xref: /illumos-gate/usr/src/uts/common/rpc/rpcmod.c (revision 4a3b05278938491ea95d557939c130d3eb17cfd4)
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 2010 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /*
26  * Copyright 2012 Milan Jurik. All rights reserved.
27  * Copyright 2012 Marcel Telka <marcel@telka.sk>
28  * Copyright 2013 Nexenta Systems, Inc.  All rights reserved.
29  * Copyright 2018 OmniOS Community Edition (OmniOSce) Association.
30  */
31 /* Copyright (c) 1990 Mentat Inc. */
32 
33 /*	Copyright (c) 1984, 1986, 1987, 1988, 1989 AT&T	*/
34 /*	All Rights Reserved	*/
35 
36 /*
37  * Kernel RPC filtering module
38  */
39 
40 #include <sys/param.h>
41 #include <sys/types.h>
42 #include <sys/stream.h>
43 #include <sys/stropts.h>
44 #include <sys/strsubr.h>
45 #include <sys/tihdr.h>
46 #include <sys/timod.h>
47 #include <sys/tiuser.h>
48 #include <sys/debug.h>
49 #include <sys/signal.h>
50 #include <sys/pcb.h>
51 #include <sys/user.h>
52 #include <sys/errno.h>
53 #include <sys/cred.h>
54 #include <sys/policy.h>
55 #include <sys/inline.h>
56 #include <sys/cmn_err.h>
57 #include <sys/kmem.h>
58 #include <sys/file.h>
59 #include <sys/sysmacros.h>
60 #include <sys/systm.h>
61 #include <sys/t_lock.h>
62 #include <sys/ddi.h>
63 #include <sys/vtrace.h>
64 #include <sys/callb.h>
65 #include <sys/strsun.h>
66 
67 #include <sys/strlog.h>
68 #include <rpc/rpc_com.h>
69 #include <inet/common.h>
70 #include <rpc/types.h>
71 #include <sys/time.h>
72 #include <rpc/xdr.h>
73 #include <rpc/auth.h>
74 #include <rpc/clnt.h>
75 #include <rpc/rpc_msg.h>
76 #include <rpc/clnt.h>
77 #include <rpc/svc.h>
78 #include <rpc/rpcsys.h>
79 #include <rpc/rpc_rdma.h>
80 
81 /*
82  * This is the loadable module wrapper.
83  */
84 #include <sys/conf.h>
85 #include <sys/modctl.h>
86 #include <sys/syscall.h>
87 
88 extern struct streamtab rpcinfo;
89 
90 static struct fmodsw fsw = {
91 	"rpcmod",
92 	&rpcinfo,
93 	D_NEW|D_MP,
94 };
95 
96 /*
97  * Module linkage information for the kernel.
98  */
99 
100 static struct modlstrmod modlstrmod = {
101 	&mod_strmodops, "rpc interface str mod", &fsw
102 };
103 
104 /*
105  * For the RPC system call.
106  */
107 static struct sysent rpcsysent = {
108 	2,
109 	SE_32RVAL1 | SE_ARGC | SE_NOUNLOAD,
110 	rpcsys
111 };
112 
113 static struct modlsys modlsys = {
114 	&mod_syscallops,
115 	"RPC syscall",
116 	&rpcsysent
117 };
118 
119 #ifdef _SYSCALL32_IMPL
120 static struct modlsys modlsys32 = {
121 	&mod_syscallops32,
122 	"32-bit RPC syscall",
123 	&rpcsysent
124 };
125 #endif /* _SYSCALL32_IMPL */
126 
127 static struct modlinkage modlinkage = {
128 	MODREV_1,
129 	{
130 		&modlsys,
131 #ifdef _SYSCALL32_IMPL
132 		&modlsys32,
133 #endif
134 		&modlstrmod,
135 		NULL
136 	}
137 };
138 
139 int
140 _init(void)
141 {
142 	int error = 0;
143 	callb_id_t cid;
144 	int status;
145 
146 	svc_init();
147 	clnt_init();
148 	cid = callb_add(connmgr_cpr_reset, 0, CB_CL_CPR_RPC, "rpc");
149 
150 	if (error = mod_install(&modlinkage)) {
151 		/*
152 		 * Could not install module, cleanup previous
153 		 * initialization work.
