xref: /illumos-gate/usr/src/uts/common/inet/sockmods/socksctp.c (revision 8a2b682e57a046b828f37bcde1776f131ef4629f)
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 /*
23  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24  * Copyright (c) 2015 Joyent, Inc.  All rights reserved.
25  * Copyright 2019 OmniOS Community Edition (OmniOSce) Association.
26  */
27 
28 #include <sys/types.h>
29 #include <sys/t_lock.h>
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/buf.h>
33 #include <sys/vfs.h>
34 #include <sys/vnode.h>
35 #include <sys/debug.h>
36 #include <sys/errno.h>
37 #include <sys/stropts.h>
38 #include <sys/cmn_err.h>
39 #include <sys/sysmacros.h>
40 #include <sys/filio.h>
41 #include <sys/policy.h>
42 
43 #include <sys/project.h>
44 #include <sys/tihdr.h>
45 #include <sys/strsubr.h>
46 #include <sys/esunddi.h>
47 #include <sys/ddi.h>
48 
49 #include <sys/sockio.h>
50 #include <sys/socket.h>
51 #include <sys/socketvar.h>
52 #include <sys/strsun.h>
53 
54 #include <netinet/sctp.h>
55 #include <inet/sctp_itf.h>
56 #include <fs/sockfs/sockcommon.h>
57 #include "socksctp.h"
58 
59 /*
60  * SCTP sockfs sonode operations, 1-1 socket
61  */
62 static int sosctp_init(struct sonode *, struct sonode *, struct cred *, int);
63 static int sosctp_accept(struct sonode *, int, struct cred *, struct sonode **);
64 static int sosctp_bind(struct sonode *, struct sockaddr *, socklen_t, int,
65     struct cred *);
66 static int sosctp_listen(struct sonode *, int, struct cred *);
67 static int sosctp_connect(struct sonode *, struct sockaddr *, socklen_t,
68     int, int, struct cred *);
69 static int sosctp_recvmsg(struct sonode *, struct nmsghdr *, struct uio *,
70     struct cred *);
71 static int sosctp_sendmsg(struct sonode *, struct nmsghdr *, struct uio *,
72     struct cred *);
73 static int sosctp_getpeername(struct sonode *, struct sockaddr *, socklen_t *,
74     boolean_t, struct cred *);
75 static int sosctp_getsockname(struct sonode *, struct sockaddr *, socklen_t *,
76     struct cred *);
77 static int sosctp_shutdown(struct sonode *, int, struct cred *);
78 static int sosctp_getsockopt(struct sonode *, int, int, void *, socklen_t *,
79     int, struct cred *);
80 static int sosctp_setsockopt(struct sonode *, int, int, const void *,
81     socklen_t, struct cred *);
82 static int sosctp_ioctl(struct sonode *, int, intptr_t, int, struct cred *,
83     int32_t *);
84 static int sosctp_close(struct sonode *, int, struct cred *);
85 void sosctp_fini(struct sonode *, struct cred *);
86 
87 /*
88  * SCTP sockfs sonode operations, 1-N socket
89  */
90 static int sosctp_seq_connect(struct sonode *, struct sockaddr *,
91     socklen_t, int, int, struct cred *);
92 static int sosctp_seq_sendmsg(struct sonode *, struct nmsghdr *, struct uio *,
93     struct cred *);
94 
95 /*
96  * Socket association upcalls, 1-N socket connection
97  */
98 sock_upper_handle_t sctp_assoc_newconn(sock_upper_handle_t,
99     sock_lower_handle_t, sock_downcalls_t *, struct cred *, pid_t,
100     sock_upcalls_t **);
101 static void sctp_assoc_connected(sock_upper_handle_t, sock_connid_t,
102     struct cred *, pid_t);
103 static int sctp_assoc_disconnected(sock_upper_handle_t, sock_connid_t, int);
104 static void sctp_assoc_disconnecting(sock_upper_handle_t, sock_opctl_action_t,
105     uintptr_t arg);
106 static ssize_t sctp_assoc_recv(sock_upper_handle_t, mblk_t *, size_t, int,
107     int *, boolean_t *);
108 static void sctp_assoc_xmitted(sock_upper_handle_t, boolean_t);
109 static void sctp_assoc_properties(sock_upper_handle_t,
110     struct sock_proto_props *);
111 static vnode_t *sctp_assoc_get_vnode(sock_upper_handle_t);
112 
113 sonodeops_t sosctp_sonodeops = {
114 	sosctp_init,			/* sop_init	*/
115 	sosctp_accept,			/* sop_accept	*/
116 	sosctp_bind,			/* sop_bind	*/
117 	sosctp_listen,			/* sop_listen	*/
118 	sosctp_connect,			/* sop_connect	*/
119 	sosctp_recvmsg,			/* sop_recvmsg	*/
120 	sosctp_sendmsg,			/* sop_sendmsg	*/
121 	so_sendmblk_notsupp,		/* sop_sendmblk	*/
122 	sosctp_getpeername,		/* sop_getpeername */
123 	sosctp_getsockname,		/* sop_getsockname */
124 	sosctp_shutdown,		/* sop_shutdown */
125 	sosctp_getsockopt,		/* sop_getsockopt */
126 	sosctp_setsockopt,		/* sop_setsockopt */
127 	sosctp_ioctl,			/* sop_ioctl	*/
128 	so_poll,			/* sop_poll	*/
129 	sosctp_close,			/* sop_close	*/
130 };
131 
132 sonodeops_t sosctp_seq_sonodeops = {
133 	sosctp_init,			/* sop_init	*/
134 	so_accept_notsupp,		/* sop_accept	*/
135 	sosctp_bind,			/* sop_bind	*/
136 	sosctp_listen,			/* sop_listen	*/
137 	sosctp_seq_connect,		/* sop_connect	*/
138 	sosctp_recvmsg,			/* sop_recvmsg	*/
139 	sosctp_seq_sendmsg,		/* sop_sendmsg	*/
140 	so_sendmblk_notsupp,		/* sop_sendmblk	*/
141 	so_getpeername_notsupp,		/* sop_getpeername */
142 	sosctp_getsockname,		/* sop_getsockname */
143 	so_shutdown_notsupp,		/* sop_shutdown */
144 	sosctp_getsockopt,		/* sop_getsockopt */
145 	sosctp_setsockopt,		/* sop_setsockopt */
146 	sosctp_ioctl,			/* sop_ioctl	*/
147 	so_poll,			/* sop_poll	*/
148 	sosctp_close,			/* sop_close	*/
149 };
150 
151 /* All the upcalls expect the upper handle to be sonode. */
152 sock_upcalls_t sosctp_sock_upcalls = {
153 	so_newconn,
154 	so_connected,
155 	so_disconnected,
156 	so_opctl,
157 	so_queue_msg,
158 	so_set_prop,
159 	so_txq_full,
160 	NULL,			/* su_signal_oob */
161 	NULL,			/* su_signal_oob */
162 	NULL,			/* su_set_error */
163 	NULL,			/* su_closed */
164 	so_get_vnode		/* su_get_vnode */
165 };
166 
167 /* All the upcalls expect the upper handle to be sctp_sonode/sctp_soassoc. */
168 sock_upcalls_t sosctp_assoc_upcalls = {
169 	sctp_assoc_newconn,
170 	sctp_assoc_connected,
171 	sctp_assoc_disconnected,
172 	sctp_assoc_disconnecting,
173 	sctp_assoc_recv,
174 	sctp_assoc_properties,
175 	sctp_assoc_xmitted,
176 	NULL,			/* su_recv_space */
177 	NULL,			/* su_signal_oob */
178 	NULL,			/* su_set_error */
179 	NULL,			/* su_closed */
180 	sctp_assoc_get_vnode
181 };
182 
183 /* ARGSUSED */
184 static int
185 sosctp_init(struct sonode *so, struct sonode *pso, struct cred *cr, int flags)
186 {
187 	struct sctp_sonode *ss;
188 	struct sctp_sonode *pss;
189 	sctp_sockbuf_limits_t sbl;
190 	int err;
191 
192 	ss = SOTOSSO(so);
193 
194 	if (pso != NULL) {
195 		/*
196 		 * Passive open, just inherit settings from parent. We should
197 		 * not end up here for SOCK_SEQPACKET type sockets, since no
198 		 * new sonode is created in that case.
199 		 */
200 		ASSERT(so->so_type == SOCK_STREAM);
201 		pss = SOTOSSO(pso);
202 
203 		mutex_enter(&pso->so_lock);
204 		so->so_state |= (SS_ISBOUND | SS_ISCONNECTED |
205 		    (pso->so_state & SS_ASYNC));
206 		sosctp_so_inherit(pss, ss);
207 		so->so_proto_props = pso->so_proto_props;
208 		so->so_mode = pso->so_mode;
209 		mutex_exit(&pso->so_lock);
210 
211 		return (0);
212 	}
213 
214 	if ((err = secpolicy_basic_net_access(cr)) != 0)
215 		return (err);
216 
217 	if (so->so_type == SOCK_STREAM) {
218 		so->so_proto_handle = (sock_lower_handle_t)sctp_create(so,
219 		    NULL, so->so_family, so->so_type, SCTP_CAN_BLOCK,
220 		    &sosctp_sock_upcalls, &sbl, cr);
221 		so->so_mode = SM_CONNREQUIRED;
222 	} else {
223 		ASSERT(so->so_type == SOCK_SEQPACKET);
224 		so->so_proto_handle = (sock_lower_handle_t)sctp_create(ss,
225 		    NULL, so->so_family, so->so_type, SCTP_CAN_BLOCK,
226 		    &sosctp_assoc_upcalls, &sbl, cr);
227 	}
228 
229 	if (so->so_proto_handle == NULL)
230 		return (ENOMEM);
231 
232 	so->so_rcvbuf = sbl.sbl_rxbuf;
233 	so->so_rcvlowat = sbl.sbl_rxlowat;
234 	so->so_sndbuf = sbl.sbl_txbuf;
235 	so->so_sndlowat = sbl.sbl_txlowat;
236 
237 	return (0);
238 }
239 
240 /*
241  * Accept incoming connection.
242  */
243 /*ARGSUSED*/
244 static int
245 sosctp_accept(struct sonode *so, int fflag, struct cred *cr,
246     struct sonode **nsop)
247 {
248 	int error = 0;
249 
250 	if ((so->so_state & SS_ACCEPTCONN) == 0)
251 		return (EINVAL);
252 
253 	error = so_acceptq_dequeue(so, (fflag & (FNONBLOCK|FNDELAY)), nsop);
254 
255 	return (error);
256 }
257 
258 /*
259  * Bind local endpoint.
