384 	 * looks for free port.  Use in the option UDP_ANONPRIVBIND.
8385 	 */
8386 	us->us_min_anonpriv_port = 512;
8387 
8388 	us->us_bind_fanout_size = udp_bind_fanout_size;
8389 
8390 	/* Roundup variable that might have been modified in /etc/system */
8391 	if (us->us_bind_fanout_size & (us->us_bind_fanout_size - 1)) {
8392 		/* Not a power of two. Round up to nearest power of two */
8393 		for (i = 0; i < 31; i++) {
8394 			if (us->us_bind_fanout_size < (1 << i))
8395 				break;
8396 		}
8397 		us->us_bind_fanout_size = 1 << i;
8398 	}
8399 	us->us_bind_fanout = kmem_zalloc(us->us_bind_fanout_size *
8400 	    sizeof (udp_fanout_t), KM_SLEEP);
8401 	for (i = 0; i < us->us_bind_fanout_size; i++) {
8402 		mutex_init(&us->us_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT,
8403 		    NULL);
8404 	}
8405 
8406 	pa = (udpparam_t *)kmem_alloc(sizeof (udp_param_arr), KM_SLEEP);
8407 
8408 	us->us_param_arr = pa;
8409 	bcopy(udp_param_arr, us->us_param_arr, sizeof (udp_param_arr));
8410 
8411 	(void) udp_param_register(&us->us_nd,
8412 	    us->us_param_arr, A_CNT(udp_param_arr));
8413 
8414 	us->us_kstat = udp_kstat2_init(stackid, &us->us_statistics);
8415 	us->us_mibkp = udp_kstat_init(stackid);
8416 	return (us);
8417 }
8418 
8419 /*
8420  * Free the UDP stack instance.
8421  */
8422 static void
8423 udp_stack_fini(netstackid_t stackid, void *arg)
8424 {
8425 	udp_stack_t *us = (udp_stack_t *)arg;
8426 	int i;
8427 
8428 	for (i = 0; i < us->us_bind_fanout_size; i++) {
8429 		mutex_destroy(&us->us_bind_fanout[i].uf_lock);
8430 	}
8431 
8432 	kmem_free(us->us_bind_fanout, us->us_bind_fanout_size *
8433 	    sizeof (udp_fanout_t));
8434 
8435 	us->us_bind_fanout = NULL;
8436 
8437 	nd_free(&us->us_nd);
8438 	kmem_free(us->us_param_arr, sizeof (udp_param_arr));
8439 	us->us_param_arr = NULL;
8440 
8441 	udp_kstat_fini(stackid, us->us_mibkp);
8442 	us->us_mibkp = NULL;
8443 
8444 	udp_kstat2_fini(stackid, us->us_kstat);
8445 	us->us_kstat = NULL;
8446 	bzero(&us->us_statistics, sizeof (us->us_statistics));
8447 	kmem_free(us, sizeof (*us));
8448 }
8449 
8450 static void *
8451 udp_kstat2_init(netstackid_t stackid, udp_stat_t *us_statisticsp)
8452 {
8453 	kstat_t *ksp;
8454 
8455 	udp_stat_t template = {
8456 		{ "udp_ip_send",		KSTAT_DATA_UINT64 },
8457 		{ "udp_ip_ire_send",		KSTAT_DATA_UINT64 },
8458 		{ "udp_ire_null",		KSTAT_DATA_UINT64 },
8459 		{ "udp_drain",			KSTAT_DATA_UINT64 },
8460 		{ "udp_sock_fallback",		KSTAT_DATA_UINT64 },
8461 		{ "udp_rrw_busy",		KSTAT_DATA_UINT64 },
8462 		{ "udp_rrw_msgcnt",		KSTAT_DATA_UINT64 },
8463 		{ "udp_out_sw_cksum",		KSTAT_DATA_UINT64 },
8464 		{ "udp_out_sw_cksum_bytes",	KSTAT_DATA_UINT64 },
8465 		{ "udp_out_opt",		KSTAT_DATA_UINT64 },
8466 		{ "udp_out_err_notconn",	KSTAT_DATA_UINT64 },