154 		 */
155 		clnt_fini();
156 		if (cid != NULL)
157 			(void) callb_delete(cid);
158 
159 		return (error);
160 	}
161 
162 	/*
163 	 * Load up the RDMA plugins and initialize the stats. Even if the
164 	 * plugins loadup fails, but rpcmod was successfully installed the
165 	 * counters still get initialized.
166 	 */
167 	rw_init(&rdma_lock, NULL, RW_DEFAULT, NULL);
168 	mutex_init(&rdma_modload_lock, NULL, MUTEX_DEFAULT, NULL);
169 
170 	cv_init(&rdma_wait.svc_cv, NULL, CV_DEFAULT, NULL);
171 	mutex_init(&rdma_wait.svc_lock, NULL, MUTEX_DEFAULT, NULL);
172 
173 	mt_kstat_init();
174 
175 	/*
176 	 * Get our identification into ldi.  This is used for loading
177 	 * other modules, e.g. rpcib.
178 	 */
179 	status = ldi_ident_from_mod(&modlinkage, &rpcmod_li);
180 	if (status != 0) {
181 		cmn_err(CE_WARN, "ldi_ident_from_mod fails with %d", status);
182 		rpcmod_li = NULL;
183 	}
184 
185 	return (error);
186 }
187 
188 /*
189  * The unload entry point fails, because we advertise entry points into
190  * rpcmod from the rest of kRPC: rpcmod_release().
191  */
192 int
193 _fini(void)
194 {
195 	return (EBUSY);
196 }
197 
198 int
199 _info(struct modinfo *modinfop)
200 {
201 	return (mod_info(&modlinkage, modinfop));
202 }
203 
204 extern int nulldev();
205 
206 #define	RPCMOD_ID	2049
207 
208 int rmm_open(queue_t *, dev_t *, int, int, cred_t *);
209 int rmm_close(queue_t *, int, cred_t *);
210 
211 /*
212  * To save instructions, since STREAMS ignores the return value
213  * from these functions, they are defined as void here. Kind of icky, but...
214  */
215 void rmm_rput(queue_t *, mblk_t *);
216 void rmm_wput(queue_t *, mblk_t *);
217 void rmm_rsrv(queue_t *);
218 void rmm_wsrv(queue_t *);
219 
220 int rpcmodopen(queue_t *, dev_t *, int, int, cred_t *);
221 int rpcmodclose(queue_t *, int, cred_t *);
222 void rpcmodrput(queue_t *, mblk_t *);
223 void rpcmodwput(queue_t *, mblk_t *);
224 void rpcmodrsrv();
225 void rpcmodwsrv(queue_t *);
226 
227 static	void	rpcmodwput_other(queue_t *, mblk_t *);
228 static	int	mir_close(queue_t *q);
229 static	int	mir_open(queue_t *q, dev_t *devp, int flag, int sflag,
230 		    cred_t *credp);
231 static	void	mir_rput(queue_t *q, mblk_t *mp);
232 static	void	mir_rsrv(queue_t *q);
233 static	void	mir_wput(queue_t *q, mblk_t *mp);
234 static	void	mir_wsrv(queue_t *q);
235 
236 static struct module_info rpcmod_info =
237 	{RPCMOD_ID, "rpcmod", 0, INFPSZ, 256*1024, 1024};
238 
239 static struct qinit rpcmodrinit = {
240 	(int (*)())rmm_rput,
241 	(int (*)())rmm_rsrv,
242 	rmm_open,
243 	rmm_close,
244 	nulldev,
245 	&rpcmod_info,
246 	NULL
247 };
248 
249 /*
250  * The write put procedure is simply putnext to conserve stack space.
251  * The write service procedure is not used to queue data, but instead to
252  * synchronize with flow control.
253  */
254 static struct qinit rpcmodwinit = {
255 	(int (*)())rmm_wput,
256 	(int (*)())rmm_wsrv,
257 	rmm_open,
258 	rmm_close,
259 	nulldev,
260 	&rpcmod_info,
261 	NULL
262 };
263 struct streamtab rpcinfo = { &rpcmodrinit, &rpcmodwinit, NULL, NULL };
264 
265 struct xprt_style_ops {
266 	int (*xo_open)();
267 	int (*xo_close)();
268 	void (*xo_wput)();
269 	void (*xo_wsrv)();
270 	void (*xo_rput)();
271 	void (*xo_rsrv)();
272 };
273 
274 /*
275  * Read side has no service procedure.