260  */
261 /*ARGSUSED*/
262 static int
263 sosctp_bind(struct sonode *so, struct sockaddr *name, socklen_t namelen,
264     int flags, struct cred *cr)
265 {
266 	int error;
267 
268 	if (!(flags & _SOBIND_LOCK_HELD)) {
269 		mutex_enter(&so->so_lock);
270 		so_lock_single(so);	/* Set SOLOCKED */
271 	} else {
272 		ASSERT(MUTEX_HELD(&so->so_lock));
273 	}
274 
275 	/*
276 	 * X/Open requires this check
277 	 */
278 	if (so->so_state & SS_CANTSENDMORE) {
279 		error = EINVAL;
280 		goto done;
281 	}
282 
283 
284 	/*
285 	 * Protocol module does address family checks.
286 	 */
287 	mutex_exit(&so->so_lock);
288 
289 	error = sctp_bind((struct sctp_s *)so->so_proto_handle, name, namelen);
290 
291 	mutex_enter(&so->so_lock);
292 	if (error == 0) {
293 		so->so_state |= SS_ISBOUND;
294 	} else {
295 		eprintsoline(so, error);
296 	}
297 done:
298 	if (!(flags & _SOBIND_LOCK_HELD)) {
299 		so_unlock_single(so, SOLOCKED);
300 		mutex_exit(&so->so_lock);
301 	} else {
302 		/* If the caller held the lock don't release it here */
303 		ASSERT(MUTEX_HELD(&so->so_lock));
304 		ASSERT(so->so_flag & SOLOCKED);
305 	}
306 
307 	return (error);
308 }
309 
310 /*
311  * Turn socket into a listen socket.
312  */
313 /* ARGSUSED */
314 static int
315 sosctp_listen(struct sonode *so, int backlog, struct cred *cr)
316 {
317 	int error = 0;
318 
319 	mutex_enter(&so->so_lock);
320 	so_lock_single(so);
321 
322 	/*
323 	 * If this socket is trying to do connect, or if it has
324 	 * been connected, disallow.
325 	 */
326 	if (so->so_state & (SS_ISCONNECTING | SS_ISCONNECTED |
327 	    SS_ISDISCONNECTING | SS_CANTRCVMORE | SS_CANTSENDMORE)) {
328 		error = EINVAL;
329 		eprintsoline(so, error);
330 		goto done;
331 	}
332 
333 	if (backlog < 0) {
334 		backlog = 0;
335 	}
336 
337 	/*
338 	 * If listen() is only called to change backlog, we don't
339 	 * need to notify protocol module.
340 	 */
341 	if (so->so_state & SS_ACCEPTCONN) {
342 		so->so_backlog = backlog;
343 		goto done;
344 	}
345 
346 	mutex_exit(&so->so_lock);
347 	error = sctp_listen((struct sctp_s *)so->so_proto_handle);
348 	mutex_enter(&so->so_lock);
349 	if (error == 0) {
350 		so->so_state |= (SS_ACCEPTCONN|SS_ISBOUND);
351 		so->so_backlog = backlog;
352 	} else {
353 		eprintsoline(so, error);
354 	}
355 done:
356 	so_unlock_single(so, SOLOCKED);
357 	mutex_exit(&so->so_lock);
358 
359 	return (error);
360 }
361 
362 /*
363  * Active open.
364  */
365 /*ARGSUSED*/
366 static int
367 sosctp_connect(struct sonode *so, struct sockaddr *name,
368     socklen_t namelen, int fflag, int flags, struct cred *cr)
369 {
370 	int error = 0;
371 	pid_t pid = curproc->p_pid;
372 
373 	ASSERT(so->so_type == SOCK_STREAM);
374 
375 	mutex_enter(&so->so_lock);
376 	so_lock_single(so);
377 
378 	/*
379 	 * Can't connect() after listen(), or if the socket is already
380 	 * connected.
381 	 */
382 	if (so->so_state & (SS_ACCEPTCONN|SS_ISCONNECTED|SS_ISCONNECTING)) {
383 		if (so->so_state & SS_ISCONNECTED) {
384 			error = EISCONN;
385 		} else if (so->so_state & SS_ISCONNECTING) {
386 			error = EALREADY;
387 		} else {
388 			error = EOPNOTSUPP;
389 		}
390 		eprintsoline(so, error);
391 		goto done;
392 	}
393 
394 	/*
395 	 * Check for failure of an earlier call
396 	 */
397 	if (so->so_error != 0) {
398 		error = sogeterr(so, B_TRUE);
399 		eprintsoline(so, error);
400 		goto done;
401 	}
402 
403 	/*
404 	 * Connection is closing, or closed, don't allow reconnect.
405 	 * TCP allows this to proceed, but the socket remains unwriteable.
406 	 * BSD returns EINVAL.
407 	 */
408 	if (so->so_state & (SS_ISDISCONNECTING|SS_CANTRCVMORE|
409 	    SS_CANTSENDMORE)) {
410 		error = EINVAL;
411 		eprintsoline(so, error);
412 		goto done;
413 	}
414 
415 	if (name == NULL || namelen == 0) {
416 		error = EINVAL;
417 		eprintsoline(so, error);
418 		goto done;
419 	}
420 
421 	soisconnecting(so);
422 	mutex_exit(&so->so_lock);
423 
424 	error = sctp_connect((struct sctp_s *)so->so_proto_handle,
425 	    name, namelen, cr, pid);
426 
427 	mutex_enter(&so->so_lock);
428 	if (error == 0) {
429 		/*
430 		 * Allow other threads to access the socket
431 		 */
432 		error = sowaitconnected(so, fflag, 0);
433 	}
434 done:
435 	so_unlock_single(so, SOLOCKED);
436 	mutex_exit(&so->so_lock);
437 	return (error);
438 }
439 
440 /*
441  * Active open for 1-N sockets, create a new association and
442  * call connect on that.
443  * If there parent hasn't been bound yet (this is the first association),
444  * make it so.
445  */
446 static int
447 sosctp_seq_connect(struct sonode *so, struct sockaddr *name,
448     socklen_t namelen, int fflag, int flags, struct cred *cr)
449 {
450 	struct sctp_soassoc *ssa;
451 	struct sctp_sonode *ss;
452 	int error;
453 
454 	ASSERT(so->so_type == SOCK_SEQPACKET);
455 
456 	mutex_enter(&so->so_lock);
457 	so_lock_single(so);
458 
459 	if (name == NULL || namelen == 0) {
460 		error = EINVAL;
461 		eprintsoline(so, error);
462 		goto done;
463 	}
464 
465 	ss = SOTOSSO(so);
466 
467 	error = sosctp_assoc_createconn(ss, name, namelen, NULL, 0, fflag,
468 	    cr, &ssa);
469 	if (error != 0) {
470 		if ((error == EHOSTUNREACH) && (flags & _SOCONNECT_XPG4_2)) {
471 			error = ENETUNREACH;
472 		}
473 	}
474 	if (ssa != NULL) {
475 		SSA_REFRELE(ss, ssa);
476 	}
477 
478 done:
479 	so_unlock_single(so, SOLOCKED);
480 	mutex_exit(&so->so_lock);
481 	return (error);
482 }
483 
484 /*
485  * Receive data.
486  */
487 /* ARGSUSED */
488 static int
489 sosctp_recvmsg(struct sonode *so, struct nmsghdr *msg, struct uio *uiop,
490     struct cred *cr)
491 {
492 	struct sctp_sonode *ss = SOTOSSO(so);
493 	struct sctp_soassoc *ssa = NULL;
494 	int flags, error = 0;
495 	struct T_unitdata_ind *tind;
496 	ssize_t orig_resid = uiop->uio_resid;
497 	int len, count, readcnt = 0;
498 	socklen_t controllen, namelen;
499 	void *opt;
500 	mblk_t *mp;
501 	rval_t	rval;
502 
503 	controllen = msg->msg_controllen;
504 	namelen = msg->msg_namelen;
505 	flags = msg->msg_flags;
506 	msg->msg_flags = 0;
507 	msg->msg_controllen = 0;
508 	msg->msg_namelen = 0;
509 
510 	if (so->so_type == SOCK_STREAM) {
511 		if (!(so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING|
512 		    SS_CANTRCVMORE))) {
513 			return (ENOTCONN);
514 		}
515 	} else {
516 		/* NOTE: Will come here from vop_read() as well */
517 		/* For 1-N socket, recv() cannot be used. */
518 		if (namelen == 0)
519 			return (EOPNOTSUPP);
520 		/*
521 		 * If there are no associations, and no new connections are
522 		 * coming in, there's not going to be new messages coming
523 		 * in either.
524 		 */
525 		if (so->so_rcv_q_head == NULL && so->so_rcv_head == NULL &&
526 		    ss->ss_assoccnt == 0 && !(so->so_state & SS_ACCEPTCONN)) {
527 			return (ENOTCONN);
528 		}
529 	}
530 
531 	/*
532 	 * out-of-band data not supported.
533 	 */
534 	if (flags & MSG_OOB) {
535 		return (EOPNOTSUPP);
536 	}
537 
538 	/*
539 	 * flag possibilities:
540 	 *
541 	 * MSG_PEEK	Don't consume data
542 	 * MSG_WAITALL	Wait for full quantity of data (ignored if MSG_PEEK)
543 	 * MSG_DONTWAIT Non-blocking (same as FNDELAY | FNONBLOCK)
544 	 *
545 	 * MSG_WAITALL can return less than the full buffer if either
546 	 *
547 	 * 1. we would block and we are non-blocking
548 	 * 2. a full message cannot be delivered
549 	 *
550 	 * Given that we always get a full message from proto below,
551 	 * MSG_WAITALL is not meaningful.
552 	 */
553 
554 	mutex_enter(&so->so_lock);
555 
556 	/*
557 	 * Allow just one reader at a time.