8467 		{ "udp_out_err_output",		KSTAT_DATA_UINT64 },
8468 		{ "udp_out_err_tudr",		KSTAT_DATA_UINT64 },
8469 		{ "udp_in_pktinfo",		KSTAT_DATA_UINT64 },
8470 		{ "udp_in_recvdstaddr",		KSTAT_DATA_UINT64 },
8471 		{ "udp_in_recvopts",		KSTAT_DATA_UINT64 },
8472 		{ "udp_in_recvif",		KSTAT_DATA_UINT64 },
8473 		{ "udp_in_recvslla",		KSTAT_DATA_UINT64 },
8474 		{ "udp_in_recvucred",		KSTAT_DATA_UINT64 },
8475 		{ "udp_in_recvttl",		KSTAT_DATA_UINT64 },
8476 		{ "udp_in_recvhopopts",		KSTAT_DATA_UINT64 },
8477 		{ "udp_in_recvhoplimit",	KSTAT_DATA_UINT64 },
8478 		{ "udp_in_recvdstopts",		KSTAT_DATA_UINT64 },
8479 		{ "udp_in_recvrtdstopts",	KSTAT_DATA_UINT64 },
8480 		{ "udp_in_recvrthdr",		KSTAT_DATA_UINT64 },
8481 		{ "udp_in_recvpktinfo",		KSTAT_DATA_UINT64 },
8482 		{ "udp_in_recvtclass",		KSTAT_DATA_UINT64 },
8483 		{ "udp_in_timestamp",		KSTAT_DATA_UINT64 },
8484 #ifdef DEBUG
8485 		{ "udp_data_conn",		KSTAT_DATA_UINT64 },
8486 		{ "udp_data_notconn",		KSTAT_DATA_UINT64 },
8487 #endif
8488 	};
8489 
8490 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, "udpstat", "net",
8491 	    KSTAT_TYPE_NAMED, sizeof (template) / sizeof (kstat_named_t),
8492 	    KSTAT_FLAG_VIRTUAL, stackid);
8493 
8494 	if (ksp == NULL)
8495 		return (NULL);
8496 
8497 	bcopy(&template, us_statisticsp, sizeof (template));
8498 	ksp->ks_data = (void *)us_statisticsp;
8499 	ksp->ks_private = (void *)(uintptr_t)stackid;
8500 
8501 	kstat_install(ksp);
8502 	return (ksp);
8503 }
8504 
8505 static void
8506 udp_kstat2_fini(netstackid_t stackid, kstat_t *ksp)
8507 {
8508 	if (ksp != NULL) {
8509 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
8510 		kstat_delete_netstack(ksp, stackid);
8511 	}
8512 }
8513 
8514 static void *
8515 udp_kstat_init(netstackid_t stackid)
8516 {
8517 	kstat_t	*ksp;
8518 
8519 	udp_named_kstat_t template = {
8520 		{ "inDatagrams",	KSTAT_DATA_UINT64, 0 },
8521 		{ "inErrors",		KSTAT_DATA_UINT32, 0 },
8522 		{ "outDatagrams",	KSTAT_DATA_UINT64, 0 },
8523 		{ "entrySize",		KSTAT_DATA_INT32, 0 },
8524 		{ "entry6Size",		KSTAT_DATA_INT32, 0 },
8525 		{ "outErrors",		KSTAT_DATA_UINT32, 0 },
8526 	};
8527 
8528 	ksp = kstat_create_netstack(UDP_MOD_NAME, 0, UDP_MOD_NAME, "mib2",
8529 	    KSTAT_TYPE_NAMED,
8530 	    NUM_OF_FIELDS(udp_named_kstat_t), 0, stackid);
8531 
8532 	if (ksp == NULL || ksp->ks_data == NULL)
8533 		return (NULL);
8534 
8535 	template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t);
8536 	template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t);
8537 
8538 	bcopy(&template, ksp->ks_data, sizeof (template));
8539 	ksp->ks_update = udp_kstat_update;
8540 	ksp->ks_private = (void *)(uintptr_t)stackid;
8541 
8542 	kstat_install(ksp);
8543 	return (ksp);
8544 }
8545 
8546 static void
8547 udp_kstat_fini(netstackid_t stackid, kstat_t *ksp)
8548 {