276  */
277 static struct xprt_style_ops xprt_clts_ops = {
278 	rpcmodopen,
279 	rpcmodclose,
280 	rpcmodwput,
281 	rpcmodwsrv,
282 	rpcmodrput,
283 	NULL
284 };
285 
286 static struct xprt_style_ops xprt_cots_ops = {
287 	mir_open,
288 	mir_close,
289 	mir_wput,
290 	mir_wsrv,
291 	mir_rput,
292 	mir_rsrv
293 };
294 
295 /*
296  * Per rpcmod "slot" data structure. q->q_ptr points to one of these.
297  */
298 struct rpcm {
299 	void		*rm_krpc_cell;	/* Reserved for use by kRPC */
300 	struct		xprt_style_ops	*rm_ops;
301 	int		rm_type;	/* Client or server side stream */
302 #define	RM_CLOSING	0x1		/* somebody is trying to close slot */
303 	uint_t		rm_state;	/* state of the slot. see above */
304 	uint_t		rm_ref;		/* cnt of external references to slot */
305 	kmutex_t	rm_lock;	/* mutex protecting above fields */
306 	kcondvar_t	rm_cwait;	/* condition for closing */
307 	zoneid_t	rm_zoneid;	/* zone which pushed rpcmod */
308 };
309 
310 struct temp_slot {
311 	void *cell;
312 	struct xprt_style_ops *ops;
313 	int type;
314 	mblk_t *info_ack;
315 	kmutex_t lock;
316 	kcondvar_t wait;
317 };
318 
319 typedef struct mir_s {
320 	void	*mir_krpc_cell;	/* Reserved for kRPC use. This field */
321 					/* must be first in the structure. */
322 	struct xprt_style_ops	*rm_ops;
323 	int	mir_type;		/* Client or server side stream */
324 
325 	mblk_t	*mir_head_mp;		/* RPC msg in progress */
326 		/*
327 		 * mir_head_mp points the first mblk being collected in
328 		 * the current RPC message.  Record headers are removed
329 		 * before data is linked into mir_head_mp.
330 		 */
331 	mblk_t	*mir_tail_mp;		/* Last mblk in mir_head_mp */
332 		/*
333 		 * mir_tail_mp points to the last mblk in the message
334 		 * chain starting at mir_head_mp.  It is only valid
335 		 * if mir_head_mp is non-NULL and is used to add new
336 		 * data blocks to the end of chain quickly.
337 		 */
338 
339 	int32_t	mir_frag_len;		/* Bytes seen in the current frag */
340 		/*
341 		 * mir_frag_len starts at -4 for beginning of each fragment.
342 		 * When this length is negative, it indicates the number of
343 		 * bytes that rpcmod needs to complete the record marker
344 		 * header.  When it is positive or zero, it holds the number
345 		 * of bytes that have arrived for the current fragment and
346 		 * are held in mir_header_mp.
347 		 */
348 
349 	int32_t	mir_frag_header;
350 		/*
351 		 * Fragment header as collected for the current fragment.
352 		 * It holds the last-fragment indicator and the number
353 		 * of bytes in the fragment.
354 		 */
355 
356 	unsigned int
357 		mir_ordrel_pending : 1,	/* Sent T_ORDREL_REQ */
358 		mir_hold_inbound : 1,	/* Hold inbound messages on server */
359 					/* side until outbound flow control */
360 					/* is relieved. */
361 		mir_closing : 1,	/* The stream is being closed */
362 		mir_inrservice : 1,	/* data queued or rd srv proc running */
363 		mir_inwservice : 1,	/* data queued or wr srv proc running */
364 		mir_inwflushdata : 1,	/* flush M_DATAs when srv runs */
365 		/*
366 		 * On client streams, mir_clntreq is 0 or 1; it is set
367 		 * to 1 whenever a new request is sent out (mir_wput)
368 		 * and cleared when the timer fires (mir_timer).  If
369 		 * the timer fires with this value equal to 0, then the
370 		 * stream is considered idle and kRPC is notified.