558 	 */
559 	error = so_lock_read_intr(so,
560 	    uiop->uio_fmode | ((flags & MSG_DONTWAIT) ? FNONBLOCK : 0));
561 	if (error) {
562 		mutex_exit(&so->so_lock);
563 		return (error);
564 	}
565 	mutex_exit(&so->so_lock);
566 again:
567 	error = so_dequeue_msg(so, &mp, uiop, &rval, flags | MSG_DUPCTRL);
568 	if (mp != NULL) {
569 		if (so->so_type == SOCK_SEQPACKET) {
570 			ssa = *(struct sctp_soassoc **)DB_BASE(mp);
571 		}
572 
573 		tind = (struct T_unitdata_ind *)mp->b_rptr;
574 
575 		len = tind->SRC_length;
576 
577 		if (namelen > 0 && len > 0) {
578 
579 			opt = sogetoff(mp, tind->SRC_offset, len, 1);
580 
581 			ASSERT(opt != NULL);
582 
583 			msg->msg_name = kmem_alloc(len, KM_SLEEP);
584 			msg->msg_namelen = len;
585 
586 			bcopy(opt, msg->msg_name, len);
587 		}
588 
589 		len = tind->OPT_length;
590 		if (controllen == 0) {
591 			if (len > 0) {
592 				msg->msg_flags |= MSG_CTRUNC;
593 			}
594 		} else if (len > 0) {
595 			opt = sogetoff(mp, tind->OPT_offset, len,
596 			    __TPI_ALIGN_SIZE);
597 
598 			ASSERT(opt != NULL);
599 			sosctp_pack_cmsg(opt, msg, len);
600 		}
601 
602 		if (mp->b_flag & SCTP_NOTIFICATION) {
603 			msg->msg_flags |= MSG_NOTIFICATION;
604 		}
605 
606 		if (!(mp->b_flag & SCTP_PARTIAL_DATA) &&
607 		    !(rval.r_val1 & MOREDATA)) {
608 			msg->msg_flags |= MSG_EOR;
609 		}
610 		freemsg(mp);
611 	}
612 done:
613 	if (!(flags & MSG_PEEK))
614 		readcnt = orig_resid - uiop->uio_resid;
615 	/*
616 	 * Determine if we need to update SCTP about the buffer
617 	 * space.  For performance reason, we cannot update SCTP
618 	 * every time a message is read.  The socket buffer low
619 	 * watermark is used as the threshold.
620 	 */
621 	if (ssa == NULL) {
622 		mutex_enter(&so->so_lock);
623 		count = so->so_rcvbuf - so->so_rcv_queued;
624 
625 		ASSERT(so->so_rcv_q_head != NULL ||
626 		    so->so_rcv_head != NULL ||
627 		    so->so_rcv_queued == 0);
628 
629 		so_unlock_read(so);
630 
631 		/*
632 		 * so_dequeue_msg() sets r_val2 to true if flow control was
633 		 * cleared and we need to update SCTP.  so_flowctrld was
634 		 * cleared in so_dequeue_msg() via so_check_flow_control().
635 		 */
636 		if (rval.r_val2) {
637 			mutex_exit(&so->so_lock);
638 			sctp_recvd((struct sctp_s *)so->so_proto_handle, count);
639 		} else {
640 			mutex_exit(&so->so_lock);
641 		}
642 	} else {
643 		/*
644 		 * Each association keeps track of how much data it has
645 		 * queued; we need to update the value here. Note that this
646 		 * is slightly different from SOCK_STREAM type sockets, which
647 		 * does not need to update the byte count, as it is already
648 		 * done in so_dequeue_msg().
649 		 */
650 		mutex_enter(&so->so_lock);
651 		ssa->ssa_rcv_queued -= readcnt;
652 		count = so->so_rcvbuf - ssa->ssa_rcv_queued;
653 
654 		so_unlock_read(so);
655 
656 		if (readcnt > 0 && ssa->ssa_flowctrld &&
657 		    ssa->ssa_rcv_queued < so->so_rcvlowat) {
658 			/*
659 			 * Need to clear ssa_flowctrld, different from 1-1
660 			 * style.
661 			 */
662 			ssa->ssa_flowctrld = B_FALSE;
663 			mutex_exit(&so->so_lock);
664 			sctp_recvd(ssa->ssa_conn, count);
665 			mutex_enter(&so->so_lock);
666 		}
667 
668 		/*
669 		 * MOREDATA flag is set if all data could not be copied
670 		 */
671 		if (!(flags & MSG_PEEK) && !(rval.r_val1 & MOREDATA)) {
672 			SSA_REFRELE(ss, ssa);
673 		}
674 		mutex_exit(&so->so_lock);
675 	}
676 
677 	return (error);
678 }
679 
680 int
681 sosctp_uiomove(mblk_t *hdr_mp, ssize_t count, ssize_t blk_size, int wroff,
682     struct uio *uiop, int flags)
683 {
684 	ssize_t size;
685 	int error;
686 	mblk_t *mp;
687 	dblk_t *dp;
688 
689 	if (blk_size == INFPSZ)
690 		blk_size = count;
691 
692 	/*
693 	 * Loop until we have all data copied into mblk's.
694 	 */
695 	while (count > 0) {
696 		size = MIN(count, blk_size);
697 
698 		/*
699 		 * As a message can be splitted up and sent in different
700 		 * packets, each mblk will have the extra space before
701 		 * data to accommodate what SCTP wants to put in there.
702 		 */
703 		while ((mp = allocb(size + wroff, BPRI_MED)) == NULL) {
704 			if ((uiop->uio_fmode & (FNDELAY|FNONBLOCK)) ||
705 			    (flags & MSG_DONTWAIT)) {
706 				return (EAGAIN);
707 			}
708 			if ((error = strwaitbuf(size + wroff, BPRI_MED))) {
709 				return (error);
710 			}
711 		}
712 
713 		dp = mp->b_datap;
714 		dp->db_cpid = curproc->p_pid;
715 		ASSERT(wroff <= dp->db_lim - mp->b_wptr);
716 		mp->b_rptr += wroff;
717 		error = uiomove(mp->b_rptr, size, UIO_WRITE, uiop);
718 		if (error != 0) {
719 			freeb(mp);
720 			return (error);
721 		}
722 		mp->b_wptr = mp->b_rptr + size;
723 		count -= size;
724 		hdr_mp->b_cont = mp;
725 		hdr_mp = mp;
726 	}
727 	return (0);
728 }
729 
730 /*
731  * Send message.
732  */
733 static int
734 sosctp_sendmsg(struct sonode *so, struct nmsghdr *msg, struct uio *uiop,
735     struct cred *cr)
736 {
737 	mblk_t *mctl;
738 	struct cmsghdr *cmsg;
739 	struct sctp_sndrcvinfo *sinfo;
740 	int optlen, flags, fflag;
741 	ssize_t count, msglen;
742 	int error;
743 
744 	ASSERT(so->so_type == SOCK_STREAM);
745 
746 	flags = msg->msg_flags;
747 	if (flags & MSG_OOB) {
748 		/*
749 		 * No out-of-band data support.
750 		 */
751 		return (EOPNOTSUPP);
752 	}
753 
754 	if (msg->msg_controllen != 0) {
755 		optlen = msg->msg_controllen;
756 		cmsg = sosctp_find_cmsg(msg->msg_control, optlen, SCTP_SNDRCV);
757 		if (cmsg != NULL) {
758 			if (cmsg->cmsg_len <
759 			    (sizeof (*sinfo) + sizeof (*cmsg))) {
760 				eprintsoline(so, EINVAL);
761 				return (EINVAL);
762 			}
763 			sinfo = (struct sctp_sndrcvinfo *)(cmsg + 1);
764 
765 			/* Both flags should not be set together. */
766 			if ((sinfo->sinfo_flags & MSG_EOF) &&
767 			    (sinfo->sinfo_flags & MSG_ABORT)) {
768 				eprintsoline(so, EINVAL);
769 				return (EINVAL);
770 			}
771 
772 			/* Initiate a graceful shutdown. */
773 			if (sinfo->sinfo_flags & MSG_EOF) {
774 				/* Can't include data in MSG_EOF message. */
775 				if (uiop->uio_resid != 0) {
776 					eprintsoline(so, EINVAL);
777 					return (EINVAL);
778 				}
779 
780 				/*
781 				 * This is the same sequence as done in
782 				 * shutdown(SHUT_WR).
783 				 */
784 				mutex_enter(&so->so_lock);
785 				so_lock_single(so);
786 				socantsendmore(so);
787 				cv_broadcast(&so->so_snd_cv);
788 				so->so_state |= SS_ISDISCONNECTING;
789 				mutex_exit(&so->so_lock);
790 
791 				pollwakeup(&so->so_poll_list, POLLOUT);
792 				sctp_recvd((struct sctp_s *)so->so_proto_handle,
793 				    so->so_rcvbuf);
794 				error = sctp_disconnect(
795 				    (struct sctp_s *)so->so_proto_handle);
796 
797 				mutex_enter(&so->so_lock);
798 				so_unlock_single(so, SOLOCKED);
799 				mutex_exit(&so->so_lock);
800 				return (error);
801 			}
802 		}
803 	} else {
804 		optlen = 0;
805 	}
806 
807 	mutex_enter(&so->so_lock);
808 	for (;;) {
809 		if (so->so_state & SS_CANTSENDMORE) {
810 			mutex_exit(&so->so_lock);
811 			return (EPIPE);
812 		}
813 
814 		if (so->so_error != 0) {
815 			error = sogeterr(so, B_TRUE);
816 			mutex_exit(&so->so_lock);
817 			return (error);
818 		}
819 
820 		if (!so->so_snd_qfull)
821 			break;
822 
823 		if (so->so_state & SS_CLOSING) {
824 			mutex_exit(&so->so_lock);
825 			return (EINTR);
826 		}
827 		/*
828 		 * Xmit window full in a blocking socket.
829 		 */
830 		if ((uiop->uio_fmode & (FNDELAY|FNONBLOCK)) ||
831 		    (flags & MSG_DONTWAIT)) {
832 			mutex_exit(&so->so_lock);
833 			return (EAGAIN);
834 		} else {
835 			/*
836 			 * Wait for space to become available and try again.
837 			 */
838 			error = cv_wait_sig(&so->so_snd_cv, &so->so_lock);
839 			if (!error) { /* signal */
840 				mutex_exit(&so->so_lock);
841 				return (EINTR);
842 			}
843 		}
844 	}
845 	msglen = count = uiop->uio_resid;
846 
847 	/* Don't allow sending a message larger than the send buffer size. */
848 	/* XXX Transport module need to enforce this */
849 	if (msglen > so->so_sndbuf) {
850 		mutex_exit(&so->so_lock);
851 		return (EMSGSIZE);
852 	}
853 
854 	/*
855 	 * Allow piggybacking data on handshake messages (SS_ISCONNECTING).
856 	 */
857 	if (!(so->so_state & (SS_ISCONNECTING | SS_ISCONNECTED))) {
858 		/*
859 		 * We need to check here for listener so that the
860 		 * same error will be returned as with a TCP socket.
861 		 * In this case, sosctp_connect() returns EOPNOTSUPP
862 		 * while a TCP socket returns ENOTCONN instead.  Catch it
863 		 * here to have the same behavior as a TCP socket.
864 		 *
865 		 * We also need to make sure that the peer address is
866 		 * provided before we attempt to do the connect.
867 		 */
868 		if ((so->so_state & SS_ACCEPTCONN) ||
869 		    msg->msg_name == NULL) {
870 			mutex_exit(&so->so_lock);
871 			error = ENOTCONN;
872 			goto error_nofree;
873 		}
874 		mutex_exit(&so->so_lock);
875 		fflag = uiop->uio_fmode;
876 		if (flags & MSG_DONTWAIT) {
877 			fflag |= FNDELAY;
878 		}
879 		error = sosctp_connect(so, msg->msg_name, msg->msg_namelen,
880 		    fflag, (so->so_version == SOV_XPG4_2) * _SOCONNECT_XPG4_2,
881 		    cr);
882 		if (error) {
883 			/*
884 			 * Check for non-fatal errors, socket connected
885 			 * while the lock had been lifted.