8549 	if (ksp != NULL) {
8550 		ASSERT(stackid == (netstackid_t)(uintptr_t)ksp->ks_private);
8551 		kstat_delete_netstack(ksp, stackid);
8552 	}
8553 }
8554 
8555 static int
8556 udp_kstat_update(kstat_t *kp, int rw)
8557 {
8558 	udp_named_kstat_t *udpkp;
8559 	netstackid_t	stackid = (netstackid_t)(uintptr_t)kp->ks_private;
8560 	netstack_t	*ns;
8561 	udp_stack_t	*us;
8562 
8563 	if ((kp == NULL) || (kp->ks_data == NULL))
8564 		return (EIO);
8565 
8566 	if (rw == KSTAT_WRITE)
8567 		return (EACCES);
8568 
8569 	ns = netstack_find_by_stackid(stackid);
8570 	if (ns == NULL)
8571 		return (-1);
8572 	us = ns->netstack_udp;
8573 	if (us == NULL) {
8574 		netstack_rele(ns);
8575 		return (-1);
8576 	}
8577 	udpkp = (udp_named_kstat_t *)kp->ks_data;
8578 
8579 	udpkp->inDatagrams.value.ui64 =	us->us_udp_mib.udpHCInDatagrams;
8580 	udpkp->inErrors.value.ui32 =	us->us_udp_mib.udpInErrors;
8581 	udpkp->outDatagrams.value.ui64 = us->us_udp_mib.udpHCOutDatagrams;
8582 	udpkp->outErrors.value.ui32 =	us->us_udp_mib.udpOutErrors;
8583 	netstack_rele(ns);
8584 	return (0);
8585 }
8586 
8587 /* ARGSUSED */
8588 static void
8589 udp_rput(queue_t *q, mblk_t *mp)
8590 {
8591 	/*
8592 	 * We get here whenever we do qreply() from IP,
8593 	 * i.e as part of handlings ioctls, etc.
8594 	 */
8595 	putnext(q, mp);
8596 }
8597 
8598 /*
8599  * Read-side synchronous stream info entry point, called as a
8600  * result of handling certain STREAMS ioctl operations.
8601  */
8602 static int
8603 udp_rinfop(queue_t *q, infod_t *dp)
8604 {
8605 	mblk_t	*mp;
8606 	uint_t	cmd = dp->d_cmd;
8607 	int	res = 0;
8608 	int	error = 0;
8609 	udp_t	*udp = Q_TO_UDP(RD(UDP_WR(q)));
8610 	struct stdata *stp = STREAM(q);
8611 
8612 	mutex_enter(&udp->udp_drain_lock);
8613 	/* If shutdown on read has happened, return nothing */
8614 	mutex_enter(&stp->sd_lock);
8615 	if (stp->sd_flag & STREOF) {
8616 		mutex_exit(&stp->sd_lock);
8617 		goto done;
8618 	}
8619 	mutex_exit(&stp->sd_lock);
8620 
8621 	if ((mp = udp->udp_rcv_list_head) == NULL)
8622 		goto done;
8623 
8624 	ASSERT(DB_TYPE(mp) != M_DATA && mp->b_cont != NULL);
8625 
8626 	if (cmd & INFOD_COUNT) {
8627 		/*
8628 		 * Return the number of messages.
8629 		 */
8630 		dp->d_count += udp->udp_rcv_msgcnt;
8631 		res |= INFOD_COUNT;
8632 	}
8633 	if (cmd & INFOD_BYTES) {
8634 		/*
8635 		 * Return size of all data messages.
8636 		 */
8637 		dp->d_bytes += udp->udp_rcv_cnt;
8638 		res |= INFOD_BYTES;
8639 	}
8640 	if (cmd & INFOD_FIRSTBYTES) {
8641 		/*
8642 		 * Return size of first data message.
8643 		 */
8644 		dp->d_bytes = msgdsize(mp);
8645 		res |= INFOD_FIRSTBYTES;
8646 		dp->d_cmd &= ~INFOD_FIRSTBYTES;
8647 	}
8648 	if (cmd & INFOD_COPYOUT) {
8649 		mblk_t *mp1 = mp->b_cont;
8650 		int n;
8651 		/*
8652 		 * Return data contents of first message.