371 		 */
372 		mir_clntreq : 1,
373 		/*
374 		 * On server streams, stop accepting messages
375 		 */
376 		mir_svc_no_more_msgs : 1,
377 		mir_listen_stream : 1,	/* listen end point */
378 		mir_unused : 1,	/* no longer used */
379 		mir_timer_call : 1,
380 		mir_junk_fill_thru_bit_31 : 21;
381 
382 	int	mir_setup_complete;	/* server has initialized everything */
383 	timeout_id_t mir_timer_id;	/* Timer for idle checks */
384 	clock_t	mir_idle_timeout;	/* Allowed idle time before shutdown */
385 		/*
386 		 * This value is copied from clnt_idle_timeout or
387 		 * svc_idle_timeout during the appropriate ioctl.
388 		 * Kept in milliseconds
389 		 */
390 	clock_t	mir_use_timestamp;	/* updated on client with each use */
391 		/*
392 		 * This value is set to lbolt
393 		 * every time a client stream sends or receives data.
394 		 * Even if the timer message arrives, we don't shutdown
395 		 * client unless:
396 		 *    lbolt >= MSEC_TO_TICK(mir_idle_timeout)+mir_use_timestamp.
397 		 * This value is kept in HZ.
398 		 */
399 
400 	uint_t	*mir_max_msg_sizep;	/* Reference to sanity check size */
401 		/*
402 		 * This pointer is set to &clnt_max_msg_size or
403 		 * &svc_max_msg_size during the appropriate ioctl.
404 		 */
405 	zoneid_t mir_zoneid;	/* zone which pushed rpcmod */
406 	/* Server-side fields. */
407 	int	mir_ref_cnt;		/* Reference count: server side only */
408 					/* counts the number of references */
409 					/* that a kernel RPC server thread */
410 					/* (see svc_run()) has on this rpcmod */
411 					/* slot. Effectively, it is the */
412 					/* number of unprocessed messages */
413 					/* that have been passed up to the */
414 					/* kRPC layer */
415 
416 	mblk_t	*mir_svc_pend_mp;	/* Pending T_ORDREL_IND or */
417 					/* T_DISCON_IND */
418 
419 	/*
420 	 * these fields are for both client and server, but for debugging,
421 	 * it is easier to have these last in the structure.
422 	 */
423 	kmutex_t	mir_mutex;	/* Mutex and condvar for close */
424 	kcondvar_t	mir_condvar;	/* synchronization. */
425 	kcondvar_t	mir_timer_cv;	/* Timer routine sync. */
426 } mir_t;
427 
428 void tmp_rput(queue_t *q, mblk_t *mp);
429 
430 struct xprt_style_ops tmpops = {
431 	NULL,
432 	NULL,
433 	putnext,
434 	NULL,
435 	tmp_rput,
436 	NULL
437 };
438 
439 void
440 tmp_rput(queue_t *q, mblk_t *mp)
441 {
442 	struct temp_slot *t = (struct temp_slot *)(q->q_ptr);
443 	struct T_info_ack *pptr;
444 
445 	switch (mp->b_datap->db_type) {
446 	case M_PCPROTO:
447 		pptr = (struct T_info_ack *)mp->b_rptr;
448 		switch (pptr->PRIM_type) {
449 		case T_INFO_ACK:
450 			mutex_enter(&t->lock);
451 			t->info_ack = mp;
452 			cv_signal(&t->wait);
453 			mutex_exit(&t->lock);
454 			return;
455 		default:
456 			break;
457 		}
458 	default:
459 		break;
460 	}
461 
462 	/*
463 	 * Not an info-ack, so free it. This is ok because we should
464 	 * not be receiving data until the open finishes: rpcmod
465 	 * is pushed well before the end-point is bound to an address.
466 	 */
467 	freemsg(mp);
468 }
469 
470 int
471 rmm_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *crp)
472 {
473 	mblk_t *bp;
474 	struct temp_slot ts, *t;
475 	struct T_info_ack *pptr;
476 	int error = 0;
477 
478 	ASSERT(q != NULL);
479 	/*
480 	 * Check for re-opens.
481 	 */
482 	if (q->q_ptr) {
483 		TRACE_1(TR_FAC_KRPC, TR_RPCMODOPEN_END,
484 		    "rpcmodopen_end:(%s)", "q->qptr");
485 		return (0);
486 	}
487 
488 	t = &ts;
489 	bzero(t, sizeof (*t));
490 	q->q_ptr = (void *)t;
491 	WR(q)->q_ptr = (void *)t;
492 
493 	/*
494 	 * Allocate the required messages upfront.