886 			 */
887 			if (error != EISCONN && error != EALREADY) {
888 				goto error_nofree;
889 			}
890 			error = 0;
891 		}
892 	} else {
893 		mutex_exit(&so->so_lock);
894 	}
895 
896 	mctl = sctp_alloc_hdr(msg->msg_name, msg->msg_namelen,
897 	    msg->msg_control, optlen, SCTP_CAN_BLOCK);
898 	if (mctl == NULL) {
899 		error = EINTR;
900 		goto error_nofree;
901 	}
902 
903 	/* Copy in the message. */
904 	if ((error = sosctp_uiomove(mctl, count, so->so_proto_props.sopp_maxblk,
905 	    so->so_proto_props.sopp_wroff, uiop, flags)) != 0) {
906 		goto error_ret;
907 	}
908 	error = sctp_sendmsg((struct sctp_s *)so->so_proto_handle, mctl, 0);
909 	if (error == 0)
910 		return (0);
911 
912 error_ret:
913 	freemsg(mctl);
914 error_nofree:
915 	mutex_enter(&so->so_lock);
916 	if ((error == EPIPE) && (so->so_state & SS_CANTSENDMORE)) {
917 		/*
918 		 * We received shutdown between the time lock was
919 		 * lifted and call to sctp_sendmsg().
920 		 */
921 		mutex_exit(&so->so_lock);
922 		return (EPIPE);
923 	}
924 	mutex_exit(&so->so_lock);
925 	return (error);
926 }
927 
928 /*
929  * Send message on 1-N socket. Connects automatically if there is
930  * no association.
931  */
932 static int
933 sosctp_seq_sendmsg(struct sonode *so, struct nmsghdr *msg, struct uio *uiop,
934     struct cred *cr)
935 {
936 	struct sctp_sonode *ss;
937 	struct sctp_soassoc *ssa;
938 	struct cmsghdr *cmsg;
939 	struct sctp_sndrcvinfo *sinfo;
940 	int aid = 0;
941 	mblk_t *mctl;
942 	int namelen, optlen, flags;
943 	ssize_t count, msglen;
944 	int error;
945 	uint16_t s_flags = 0;
946 
947 	ASSERT(so->so_type == SOCK_SEQPACKET);
948 
949 	/*
950 	 * There shouldn't be problems with alignment, as the memory for
951 	 * msg_control was alloced with kmem_alloc.
952 	 */
953 	cmsg = sosctp_find_cmsg(msg->msg_control, msg->msg_controllen,
954 	    SCTP_SNDRCV);
955 	if (cmsg != NULL) {
956 		if (cmsg->cmsg_len < (sizeof (*sinfo) + sizeof (*cmsg))) {
957 			eprintsoline(so, EINVAL);
958 			return (EINVAL);
959 		}
960 		sinfo = (struct sctp_sndrcvinfo *)(cmsg + 1);
961 		s_flags = sinfo->sinfo_flags;
962 		aid = sinfo->sinfo_assoc_id;
963 	}
964 
965 	ss = SOTOSSO(so);
966 	namelen = msg->msg_namelen;
967 
968 	if (msg->msg_controllen > 0) {
969 		optlen = msg->msg_controllen;
970 	} else {
971 		optlen = 0;
972 	}
973 
974 	mutex_enter(&so->so_lock);
975 
976 	/*
977 	 * If there is no association id, connect to address specified
978 	 * in msg_name.  Otherwise look up the association using the id.
979 	 */
980 	if (aid == 0) {
981 		/*
982 		 * Connect and shutdown cannot be done together, so check for
983 		 * MSG_EOF.
984 		 */
985 		if (msg->msg_name == NULL || namelen == 0 ||
986 		    (s_flags & MSG_EOF)) {
987 			error = EINVAL;
988 			eprintsoline(so, error);
989 			goto done;
990 		}
991 		flags = uiop->uio_fmode;
992 		if (msg->msg_flags & MSG_DONTWAIT) {
993 			flags |= FNDELAY;
994 		}
995 		so_lock_single(so);
996 		error = sosctp_assoc_createconn(ss, msg->msg_name, namelen,
997 		    msg->msg_control, optlen, flags, cr, &ssa);
998 		if (error) {
999 			if ((so->so_version == SOV_XPG4_2) &&
1000 			    (error == EHOSTUNREACH)) {
1001 				error = ENETUNREACH;
1002 			}
1003 			if (ssa == NULL) {
1004 				/*
1005 				 * Fatal error during connect(). Bail out.
1006 				 * If ssa exists, it means that the handshake
1007 				 * is in progress.
1008 				 */
1009 				eprintsoline(so, error);
1010 				so_unlock_single(so, SOLOCKED);
1011 				goto done;
1012 			}
1013 			/*
1014 			 * All the errors are non-fatal ones, don't return
1015 			 * e.g. EINPROGRESS from sendmsg().
1016 			 */
1017 			error = 0;
1018 		}
1019 		so_unlock_single(so, SOLOCKED);
1020 	} else {
1021 		if ((error = sosctp_assoc(ss, aid, &ssa)) != 0) {
1022 			eprintsoline(so, error);
1023 			goto done;
1024 		}
1025 	}
1026 
1027 	/*
1028 	 * Now we have an association.
1029 	 */
1030 	flags = msg->msg_flags;
1031 
1032 	/*
1033 	 * MSG_EOF initiates graceful shutdown.
1034 	 */
1035 	if (s_flags & MSG_EOF) {
1036 		if (uiop->uio_resid) {
1037 			/*
1038 			 * Can't include data in MSG_EOF message.
1039 			 */
1040 			error = EINVAL;
1041 		} else {
1042 			mutex_exit(&so->so_lock);
1043 			ssa->ssa_state |= SS_ISDISCONNECTING;
1044 			sctp_recvd(ssa->ssa_conn, so->so_rcvbuf);
1045 			error = sctp_disconnect(ssa->ssa_conn);
1046 			mutex_enter(&so->so_lock);
1047 		}
1048 		goto refrele;
1049 	}
1050 
1051 	for (;;) {
1052 		if (ssa->ssa_state & SS_CANTSENDMORE) {
1053 			SSA_REFRELE(ss, ssa);
1054 			mutex_exit(&so->so_lock);
1055 			return (EPIPE);
1056 		}
1057 		if (ssa->ssa_error != 0) {
1058 			error = ssa->ssa_error;
1059 			ssa->ssa_error = 0;
1060 			goto refrele;
1061 		}
1062 
1063 		if (!ssa->ssa_snd_qfull)
1064 			break;
1065 
1066 		if (so->so_state & SS_CLOSING) {
1067 			error = EINTR;
1068 			goto refrele;
1069 		}
1070 		if ((uiop->uio_fmode & (FNDELAY|FNONBLOCK)) ||
1071 		    (flags & MSG_DONTWAIT)) {
1072 			error = EAGAIN;
1073 			goto refrele;
1074 		} else {
1075 			/*
1076 			 * Wait for space to become available and try again.
1077 			 */
1078 			error = cv_wait_sig(&so->so_snd_cv, &so->so_lock);
1079 			if (!error) { /* signal */
1080 				error = EINTR;
1081 				goto refrele;
1082 			}
1083 		}
1084 	}
1085 
1086 	msglen = count = uiop->uio_resid;
1087 
1088 	/* Don't allow sending a message larger than the send buffer size. */
1089 	if (msglen > so->so_sndbuf) {
1090 		error = EMSGSIZE;
1091 		goto refrele;
1092 	}
1093 
1094 	/*
1095 	 * Update TX buffer usage here so that we can lift the socket lock.
1096 	 */
1097 	mutex_exit(&so->so_lock);
1098 
1099 	mctl = sctp_alloc_hdr(msg->msg_name, namelen, msg->msg_control,
1100 	    optlen, SCTP_CAN_BLOCK);
1101 	if (mctl == NULL) {
1102 		error = EINTR;
1103 		goto lock_rele;
1104 	}
1105 
1106 	/* Copy in the message. */
1107 	if ((error = sosctp_uiomove(mctl, count, ssa->ssa_wrsize,
1108 	    ssa->ssa_wroff, uiop, flags)) != 0) {
1109 		goto lock_rele;
1110 	}
1111 	error = sctp_sendmsg((struct sctp_s *)ssa->ssa_conn, mctl, 0);
1112 lock_rele:
1113 	mutex_enter(&so->so_lock);
1114 	if (error != 0) {
1115 		freemsg(mctl);
1116 		if ((error == EPIPE) && (ssa->ssa_state & SS_CANTSENDMORE)) {
1117 			/*
1118 			 * We received shutdown between the time lock was
1119 			 * lifted and call to sctp_sendmsg().
1120 			 */
1121 			SSA_REFRELE(ss, ssa);
1122 			mutex_exit(&so->so_lock);
1123 			return (EPIPE);
1124 		}
1125 	}
1126 
1127 refrele:
1128 	SSA_REFRELE(ss, ssa);
1129 done:
1130 	mutex_exit(&so->so_lock);
1131 	return (error);
1132 }
1133 
1134 /*
1135  * Get address of remote node.
1136  */
1137 /* ARGSUSED */
1138 static int
1139 sosctp_getpeername(struct sonode *so, struct sockaddr *addr, socklen_t *addrlen,
1140     boolean_t accept, struct cred *cr)
1141 {
1142 	return (sctp_getpeername((struct sctp_s *)so->so_proto_handle, addr,
1143 	    addrlen));
1144 }
1145 
1146 /*
1147  * Get local address.
1148  */
1149 /* ARGSUSED */
1150 static int
1151 sosctp_getsockname(struct sonode *so, struct sockaddr *addr, socklen_t *addrlen,
1152     struct cred *cr)
1153 {
1154 	return (sctp_getsockname((struct sctp_s *)so->so_proto_handle, addr,
1155 	    addrlen));
1156 }
1157 
1158 /*
1159  * Called from shutdown().
1160  */
1161 /* ARGSUSED */
1162 static int
1163 sosctp_shutdown(struct sonode *so, int how, struct cred *cr)
1164 {
1165 	uint_t state_change;
1166 	int wakesig = 0;
1167 	int error = 0;
1168 
1169 	mutex_enter(&so->so_lock);
1170 	/*
1171 	 * Record the current state and then perform any state changes.
1172 	 * Then use the difference between the old and new states to
1173 	 * determine which needs to be done.