8653 		 */
8654 		ASSERT(DB_TYPE(mp1) == M_DATA);
8655 		while (mp1 != NULL && dp->d_uiop->uio_resid > 0) {
8656 			n = MIN(dp->d_uiop->uio_resid, MBLKL(mp1));
8657 			if (n != 0 && (error = uiomove((char *)mp1->b_rptr, n,
8658 			    UIO_READ, dp->d_uiop)) != 0) {
8659 				goto done;
8660 			}
8661 			mp1 = mp1->b_cont;
8662 		}
8663 		res |= INFOD_COPYOUT;
8664 		dp->d_cmd &= ~INFOD_COPYOUT;
8665 	}
8666 done:
8667 	mutex_exit(&udp->udp_drain_lock);
8668 
8669 	dp->d_res |= res;
8670 
8671 	return (error);
8672 }
8673 
8674 /*
8675  * Read-side synchronous stream entry point.  This is called as a result
8676  * of recv/read operation done at sockfs, and is guaranteed to execute
8677  * outside of the interrupt thread context.  It returns a single datagram
8678  * (b_cont chain of T_UNITDATA_IND plus data) to the upper layer.
8679  */
8680 static int
8681 udp_rrw(queue_t *q, struiod_t *dp)
8682 {
8683 	mblk_t	*mp;
8684 	udp_t	*udp = Q_TO_UDP(_RD(UDP_WR(q)));
8685 	udp_stack_t *us = udp->udp_us;
8686 
8687 	/* We should never get here when we're in SNMP mode */
8688 	ASSERT(!(udp->udp_connp->conn_flags & IPCL_UDPMOD));
8689 
8690 	/*
8691 	 * Dequeue datagram from the head of the list and return
8692 	 * it to caller; also ensure that RSLEEP sd_wakeq flag is
8693 	 * set/cleared depending on whether or not there's data
8694 	 * remaining in the list.
8695 	 */
8696 	mutex_enter(&udp->udp_drain_lock);
8697 	if (!udp->udp_direct_sockfs) {
8698 		mutex_exit(&udp->udp_drain_lock);
8699 		UDP_STAT(us, udp_rrw_busy);
8700 		return (EBUSY);
8701 	}
8702 	if ((mp = udp->udp_rcv_list_head) != NULL) {
8703 		uint_t size = msgdsize(mp);
8704 
8705 		/* Last datagram in the list? */
8706 		if ((udp->udp_rcv_list_head = mp->b_next) == NULL)
8707 			udp->udp_rcv_list_tail = NULL;
8708 		mp->b_next = NULL;
8709 
8710 		udp->udp_rcv_cnt -= size;
8711 		udp->udp_rcv_msgcnt--;
8712 		UDP_STAT(us, udp_rrw_msgcnt);
8713 
8714 		/* No longer flow-controlling? */
8715 		if (udp->udp_rcv_cnt < udp->udp_rcv_hiwat &&
8716 		    udp->udp_rcv_msgcnt < udp->udp_rcv_hiwat)
8717 			udp->udp_drain_qfull = B_FALSE;
8718 	}
8719 	if (udp->udp_rcv_list_head == NULL) {
8720 		/*
8721 		 * Either we just dequeued the last datagram or
8722 		 * we get here from sockfs and have nothing to
8723 		 * return; in this case clear RSLEEP.
8724 		 */
8725 		ASSERT(udp->udp_rcv_cnt == 0);
8726 		ASSERT(udp->udp_rcv_msgcnt == 0);
8727 		ASSERT(udp->udp_rcv_list_tail == NULL);
8728 		STR_WAKEUP_CLEAR(STREAM(q));
8729 	} else {
8730 		/*
8731 		 * More data follows; we need udp_rrw() to be
8732 		 * called in future to pick up the rest.
8733 		 */
8734 		STR_WAKEUP_SET(STREAM(q));
8735 	}
8736 	mutex_exit(&udp->udp_drain_lock);
8737 	dp->d_mp = mp;
8738 	return (0);
8739 }
8740 
8741 /*
8742  * Enqueue a completely-built T_UNITDATA_IND message into the receive
8743  * list; this is typically executed within the interrupt thread context
8744  * and so we do things as quickly as possible.
8745  */
8746 static void
8747 udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, uint_t pkt_len)
8748 {
8749 	ASSERT(q == RD(q));
8750 	ASSERT(pkt_len == msgdsize(mp));
8751 	ASSERT(mp->b_next == NULL && mp->b_cont != NULL);
8752 	ASSERT(DB_TYPE(mp) == M_PROTO && DB_TYPE(mp->b_cont) == M_DATA);
8753 	ASSERT(MBLKL(mp) >= sizeof (struct T_unitdata_ind));
8754 
8755 	mutex_enter(&udp->udp_drain_lock);
8756 	/*
8757 	 * Wake up and signal the receiving app; it is okay to do this
8758 	 * before enqueueing the mp because we are holding the drain lock.