495 	 */
496 	if ((bp = allocb_cred(sizeof (struct T_info_req) +
497 	    sizeof (struct T_info_ack), crp, curproc->p_pid)) == NULL) {
498 		return (ENOBUFS);
499 	}
500 
501 	mutex_init(&t->lock, NULL, MUTEX_DEFAULT, NULL);
502 	cv_init(&t->wait, NULL, CV_DEFAULT, NULL);
503 
504 	t->ops = &tmpops;
505 
506 	qprocson(q);
507 	bp->b_datap->db_type = M_PCPROTO;
508 	*(int32_t *)bp->b_wptr = (int32_t)T_INFO_REQ;
509 	bp->b_wptr += sizeof (struct T_info_req);
510 	putnext(WR(q), bp);
511 
512 	mutex_enter(&t->lock);
513 	while (t->info_ack == NULL) {
514 		if (cv_wait_sig(&t->wait, &t->lock) == 0) {
515 			error = EINTR;
516 			break;
517 		}
518 	}
519 	mutex_exit(&t->lock);
520 
521 	if (error)
522 		goto out;
523 
524 	pptr = (struct T_info_ack *)t->info_ack->b_rptr;
525 
526 	if (pptr->SERV_type == T_CLTS) {
527 		if ((error = rpcmodopen(q, devp, flag, sflag, crp)) == 0)
528 			((struct rpcm *)q->q_ptr)->rm_ops = &xprt_clts_ops;
529 	} else {
530 		if ((error = mir_open(q, devp, flag, sflag, crp)) == 0)
531 			((mir_t *)q->q_ptr)->rm_ops = &xprt_cots_ops;
532 	}
533 
534 out:
535 	if (error)
536 		qprocsoff(q);
537 
538 	freemsg(t->info_ack);
539 	mutex_destroy(&t->lock);
540 	cv_destroy(&t->wait);
541 
542 	return (error);
543 }
544 
545 void
546 rmm_rput(queue_t *q, mblk_t  *mp)
547 {
548 	(*((struct temp_slot *)q->q_ptr)->ops->xo_rput)(q, mp);
549 }
550 
551 void
552 rmm_rsrv(queue_t *q)
553 {
554 	(*((struct temp_slot *)q->q_ptr)->ops->xo_rsrv)(q);
555 }
556 
557 void
558 rmm_wput(queue_t *q, mblk_t *mp)
559 {
560 	(*((struct temp_slot *)q->q_ptr)->ops->xo_wput)(q, mp);
561 }
562 
563 void
564 rmm_wsrv(queue_t *q)
565 {
566 	(*((struct temp_slot *)q->q_ptr)->ops->xo_wsrv)(q);
567 }
568 
569 int
570 rmm_close(queue_t *q, int flag, cred_t *crp)
571 {
572 	return ((*((struct temp_slot *)q->q_ptr)->ops->xo_close)(q, flag, crp));
573 }
574 
575 /*
576  * rpcmodopen -	open routine gets called when the module gets pushed
577  *		onto the stream.
578  */
579 /*ARGSUSED*/
580 int
581 rpcmodopen(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *crp)
582 {
583 	struct rpcm *rmp;
584 
585 	TRACE_0(TR_FAC_KRPC, TR_RPCMODOPEN_START, "rpcmodopen_start:");
586 
587 	/*
588 	 * Only sufficiently privileged users can use this module, and it
589 	 * is assumed that they will use this module properly, and NOT send
590 	 * bulk data from downstream.
591 	 */
592 	if (secpolicy_rpcmod_open(crp) != 0)
593 		return (EPERM);
594 
595 	/*
596 	 * Allocate slot data structure.