1174 	 */
1175 	state_change = so->so_state;
1176 
1177 	switch (how) {
1178 	case SHUT_RD:
1179 		socantrcvmore(so);
1180 		break;
1181 	case SHUT_WR:
1182 		socantsendmore(so);
1183 		break;
1184 	case SHUT_RDWR:
1185 		socantsendmore(so);
1186 		socantrcvmore(so);
1187 		break;
1188 	default:
1189 		mutex_exit(&so->so_lock);
1190 		return (EINVAL);
1191 	}
1192 
1193 	state_change = so->so_state & ~state_change;
1194 
1195 	if (state_change & SS_CANTRCVMORE) {
1196 		if (so->so_rcv_q_head == NULL) {
1197 			cv_signal(&so->so_rcv_cv);
1198 		}
1199 		wakesig = POLLIN|POLLRDNORM;
1200 
1201 		socket_sendsig(so, SOCKETSIG_READ);
1202 	}
1203 	if (state_change & SS_CANTSENDMORE) {
1204 		cv_broadcast(&so->so_snd_cv);
1205 		wakesig |= POLLOUT;
1206 
1207 		so->so_state |= SS_ISDISCONNECTING;
1208 	}
1209 	mutex_exit(&so->so_lock);
1210 
1211 	pollwakeup(&so->so_poll_list, wakesig);
1212 
1213 	if (state_change & SS_CANTSENDMORE) {
1214 		sctp_recvd((struct sctp_s *)so->so_proto_handle, so->so_rcvbuf);
1215 		error = sctp_disconnect((struct sctp_s *)so->so_proto_handle);
1216 	}
1217 
1218 	/*
1219 	 * HACK: sctp_disconnect() may return EWOULDBLOCK.  But this error is
1220 	 * not documented in standard socket API.  Catch it here.
1221 	 */
1222 	if (error == EWOULDBLOCK)
1223 		error = 0;
1224 	return (error);
1225 }
1226 
1227 /*
1228  * Get socket options.
1229  */
1230 /*ARGSUSED5*/
1231 static int
1232 sosctp_getsockopt(struct sonode *so, int level, int option_name,
1233     void *optval, socklen_t *optlenp, int flags, struct cred *cr)
1234 {
1235 	socklen_t maxlen = *optlenp;
1236 	socklen_t len;
1237 	socklen_t optlen;
1238 	uint8_t	buffer[4];
1239 	void	*optbuf = &buffer;
1240 	int	error = 0;
1241 
1242 	if (level == SOL_SOCKET) {
1243 		switch (option_name) {
1244 		/* Not supported options */
1245 		case SO_SNDTIMEO:
1246 		case SO_RCVTIMEO:
1247 		case SO_EXCLBIND:
1248 			eprintsoline(so, ENOPROTOOPT);
1249 			return (ENOPROTOOPT);
1250 		default:
1251 			error = socket_getopt_common(so, level, option_name,
1252 			    optval, optlenp, flags);
1253 			if (error >= 0)
1254 				return (error);
1255 			/* Pass the request to the protocol */
1256 			break;
1257 		}
1258 	}
1259 
1260 	if (level == IPPROTO_SCTP) {
1261 		/*
1262 		 * Should go through ioctl().
1263 		 */
1264 		return (EINVAL);
1265 	}
1266 
1267 	if (maxlen > sizeof (buffer)) {
1268 		optbuf = kmem_alloc(maxlen, KM_SLEEP);
1269 	}
1270 	optlen = maxlen;
1271 
1272 	/*
1273 	 * If the resulting optlen is greater than the provided maxlen, then
1274 	 * we sliently trucate.
1275 	 */
1276 	error = sctp_get_opt((struct sctp_s *)so->so_proto_handle, level,
1277 	    option_name, optbuf, &optlen);
1278 
1279 	if (error != 0) {
1280 		eprintsoline(so, error);
1281 		goto free;
1282 	}
1283 	len = optlen;
1284 
1285 copyout:
1286 
1287 	len = MIN(len, maxlen);
1288 	bcopy(optbuf, optval, len);
1289 	*optlenp = optlen;
1290 free:
1291 	if (optbuf != &buffer) {
1292 		kmem_free(optbuf, maxlen);
1293 	}
1294 
1295 	return (error);
1296 }
1297 
1298 /*
1299  * Set socket options
1300  */
1301 /* ARGSUSED */
1302 static int
1303 sosctp_setsockopt(struct sonode *so, int level, int option_name,
1304     const void *optval, t_uscalar_t optlen, struct cred *cr)
1305 {
1306 	struct sctp_sonode *ss = SOTOSSO(so);
1307 	struct sctp_soassoc *ssa = NULL;
1308 	sctp_assoc_t id;
1309 	int error, rc;
1310 	void *conn = NULL;
1311 
1312 	mutex_enter(&so->so_lock);
1313 
1314 	/*
1315 	 * For some SCTP level options, one can select the association this
1316 	 * applies to.
1317 	 */
1318 	if (so->so_type == SOCK_STREAM) {
1319 		conn = so->so_proto_handle;
1320 	} else {
1321 		/*
1322 		 * SOCK_SEQPACKET only
1323 		 */
1324 		id = 0;
1325 		if (level == IPPROTO_SCTP) {
1326 			switch (option_name) {
1327 			case SCTP_RTOINFO:
1328 			case SCTP_ASSOCINFO:
1329 			case SCTP_SET_PEER_PRIMARY_ADDR:
1330 			case SCTP_PRIMARY_ADDR:
1331 			case SCTP_PEER_ADDR_PARAMS:
1332 				/*
1333 				 * Association ID is the first element
1334 				 * params struct
1335 				 */
1336 				if (optlen < sizeof (sctp_assoc_t)) {
1337 					error = EINVAL;
1338 					eprintsoline(so, error);
1339 					goto done;
1340 				}
1341 				id = *(sctp_assoc_t *)optval;
1342 				break;
1343 			case SCTP_DEFAULT_SEND_PARAM:
1344 				if (optlen != sizeof (struct sctp_sndrcvinfo)) {
1345 					error = EINVAL;
1346 					eprintsoline(so, error);
1347 					goto done;
1348 				}
1349 				id = ((struct sctp_sndrcvinfo *)
1350 				    optval)->sinfo_assoc_id;
1351 				break;
1352 			case SCTP_INITMSG:
1353 				/*
1354 				 * Only applies to future associations
1355 				 */
1356 				conn = so->so_proto_handle;
1357 				break;
1358 			default:
1359 				break;
1360 			}
1361 		} else if (level == SOL_SOCKET) {
1362 			if (option_name == SO_LINGER) {
1363 				error = EOPNOTSUPP;
1364 				eprintsoline(so, error);
1365 				goto done;
1366 			}
1367 			/*
1368 			 * These 2 options are applied to all associations.
1369 			 * The other socket level options are only applied
1370 			 * to the socket (not associations).
1371 			 */
1372 			if ((option_name != SO_RCVBUF) &&
1373 			    (option_name != SO_SNDBUF)) {
1374 				conn = so->so_proto_handle;
1375 			}
1376 		} else {
1377 			conn = NULL;
1378 		}
1379 
1380 		/*
1381 		 * If association ID was specified, do op on that assoc.
1382 		 * Otherwise set the default setting of a socket.
1383 		 */
1384 		if (id != 0) {
1385 			if ((error = sosctp_assoc(ss, id, &ssa)) != 0) {
1386 				eprintsoline(so, error);
1387 				goto done;
1388 			}
1389 			conn = ssa->ssa_conn;
1390 		}
1391 	}
1392 	dprint(2, ("sosctp_setsockopt %p (%d) - conn %p %d %d id:%d\n",
1393 	    (void *)ss, so->so_type, (void *)conn, level, option_name, id));
1394 
1395 	ASSERT(ssa == NULL || (ssa != NULL && conn != NULL));
1396 	if (conn != NULL) {
1397 		mutex_exit(&so->so_lock);
1398 		error = sctp_set_opt((struct sctp_s *)conn, level, option_name,
1399 		    optval, optlen);
1400 		mutex_enter(&so->so_lock);
1401 		if (ssa != NULL)
1402 			SSA_REFRELE(ss, ssa);
1403 	} else {
1404 		/*
1405 		 * 1-N socket, and we have to apply the operation to ALL
1406 		 * associations. Like with anything of this sort, the
1407 		 * problem is what to do if the operation fails.
1408 		 * Just try to apply the setting to everyone, but store
1409 		 * error number if someone returns such.  And since we are
1410 		 * looping through all possible aids, some of them can be
1411 		 * invalid.  We just ignore this kind (sosctp_assoc()) of
1412 		 * errors.
1413 		 */
1414 		sctp_assoc_t aid;
1415 
1416 		mutex_exit(&so->so_lock);
1417 		error = sctp_set_opt((struct sctp_s *)so->so_proto_handle,
1418 		    level, option_name, optval, optlen);
1419 		mutex_enter(&so->so_lock);
1420 		for (aid = 1; aid < ss->ss_maxassoc; aid++) {
1421 			if (sosctp_assoc(ss, aid, &ssa) != 0)
1422 				continue;
1423 			mutex_exit(&so->so_lock);
1424 			rc = sctp_set_opt((struct sctp_s *)ssa->ssa_conn, level,
1425 			    option_name, optval, optlen);
1426 			mutex_enter(&so->so_lock);
1427 			SSA_REFRELE(ss, ssa);
1428 			if (error == 0) {
1429 				error = rc;
1430 			}
1431 		}
1432 	}
1433 done:
1434 	mutex_exit(&so->so_lock);
1435 	return (error);
1436 }
1437 
1438 /*ARGSUSED*/
1439 static int
1440 sosctp_ioctl(struct sonode *so, int cmd, intptr_t arg, int mode,
1441     struct cred *cr, int32_t *rvalp)
1442 {
1443 	struct sctp_sonode	*ss;
1444 	int32_t			value;
1445 	int			error;
1446 	int			intval;
1447 	pid_t			pid;
1448 	struct sctp_soassoc	*ssa;
1449 	void			*conn;
1450 	void			*buf;
1451 	STRUCT_DECL(sctpopt, opt);
1452 	uint32_t		optlen;
1453 	int			buflen;
1454 
1455 	ss = SOTOSSO(so);
1456 
1457 	/* handle socket specific ioctls */
1458 	switch (cmd) {
1459 	case FIONBIO:
1460 		if (so_copyin((void *)arg, &value, sizeof (int32_t),
1461 		    (mode & (int)FKIOCTL))) {
1462 			return (EFAULT);
1463 		}
1464 		mutex_enter(&so->so_lock);
1465 		if (value) {
1466 			so->so_state |= SS_NDELAY;
1467 		} else {
1468 			so->so_state &= ~SS_NDELAY;
1469 		}
1470 		mutex_exit(&so->so_lock);
1471 		return (0);
1472 
1473 	case FIOASYNC:
1474 		if (so_copyin((void *)arg, &value, sizeof (int32_t),
1475 		    (mode & (int)FKIOCTL))) {
1476 			return (EFAULT);
1477 		}
1478 		mutex_enter(&so->so_lock);
1479 
1480 		if (value) {
1481 			/* Turn on SIGIO */
1482 			so->so_state |= SS_ASYNC;
1483 		} else {
1484 			/* Turn off SIGIO */
1485 			so->so_state &= ~SS_ASYNC;
1486 		}
1487 		mutex_exit(&so->so_lock);
1488 		return (0);
1489 
1490 	case SIOCSPGRP:
1491 	case FIOSETOWN:
1492 		if (so_copyin((void *)arg, &pid, sizeof (pid_t),
1493 		    (mode & (int)FKIOCTL))) {
1494 			return (EFAULT);
1495 		}
1496 		mutex_enter(&so->so_lock);
1497 
1498 		error = (pid != so->so_pgrp) ? socket_chgpgrp(so, pid) : 0;
1499 		mutex_exit(&so->so_lock);
1500 		return (error);
1501 
1502 	case SIOCGPGRP:
1503 	case FIOGETOWN:
1504 		if (so_copyout(&so->so_pgrp, (void *)arg,
1505 		    sizeof (pid_t), (mode & (int)FKIOCTL)))
1506 			return (EFAULT);
1507 		return (0);
1508 
1509 	case FIONREAD:
1510 		/* XXX: Cannot be used unless standard buffer is used */
1511 		/*
1512 		 * Return number of bytes of data in all data messages
1513 		 * in queue in "arg".