8759 	 * One of the advantages of synchronous stream is the ability for
8760 	 * us to find out when the application performs a read on the
8761 	 * socket by way of udp_rrw() entry point being called.  We need
8762 	 * to generate SIGPOLL/SIGIO for each received data in the case
8763 	 * of asynchronous socket just as in the strrput() case.  However,
8764 	 * we only wake the application up when necessary, i.e. during the
8765 	 * first enqueue.  When udp_rrw() is called, we send up a single
8766 	 * datagram upstream and call STR_WAKEUP_SET() again when there
8767 	 * are still data remaining in our receive queue.
8768 	 */
8769 	if (udp->udp_rcv_list_head == NULL) {
8770 		STR_WAKEUP_SET(STREAM(q));
8771 		udp->udp_rcv_list_head = mp;
8772 	} else {
8773 		udp->udp_rcv_list_tail->b_next = mp;
8774 	}
8775 	udp->udp_rcv_list_tail = mp;
8776 	udp->udp_rcv_cnt += pkt_len;
8777 	udp->udp_rcv_msgcnt++;
8778 
8779 	/* Need to flow-control? */
8780 	if (udp->udp_rcv_cnt >= udp->udp_rcv_hiwat ||
8781 	    udp->udp_rcv_msgcnt >= udp->udp_rcv_hiwat)
8782 		udp->udp_drain_qfull = B_TRUE;
8783 
8784 	/* Update poll events and send SIGPOLL/SIGIO if necessary */
8785 	STR_SENDSIG(STREAM(q));
8786 	mutex_exit(&udp->udp_drain_lock);
8787 }
8788 
8789 /*
8790  * Drain the contents of receive list to the module upstream; we do
8791  * this during close or when we fallback to the slow mode due to
8792  * sockmod being popped or a module being pushed on top of us.
8793  */
8794 static void
8795 udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing)
8796 {
8797 	mblk_t *mp;
8798 	udp_stack_t *us = udp->udp_us;
8799 
8800 	ASSERT(q == RD(q));
8801 
8802 	mutex_enter(&udp->udp_drain_lock);
8803 	/*
8804 	 * There is no race with a concurrent udp_input() sending
8805 	 * up packets using putnext() after we have cleared the
8806 	 * udp_direct_sockfs flag but before we have completed
8807 	 * sending up the packets in udp_rcv_list, since we are
8808 	 * either a writer or we have quiesced the conn.
8809 	 */
8810 	udp->udp_direct_sockfs = B_FALSE;
8811 	mutex_exit(&udp->udp_drain_lock);
8812 
8813 	if (udp->udp_rcv_list_head != NULL)
8814 		UDP_STAT(us, udp_drain);
8815 
8816 	/*
8817 	 * Send up everything via putnext(); note here that we
8818 	 * don't need the udp_drain_lock to protect us since
8819 	 * nothing can enter udp_rrw() and that we currently
8820 	 * have exclusive access to this udp.
8821 	 */
8822 	while ((mp = udp->udp_rcv_list_head) != NULL) {
8823 		udp->udp_rcv_list_head = mp->b_next;
8824 		mp->b_next = NULL;
8825 		udp->udp_rcv_cnt -= msgdsize(mp);
8826 		udp->udp_rcv_msgcnt--;
8827 		if (closing) {
8828 			freemsg(mp);
8829 		} else {
8830 			putnext(q, mp);
8831 		}
8832 	}
8833 	ASSERT(udp->udp_rcv_cnt == 0);
8834 	ASSERT(udp->udp_rcv_msgcnt == 0);
8835 	ASSERT(udp->udp_rcv_list_head == NULL);
8836 	udp->udp_rcv_list_tail = NULL;
8837 	udp->udp_drain_qfull = B_FALSE;
8838 }
8839 
8840 static size_t
8841 udp_set_rcv_hiwat(udp_t *udp, size_t size)
8842 {
8843 	udp_stack_t *us = udp->udp_us;
8844 
8845 	/* We add a bit of extra buffering */
8846 	size += size >> 1;
8847 	if (size > us->us_max_buf)
8848 		size = us->us_max_buf;
8849 
8850 	udp->udp_rcv_hiwat = size;
8851 	return (size);
8852 }
8853