597 	 */
598 	rmp = kmem_zalloc(sizeof (*rmp), KM_SLEEP);
599 
600 	mutex_init(&rmp->rm_lock, NULL, MUTEX_DEFAULT, NULL);
601 	cv_init(&rmp->rm_cwait, NULL, CV_DEFAULT, NULL);
602 	rmp->rm_zoneid = rpc_zoneid();
603 	/*
604 	 * slot type will be set by kRPC client and server ioctl's
605 	 */
606 	rmp->rm_type = 0;
607 
608 	q->q_ptr = (void *)rmp;
609 	WR(q)->q_ptr = (void *)rmp;
610 
611 	TRACE_1(TR_FAC_KRPC, TR_RPCMODOPEN_END, "rpcmodopen_end:(%s)", "end");
612 	return (0);
613 }
614 
615 /*
616  * rpcmodclose - This routine gets called when the module gets popped
617  * off of the stream.
618  */
619 /*ARGSUSED*/
620 int
621 rpcmodclose(queue_t *q, int flag, cred_t *crp)
622 {
623 	struct rpcm *rmp;
624 
625 	ASSERT(q != NULL);
626 	rmp = (struct rpcm *)q->q_ptr;
627 
628 	/*
629 	 * Mark our state as closing.
630 	 */
631 	mutex_enter(&rmp->rm_lock);
632 	rmp->rm_state |= RM_CLOSING;
633 
634 	/*
635 	 * Check and see if there are any messages on the queue.  If so, send
636 	 * the messages, regardless whether the downstream module is ready to
637 	 * accept data.
638 	 */
639 	if (rmp->rm_type == RPC_SERVER) {
640 		flushq(q, FLUSHDATA);
641 
642 		qenable(WR(q));
643 
644 		if (rmp->rm_ref) {
645 			mutex_exit(&rmp->rm_lock);
646 			/*
647 			 * call into SVC to clean the queue
648 			 */
649 			svc_queueclean(q);
650 			mutex_enter(&rmp->rm_lock);
651 
652 			/*
653 			 * Block while there are kRPC threads with a reference
654 			 * to this message.
655 			 */
656 			while (rmp->rm_ref)
657 				cv_wait(&rmp->rm_cwait, &rmp->rm_lock);
658 		}
659 
660 		mutex_exit(&rmp->rm_lock);
661 
662 		/*
663 		 * It is now safe to remove this queue from the stream. No kRPC
664 		 * threads have a reference to the stream, and none ever will,
665 		 * because RM_CLOSING is set.
666 		 */
667 		qprocsoff(q);
668 
669 		/* Notify kRPC that this stream is going away. */
670 		svc_queueclose(q);
671 	} else {
672 		mutex_exit(&rmp->rm_lock);
673 		qprocsoff(q);
674 	}
675 
676 	q->q_ptr = NULL;
677 	WR(q)->q_ptr = NULL;
678 	mutex_destroy(&rmp->rm_lock);
679 	cv_destroy(&rmp->rm_cwait);
680 	kmem_free(rmp, sizeof (*rmp));
681 	return (0);
682 }
683 
684 /*
685  * rpcmodrput -	Module read put procedure.  This is called from
686  *		the module, driver, or stream head downstream.
687  */
688 void
689 rpcmodrput(queue_t *q, mblk_t *mp)
690 {
691 	struct rpcm *rmp;
692 	union T_primitives *pptr;
693 	int hdrsz;
694 
695 	TRACE_0(TR_FAC_KRPC, TR_RPCMODRPUT_START, "rpcmodrput_start:");
696 
697 	ASSERT(q != NULL);
698 	rmp = (struct rpcm *)q->q_ptr;
699 
700 	if (rmp->rm_type == 0) {
701 		freemsg(mp);
702 		return;
703 	}
704 
705 	switch (mp->b_datap->db_type) {
706 	default:
707 		putnext(q, mp);
708 		break;
709 
710 	case M_PROTO:
711 	case M_PCPROTO:
712 		ASSERT((mp->b_wptr - mp->b_rptr) >= sizeof (int32_t));
713 		pptr = (union T_primitives *)mp->b_rptr;
714 
715 		/*
716 		 * Forward this message to kRPC if it is data.
717 		 */
718 		if (pptr->type == T_UNITDATA_IND) {
719 			/*
720 			 * Check if the module is being popped.
721 			 */
722 			mutex_enter(&rmp->rm_lock);
723 			if (rmp->rm_state & RM_CLOSING) {
724 				mutex_exit(&rmp->rm_lock);
725 				putnext(q, mp);
726 				break;
727 			}
728 
729 			switch (rmp->rm_type) {
730 			case RPC_CLIENT:
731 				mutex_exit(&rmp->rm_lock);
732 				hdrsz = mp->b_wptr - mp->b_rptr;
733 
734 				/*
735 				 * Make sure the header is sane.