1514 		 * For stream socket, amount of available data.
1515 		 * For sock_dgram, # of available bytes + addresses.
1516 		 */
1517 		intval = (so->so_state & SS_ACCEPTCONN) ? 0 :
1518 		    MIN(so->so_rcv_queued, INT_MAX);
1519 		if (so_copyout(&intval, (void *)arg, sizeof (intval),
1520 		    (mode & (int)FKIOCTL)))
1521 			return (EFAULT);
1522 		return (0);
1523 	case SIOCATMARK:
1524 		/*
1525 		 * No support for urgent data.
1526 		 */
1527 		intval = 0;
1528 
1529 		if (so_copyout(&intval, (void *)arg, sizeof (int),
1530 		    (mode & (int)FKIOCTL)))
1531 			return (EFAULT);
1532 		return (0);
1533 	case _I_GETPEERCRED: {
1534 		int error = 0;
1535 
1536 		if ((mode & FKIOCTL) == 0)
1537 			return (EINVAL);
1538 
1539 		mutex_enter(&so->so_lock);
1540 		if ((so->so_mode & SM_CONNREQUIRED) == 0) {
1541 			error = ENOTSUP;
1542 		} else if ((so->so_state & SS_ISCONNECTED) == 0) {
1543 			error = ENOTCONN;
1544 		} else if (so->so_peercred != NULL) {
1545 			k_peercred_t *kp = (k_peercred_t *)arg;
1546 			kp->pc_cr = so->so_peercred;
1547 			kp->pc_cpid = so->so_cpid;
1548 			crhold(so->so_peercred);
1549 		} else {
1550 			error = EINVAL;
1551 		}
1552 		mutex_exit(&so->so_lock);
1553 		return (error);
1554 	}
1555 	case SIOCSCTPGOPT:
1556 		STRUCT_INIT(opt, mode);
1557 
1558 		if (so_copyin((void *)arg, STRUCT_BUF(opt), STRUCT_SIZE(opt),
1559 		    (mode & (int)FKIOCTL))) {
1560 			return (EFAULT);
1561 		}
1562 		if ((optlen = STRUCT_FGET(opt, sopt_len)) > SO_MAXARGSIZE)
1563 			return (EINVAL);
1564 
1565 		/*
1566 		 * Find the correct sctp_t based on whether it is 1-N socket
1567 		 * or not.
1568 		 */
1569 		intval = STRUCT_FGET(opt, sopt_aid);
1570 		mutex_enter(&so->so_lock);
1571 		if ((so->so_type == SOCK_SEQPACKET) && intval) {
1572 			if ((error = sosctp_assoc(ss, intval, &ssa)) != 0) {
1573 				mutex_exit(&so->so_lock);
1574 				return (error);
1575 			}
1576 			conn = ssa->ssa_conn;
1577 			ASSERT(conn != NULL);
1578 		} else {
1579 			conn = so->so_proto_handle;
1580 			ssa = NULL;
1581 		}
1582 		mutex_exit(&so->so_lock);
1583 
1584 		/* Copyin the option buffer and then call sctp_get_opt(). */
1585 		buflen = optlen;
1586 		/* Let's allocate a buffer enough to hold an int */
1587 		if (buflen < sizeof (uint32_t))
1588 			buflen = sizeof (uint32_t);
1589 		buf = kmem_alloc(buflen, KM_SLEEP);
1590 		if (so_copyin(STRUCT_FGETP(opt, sopt_val), buf, optlen,
1591 		    (mode & (int)FKIOCTL))) {
1592 			if (ssa != NULL) {
1593 				mutex_enter(&so->so_lock);
1594 				SSA_REFRELE(ss, ssa);
1595 				mutex_exit(&so->so_lock);
1596 			}
1597 			kmem_free(buf, buflen);
1598 			return (EFAULT);
1599 		}
1600 		/* The option level has to be IPPROTO_SCTP */
1601 		error = sctp_get_opt((struct sctp_s *)conn, IPPROTO_SCTP,
1602 		    STRUCT_FGET(opt, sopt_name), buf, &optlen);
1603 		if (ssa != NULL) {
1604 			mutex_enter(&so->so_lock);
1605 			SSA_REFRELE(ss, ssa);
1606 			mutex_exit(&so->so_lock);
1607 		}
1608 		optlen = MIN(buflen, optlen);
1609 		/* No error, copyout the result with the correct buf len. */
1610 		if (error == 0) {
1611 			STRUCT_FSET(opt, sopt_len, optlen);
1612 			if (so_copyout(STRUCT_BUF(opt), (void *)arg,
1613 			    STRUCT_SIZE(opt), (mode & (int)FKIOCTL))) {
1614 				error = EFAULT;
1615 			} else if (so_copyout(buf, STRUCT_FGETP(opt, sopt_val),
1616 			    optlen, (mode & (int)FKIOCTL))) {
1617 				error = EFAULT;
1618 			}
1619 		}
1620 		kmem_free(buf, buflen);
1621 		return (error);
1622 
1623 	case SIOCSCTPSOPT:
1624 		STRUCT_INIT(opt, mode);
1625 
1626 		if (so_copyin((void *)arg, STRUCT_BUF(opt), STRUCT_SIZE(opt),
1627 		    (mode & (int)FKIOCTL))) {
1628 			return (EFAULT);
1629 		}
1630 		if ((optlen = STRUCT_FGET(opt, sopt_len)) > SO_MAXARGSIZE)
1631 			return (EINVAL);
1632 
1633 		/*
1634 		 * Find the correct sctp_t based on whether it is 1-N socket
1635 		 * or not.
1636 		 */
1637 		intval = STRUCT_FGET(opt, sopt_aid);
1638 		mutex_enter(&so->so_lock);
1639 		if (intval != 0) {
1640 			if ((error = sosctp_assoc(ss, intval, &ssa)) != 0) {
1641 				mutex_exit(&so->so_lock);
1642 				return (error);
1643 			}
1644 			conn = ssa->ssa_conn;
1645 			ASSERT(conn != NULL);
1646 		} else {
1647 			conn = so->so_proto_handle;
1648 			ssa = NULL;
1649 		}
1650 		mutex_exit(&so->so_lock);
1651 
1652 		/* Copyin the option buffer and then call sctp_set_opt(). */
1653 		buf = kmem_alloc(optlen, KM_SLEEP);
1654 		if (so_copyin(STRUCT_FGETP(opt, sopt_val), buf, optlen,
1655 		    (mode & (int)FKIOCTL))) {
1656 			if (ssa != NULL) {
1657 				mutex_enter(&so->so_lock);
1658 				SSA_REFRELE(ss, ssa);
1659 				mutex_exit(&so->so_lock);
1660 			}
1661 			kmem_free(buf, intval);
1662 			return (EFAULT);
1663 		}
1664 		/* The option level has to be IPPROTO_SCTP */
1665 		error = sctp_set_opt((struct sctp_s *)conn, IPPROTO_SCTP,
1666 		    STRUCT_FGET(opt, sopt_name), buf, optlen);
1667 		if (ssa) {
1668 			mutex_enter(&so->so_lock);
1669 			SSA_REFRELE(ss, ssa);
1670 			mutex_exit(&so->so_lock);
1671 		}
1672 		kmem_free(buf, optlen);
1673 		return (error);
1674 
1675 	case SIOCSCTPPEELOFF: {
1676 		struct sonode *nso;
1677 		struct sctp_uc_swap us;
1678 		int nfd;
1679 		struct file *nfp;
1680 		struct vnode *nvp = NULL;
1681 		struct sockparams *sp;
1682 
1683 		dprint(2, ("sctppeeloff %p\n", (void *)ss));
1684 
1685 		if (so->so_type != SOCK_SEQPACKET) {
1686 			return (EOPNOTSUPP);
1687 		}
1688 		if (so_copyin((void *)arg, &intval, sizeof (intval),
1689 		    (mode & (int)FKIOCTL))) {
1690 			return (EFAULT);
1691 		}
1692 		if (intval == 0) {
1693 			return (EINVAL);
1694 		}
1695 
1696 		/*
1697 		 * Find sockparams. This is different from parent's entry,
1698 		 * as the socket type is different.
1699 		 */
1700 		error = solookup(so->so_family, SOCK_STREAM, so->so_protocol,
1701 		    &sp);
1702 		if (error != 0)
1703 			return (error);
1704 
1705 		/*
1706 		 * Allocate the user fd.
1707 		 */
1708 		if ((nfd = ufalloc(0)) == -1) {
1709 			eprintsoline(so, EMFILE);
1710 			SOCKPARAMS_DEC_REF(sp);
1711 			return (EMFILE);
1712 		}
1713 
1714 		/*
1715 		 * Copy the fd out.
1716 		 */
1717 		if (so_copyout(&nfd, (void *)arg, sizeof (nfd),
1718 		    (mode & (int)FKIOCTL))) {
1719 			error = EFAULT;
1720 			goto err;
1721 		}
1722 		mutex_enter(&so->so_lock);
1723 
1724 		/*
1725 		 * Don't use sosctp_assoc() in order to peel off disconnected
1726 		 * associations.
1727 		 */
1728 		ssa = ((uint32_t)intval >= ss->ss_maxassoc) ? NULL :
1729 		    ss->ss_assocs[intval].ssi_assoc;
1730 		if (ssa == NULL) {
1731 			mutex_exit(&so->so_lock);
1732 			error = EINVAL;
1733 			goto err;
1734 		}
1735 		SSA_REFHOLD(ssa);
1736 
1737 		nso = socksctp_create(sp, so->so_family, SOCK_STREAM,
1738 		    so->so_protocol, so->so_version, SOCKET_NOSLEEP,
1739 		    &error, cr);
1740 		if (nso == NULL) {
1741 			SSA_REFRELE(ss, ssa);
1742 			mutex_exit(&so->so_lock);
1743 			goto err;
1744 		}
1745 		nvp = SOTOV(nso);
1746 		so_lock_single(so);
1747 		mutex_exit(&so->so_lock);
1748 
1749 		/* cannot fail, only inheriting properties */
1750 		(void) sosctp_init(nso, so, CRED(), 0);
1751 
1752 		/*
1753 		 * We have a single ref on the new socket. This is normally
1754 		 * handled by socket_{create,newconn}, but since they are not
1755 		 * used we have to do it here.
1756 		 */
1757 		nso->so_count = 1;
1758 
1759 		us.sus_handle = nso;
1760 		us.sus_upcalls = &sosctp_sock_upcalls;
1761 
1762 		/*
1763 		 * Upcalls to new socket are blocked for the duration of
1764 		 * downcall.
1765 		 */
1766 		mutex_enter(&nso->so_lock);
1767 
1768 		error = sctp_set_opt((struct sctp_s *)ssa->ssa_conn,
1769 		    IPPROTO_SCTP, SCTP_UC_SWAP, &us, sizeof (us));
1770 		if (error) {
1771 			goto peelerr;
1772 		}
1773 		error = falloc(nvp, FWRITE|FREAD, &nfp, NULL);
1774 		if (error) {
1775 			goto peelerr;
1776 		}
1777 
1778 		/*
1779 		 * fill in the entries that falloc reserved
1780 		 */
1781 		nfp->f_vnode = nvp;
1782 		mutex_exit(&nfp->f_tlock);
1783 		setf(nfd, nfp);
1784 
1785 		mutex_enter(&so->so_lock);
1786 
1787 		sosctp_assoc_move(ss, SOTOSSO(nso), ssa);
1788 
1789 		mutex_exit(&nso->so_lock);
1790 
1791 		ssa->ssa_conn = NULL;
1792 		sosctp_assoc_free(ss, ssa);
1793 
1794 		so_unlock_single(so, SOLOCKED);
1795 		mutex_exit(&so->so_lock);
1796 
1797 		return (0);
1798 
1799 err:
1800 		SOCKPARAMS_DEC_REF(sp);
1801 		setf(nfd, NULL);
1802 		eprintsoline(so, error);
1803 		return (error);
1804 
1805 peelerr:
1806 		mutex_exit(&nso->so_lock);
1807 		mutex_enter(&so->so_lock);
1808 		ASSERT(nso->so_count == 1);
1809 		nso->so_count = 0;
1810 		so_unlock_single(so, SOLOCKED);
1811 		SSA_REFRELE(ss, ssa);
1812 		mutex_exit(&so->so_lock);
1813 
1814 		setf(nfd, NULL);
1815 		ASSERT(nvp->v_count == 1);
1816 		socket_destroy(nso);
1817 		eprintsoline(so, error);
1818 		return (error);
1819 	}
1820 	default:
1821 		return (EINVAL);
1822 	}
1823 }
1824 
1825 /*ARGSUSED*/
1826 static int
1827 sosctp_close(struct sonode *so, int flag, struct cred *cr)
1828 {
1829 	struct sctp_sonode *ss;
1830 	struct sctp_sa_id *ssi;
1831 	struct sctp_soassoc *ssa;
1832 	int32_t i;
1833 
1834 	ss = SOTOSSO(so);
1835 
1836 	/*
1837 	 * Initiate connection shutdown.  Tell SCTP if there is any data
1838 	 * left unread.
1839 	 */
1840 	sctp_recvd((struct sctp_s *)so->so_proto_handle,
1841 	    so->so_rcvbuf - so->so_rcv_queued);
1842 	(void) sctp_disconnect((struct sctp_s *)so->so_proto_handle);
1843 
1844 	/*
1845 	 * New associations can't come in, but old ones might get
1846 	 * closed in upcall. Protect against that by taking a reference
1847 	 * on the association.
1848 	 */
1849 	mutex_enter(&so->so_lock);
1850 	ssi = ss->ss_assocs;
1851 	for (i = 0; i < ss->ss_maxassoc; i++, ssi++) {
1852 		if ((ssa = ssi->ssi_assoc) != NULL) {
1853 			SSA_REFHOLD(ssa);
1854 			sosctp_assoc_isdisconnected(ssa, 0);
1855 			mutex_exit(&so->so_lock);
1856 
1857 			sctp_recvd(ssa->ssa_conn, so->so_rcvbuf -
1858 			    ssa->ssa_rcv_queued);
1859 			(void) sctp_disconnect(ssa->ssa_conn);
1860 
1861 			mutex_enter(&so->so_lock);
1862 			SSA_REFRELE(ss, ssa);
1863 		}
1864 	}
1865 	mutex_exit(&so->so_lock);
1866 
1867 	return (0);
1868 }
1869 
1870 /*
1871  * Closes incoming connections which were never accepted, frees
1872  * resources.
1873  */
1874 /* ARGSUSED */
1875 void
1876 sosctp_fini(struct sonode *so, struct cred *cr)
1877 {
1878 	struct sctp_sonode *ss;
1879 	struct sctp_sa_id *ssi;
1880 	struct sctp_soassoc *ssa;
1881 	int32_t i;
1882 
1883 	ss = SOTOSSO(so);
1884 
1885 	ASSERT(so->so_ops == &sosctp_sonodeops ||
1886 	    so->so_ops == &sosctp_seq_sonodeops);
1887 
1888 	/* We are the sole owner of so now */
1889 	mutex_enter(&so->so_lock);
1890 
1891 	/* Free all pending connections */
1892 	so_acceptq_flush(so, B_TRUE);
1893 
1894 	ssi = ss->ss_assocs;
1895 	for (i = 0; i < ss->ss_maxassoc; i++, ssi++) {
1896 		if ((ssa = ssi->ssi_assoc) != NULL) {
1897 			SSA_REFHOLD(ssa);
1898 			mutex_exit(&so->so_lock);
1899 
1900 			sctp_close((struct sctp_s *)ssa->ssa_conn);
1901 
1902 			mutex_enter(&so->so_lock);
1903 			ssa->ssa_conn = NULL;
1904 			sosctp_assoc_free(ss, ssa);
1905 		}
1906 	}
1907 	if (ss->ss_assocs != NULL) {
1908 		ASSERT(ss->ss_assoccnt == 0);
1909 		kmem_free(ss->ss_assocs,
1910 		    ss->ss_maxassoc * sizeof (struct sctp_sa_id));
1911 	}
1912 	mutex_exit(&so->so_lock);
1913 
1914 	if (so->so_proto_handle)
1915 		sctp_close((struct sctp_s *)so->so_proto_handle);
1916 	so->so_proto_handle = NULL;
1917 
1918 	/*
1919 	 * Note until sctp_close() is called, SCTP can still send up
1920 	 * messages, such as event notifications.  So we should flush
1921 	 * the recevie buffer after calling sctp_close().
1922 	 */
1923 	mutex_enter(&so->so_lock);
1924 	so_rcv_flush(so);
1925 	mutex_exit(&so->so_lock);
1926 
1927 	sonode_fini(so);
1928 }
1929 
1930 /*
1931  * Upcalls from SCTP
1932  */
1933 
1934 /*
1935  * This is the upcall function for 1-N (SOCK_SEQPACKET) socket when a new
1936  * association is created.  Note that the first argument (handle) is of type
1937  * sctp_sonode *, which is the one changed to a listener for new
1938  * associations.  All the other upcalls for 1-N socket take sctp_soassoc *
1939  * as handle.  The only exception is the su_properties upcall, which
1940  * can take both types as handle.
1941  */
1942 /* ARGSUSED */
1943 sock_upper_handle_t
1944 sctp_assoc_newconn(sock_upper_handle_t parenthandle,
1945     sock_lower_handle_t connind, sock_downcalls_t *dc,
1946     struct cred *peer_cred, pid_t peer_cpid, sock_upcalls_t **ucp)
1947 {
1948 	struct sctp_sonode *lss = (struct sctp_sonode *)parenthandle;
1949 	struct sonode *lso = &lss->ss_so;
1950 	struct sctp_soassoc *ssa;
1951 	sctp_assoc_t id;
1952 
1953 	ASSERT(lss->ss_type == SOSCTP_SOCKET);
1954 	ASSERT(lso->so_state & SS_ACCEPTCONN);
1955 	ASSERT(lso->so_proto_handle != NULL); /* closed conn */
1956 	ASSERT(lso->so_type == SOCK_SEQPACKET);
1957 
1958 	mutex_enter(&lso->so_lock);
1959 
1960 	if ((id = sosctp_aid_get(lss)) == -1) {
1961 		/*
1962 		 * Array not large enough; increase size.
1963 		 */
1964 		if (sosctp_aid_grow(lss, lss->ss_maxassoc, KM_NOSLEEP) < 0) {
1965 			mutex_exit(&lso->so_lock);
1966 			return (NULL);
1967 		}
1968 		id = sosctp_aid_get(lss);
1969 		ASSERT(id != -1);
1970 	}
1971 
1972 	/*
1973 	 * Create soassoc for this connection
1974 	 */
1975 	ssa = sosctp_assoc_create(lss, KM_NOSLEEP);
1976 	if (ssa == NULL) {
1977 		mutex_exit(&lso->so_lock);
1978 		return (NULL);
1979 	}
1980 	sosctp_aid_reserve(lss, id, 1);
1981 	lss->ss_assocs[id].ssi_assoc = ssa;
1982 	++lss->ss_assoccnt;
1983 	ssa->ssa_id = id;
1984 	ssa->ssa_conn = (struct sctp_s *)connind;
1985 	ssa->ssa_state = (SS_ISBOUND | SS_ISCONNECTED);
1986 	ssa->ssa_wroff = lss->ss_wroff;
1987 	ssa->ssa_wrsize = lss->ss_wrsize;
1988 
1989 	mutex_exit(&lso->so_lock);
1990 
1991 	*ucp = &sosctp_assoc_upcalls;
1992 
1993 	return ((sock_upper_handle_t)ssa);
1994 }
1995 
1996 /* ARGSUSED */
1997 static void
1998 sctp_assoc_connected(sock_upper_handle_t handle, sock_connid_t id,
1999     struct cred *peer_cred, pid_t peer_cpid)
2000 {
2001 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2002 	struct sonode *so = &ssa->ssa_sonode->ss_so;
2003 
2004 	ASSERT(so->so_type == SOCK_SEQPACKET);
2005 	ASSERT(ssa->ssa_conn);
2006 
2007 	mutex_enter(&so->so_lock);
2008 	sosctp_assoc_isconnected(ssa);
2009 	mutex_exit(&so->so_lock);
2010 }
2011 
2012 /* ARGSUSED */
2013 static int
2014 sctp_assoc_disconnected(sock_upper_handle_t handle, sock_connid_t id, int error)
2015 {
2016 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2017 	struct sonode *so = &ssa->ssa_sonode->ss_so;
2018 	int ret;
2019 
2020 	ASSERT(so->so_type == SOCK_SEQPACKET);
2021 	ASSERT(ssa->ssa_conn != NULL);
2022 
2023 	mutex_enter(&so->so_lock);
2024 	sosctp_assoc_isdisconnected(ssa, error);
2025 	if (ssa->ssa_refcnt == 1) {
2026 		ret = 1;
2027 		ssa->ssa_conn = NULL;
2028 	} else {
2029 		ret = 0;
2030 	}
2031 	SSA_REFRELE(SOTOSSO(so), ssa);
2032 
2033 	cv_broadcast(&so->so_snd_cv);
2034 
2035 	mutex_exit(&so->so_lock);
2036 
2037 	return (ret);
2038 }
2039 
2040 /* ARGSUSED */
2041 static void
2042 sctp_assoc_disconnecting(sock_upper_handle_t handle, sock_opctl_action_t action,
2043     uintptr_t arg)
2044 {
2045 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2046 	struct sonode *so = &ssa->ssa_sonode->ss_so;
2047 
2048 	ASSERT(so->so_type == SOCK_SEQPACKET);
2049 	ASSERT(ssa->ssa_conn != NULL);
2050 	ASSERT(action == SOCK_OPCTL_SHUT_SEND);
2051 
2052 	mutex_enter(&so->so_lock);
2053 	sosctp_assoc_isdisconnecting(ssa);
2054 	mutex_exit(&so->so_lock);
2055 }
2056 
2057 /* ARGSUSED */
2058 static ssize_t
2059 sctp_assoc_recv(sock_upper_handle_t handle, mblk_t *mp, size_t len, int flags,
2060     int *errorp, boolean_t *forcepush)
2061 {
2062 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2063 	struct sctp_sonode *ss = ssa->ssa_sonode;
2064 	struct sonode *so = &ss->ss_so;
2065 	struct T_unitdata_ind *tind;
2066 	mblk_t *mp2;
2067 	union sctp_notification *sn;
2068 	struct sctp_sndrcvinfo *sinfo;
2069 	ssize_t space_available;
2070 
2071 	ASSERT(ssa->ssa_type == SOSCTP_ASSOC);
2072 	ASSERT(so->so_type == SOCK_SEQPACKET);
2073 	ASSERT(ssa->ssa_conn != NULL); /* closed conn */
2074 	ASSERT(mp != NULL);
2075 
2076 	ASSERT(errorp != NULL);
2077 	*errorp = 0;
2078 
2079 	/*
2080 	 * Should be getting T_unitdata_req's only.
2081 	 * Must have address as part of packet.
2082 	 */
2083 	tind = (struct T_unitdata_ind *)mp->b_rptr;
2084 	ASSERT((DB_TYPE(mp) == M_PROTO) &&
2085 	    (tind->PRIM_type == T_UNITDATA_IND));
2086 	ASSERT(tind->SRC_length);
2087 
2088 	mutex_enter(&so->so_lock);
2089 
2090 	/*
2091 	 * For notify messages, need to fill in association id.
2092 	 * For data messages, sndrcvinfo could be in ancillary data.
2093 	 */
2094 	if (mp->b_flag & SCTP_NOTIFICATION) {
2095 		mp2 = mp->b_cont;
2096 		sn = (union sctp_notification *)mp2->b_rptr;
2097 		switch (sn->sn_header.sn_type) {
2098 		case SCTP_ASSOC_CHANGE:
2099 			sn->sn_assoc_change.sac_assoc_id = ssa->ssa_id;
2100 			break;
2101 		case SCTP_PEER_ADDR_CHANGE:
2102 			sn->sn_paddr_change.spc_assoc_id = ssa->ssa_id;
2103 			break;
2104 		case SCTP_REMOTE_ERROR:
2105 			sn->sn_remote_error.sre_assoc_id = ssa->ssa_id;
2106 			break;
2107 		case SCTP_SEND_FAILED:
2108 			sn->sn_send_failed.ssf_assoc_id = ssa->ssa_id;
2109 			break;
2110 		case SCTP_SHUTDOWN_EVENT:
2111 			sn->sn_shutdown_event.sse_assoc_id = ssa->ssa_id;
2112 			break;
2113 		case SCTP_ADAPTATION_INDICATION:
2114 			sn->sn_adaptation_event.sai_assoc_id = ssa->ssa_id;
2115 			break;
2116 		case SCTP_PARTIAL_DELIVERY_EVENT:
2117 			sn->sn_pdapi_event.pdapi_assoc_id = ssa->ssa_id;
2118 			break;
2119 		default:
2120 			ASSERT(0);
2121 			break;
2122 		}
2123 	} else {
2124 		if (tind->OPT_length > 0) {
2125 			struct cmsghdr	*cmsg;
2126 			char		*cend;
2127 
2128 			cmsg = (struct cmsghdr *)
2129 			    ((uchar_t *)mp->b_rptr + tind->OPT_offset);
2130 			cend = (char *)cmsg + tind->OPT_length;
2131 			for (;;) {
2132 				if ((char *)(cmsg + 1) > cend ||
2133 				    ((char *)cmsg + cmsg->cmsg_len) > cend) {
2134 					break;
2135 				}
2136 				if ((cmsg->cmsg_level == IPPROTO_SCTP) &&
2137 				    (cmsg->cmsg_type == SCTP_SNDRCV)) {
2138 					sinfo = (struct sctp_sndrcvinfo *)
2139 					    (cmsg + 1);
2140 					sinfo->sinfo_assoc_id = ssa->ssa_id;
2141 					break;
2142 				}
2143 				if (cmsg->cmsg_len > 0) {
2144 					cmsg = (struct cmsghdr *)
2145 					    ((uchar_t *)cmsg + cmsg->cmsg_len);
2146 				} else {
2147 					break;
2148 				}
2149 			}
2150 		}
2151 	}
2152 
2153 	/*
2154 	 * SCTP has reserved space in the header for storing a pointer.
2155 	 * Put the pointer to assocation there, and queue the data.
2156 	 */
2157 	SSA_REFHOLD(ssa);
2158 	ASSERT((mp->b_rptr - DB_BASE(mp)) >= sizeof (ssa));
2159 	*(struct sctp_soassoc **)DB_BASE(mp) = ssa;
2160 
2161 	ssa->ssa_rcv_queued += len;
2162 	space_available = so->so_rcvbuf - ssa->ssa_rcv_queued;
2163 	if (space_available <= 0)
2164 		ssa->ssa_flowctrld = B_TRUE;
2165 
2166 	so_enqueue_msg(so, mp, len);
2167 
2168 	/* so_notify_data drops so_lock */
2169 	so_notify_data(so, len);
2170 
2171 	return (space_available);
2172 }
2173 
2174 static void
2175 sctp_assoc_xmitted(sock_upper_handle_t handle, boolean_t qfull)
2176 {
2177 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2178 	struct sctp_sonode *ss = ssa->ssa_sonode;
2179 
2180 	ASSERT(ssa->ssa_type == SOSCTP_ASSOC);
2181 	ASSERT(ss->ss_so.so_type == SOCK_SEQPACKET);
2182 	ASSERT(ssa->ssa_conn != NULL);
2183 
2184 	mutex_enter(&ss->ss_so.so_lock);
2185 
2186 	ssa->ssa_snd_qfull = qfull;
2187 
2188 	/*
2189 	 * Wake blocked writers.
2190 	 */
2191 	cv_broadcast(&ss->ss_so.so_snd_cv);
2192 
2193 	mutex_exit(&ss->ss_so.so_lock);
2194 }
2195 
2196 static vnode_t *
2197 sctp_assoc_get_vnode(sock_upper_handle_t handle)
2198 {
2199 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2200 	struct sonode *so;
2201 
2202 	if (ssa->ssa_type == SOSCTP_ASSOC)
2203 		so = &ssa->ssa_sonode->ss_so;
2204 	else
2205 		so = &((struct sctp_sonode *)handle)->ss_so;
2206 
2207 	return (so_get_vnode((sock_upper_handle_t)so));
2208 }
2209 
2210 static void
2211 sctp_assoc_properties(sock_upper_handle_t handle,
2212     struct sock_proto_props *soppp)
2213 {
2214 	struct sctp_soassoc *ssa = (struct sctp_soassoc *)handle;
2215 	struct sonode *so;
2216 
2217 	if (ssa->ssa_type == SOSCTP_ASSOC) {
2218 		so = &ssa->ssa_sonode->ss_so;
2219 
2220 		mutex_enter(&so->so_lock);
2221 
2222 		/* Per assoc_id properties. */
2223 		if (soppp->sopp_flags & SOCKOPT_WROFF)
2224 			ssa->ssa_wroff = soppp->sopp_wroff;
2225 		if (soppp->sopp_flags & SOCKOPT_MAXBLK)
2226 			ssa->ssa_wrsize = soppp->sopp_maxblk;
2227 	} else {
2228 		so = &((struct sctp_sonode *)handle)->ss_so;
2229 		mutex_enter(&so->so_lock);
2230 
2231 		if (soppp->sopp_flags & SOCKOPT_WROFF)
2232 			so->so_proto_props.sopp_wroff = soppp->sopp_wroff;
2233 		if (soppp->sopp_flags & SOCKOPT_MAXBLK)
2234 			so->so_proto_props.sopp_maxblk = soppp->sopp_maxblk;
2235 		if (soppp->sopp_flags & SOCKOPT_RCVHIWAT) {
2236 			ssize_t lowat;
2237 
2238 			so->so_rcvbuf = soppp->sopp_rxhiwat;
2239 			/*
2240 			 * The low water mark should be adjusted properly
2241 			 * if the high water mark is changed.  It should
2242 			 * not be bigger than 1/4 of high water mark.
2243 			 */
2244 			lowat = soppp->sopp_rxhiwat >> 2;
2245 			if (so->so_rcvlowat > lowat) {
2246 				/* Sanity check... */
2247 				if (lowat == 0)
2248 					so->so_rcvlowat = soppp->sopp_rxhiwat;
2249 				else
2250 					so->so_rcvlowat = lowat;
2251 			}
2252 		}
2253 	}
2254 	mutex_exit(&so->so_lock);
2255 }
2256