736 				 */
737 				if (hdrsz < TUNITDATAINDSZ ||
738 				    hdrsz < (pptr->unitdata_ind.OPT_length +
739 				    pptr->unitdata_ind.OPT_offset) ||
740 				    hdrsz < (pptr->unitdata_ind.SRC_length +
741 				    pptr->unitdata_ind.SRC_offset)) {
742 					freemsg(mp);
743 					return;
744 				}
745 
746 				/*
747 				 * Call clnt_clts_dispatch_notify, so that it
748 				 * can pass the message to the proper caller.
749 				 * Don't discard the header just yet since the
750 				 * client may need the sender's address.
751 				 */
752 				clnt_clts_dispatch_notify(mp, hdrsz,
753 				    rmp->rm_zoneid);
754 				return;
755 			case RPC_SERVER:
756 				/*
757 				 * rm_krpc_cell is exclusively used by the kRPC
758 				 * CLTS server. Try to submit the message to
759 				 * kRPC. Since this is an unreliable channel, we
760 				 * can just free the message in case the kRPC
761 				 * does not accept new messages.
762 				 */
763 				if (rmp->rm_krpc_cell &&
764 				    svc_queuereq(q, mp, TRUE)) {
765 					/*
766 					 * Raise the reference count on this
767 					 * module to prevent it from being
768 					 * popped before kRPC generates the
769 					 * reply.
770 					 */
771 					rmp->rm_ref++;
772 					mutex_exit(&rmp->rm_lock);
773 				} else {
774 					mutex_exit(&rmp->rm_lock);
775 					freemsg(mp);
776 				}
777 				return;
778 			default:
779 				mutex_exit(&rmp->rm_lock);
780 				freemsg(mp);
781 				return;
782 			} /* end switch(rmp->rm_type) */
783 		} else if (pptr->type == T_UDERROR_IND) {
784 			mutex_enter(&rmp->rm_lock);
785 			hdrsz = mp->b_wptr - mp->b_rptr;
786 
787 			/*
788 			 * Make sure the header is sane
789 			 */
790 			if (hdrsz < TUDERRORINDSZ ||
791 			    hdrsz < (pptr->uderror_ind.OPT_length +
792 			    pptr->uderror_ind.OPT_offset) ||
793 			    hdrsz < (pptr->uderror_ind.DEST_length +
794 			    pptr->uderror_ind.DEST_offset)) {
795 				mutex_exit(&rmp->rm_lock);
796 				freemsg(mp);
797 				return;
798 			}
799 
800 			/*
801 			 * In the case where a unit data error has been
802 			 * received, all we need to do is clear the message from
803 			 * the queue.
804 			 */
805 			mutex_exit(&rmp->rm_lock);
806 			freemsg(mp);
807 			RPCLOG(32, "rpcmodrput: unitdata error received at "
808 			    "%ld\n", gethrestime_sec());
809 			return;
810 		} /* end else if (pptr->type == T_UDERROR_IND) */
811 
812 		putnext(q, mp);
813 		break;
814 	} /* end switch (mp->b_datap->db_type) */
815 
816 	TRACE_0(TR_FAC_KRPC, TR_RPCMODRPUT_END,
817 	    "rpcmodrput_end:");
818 	/*
819 	 * Return codes are not looked at by the STREAMS framework.
820 	 */
821 }
822 
823 /*
824  * write put procedure
825  */
826 void
827 rpcmodwput(queue_t *q, mblk_t *mp)
828 {
829 	struct rpcm	*rmp;
830 
831 	ASSERT(q != NULL);
832 
833 	switch (mp->b_datap->db_type) {
834 		case M_PROTO:
835 		case M_PCPROTO:
836 			break;
837 		default:
838 			rpcmodwput_other(q, mp);
839 			return;
840 	}
841 
842 	/*
843 	 * Check to see if we can send the message downstream.
844 	 */
845 	if (canputnext(q)) {
846 		putnext(q, mp);
847 		return;
848 	}
849 
850 	rmp = (struct rpcm *)q->q_ptr;
851 	ASSERT(rmp != NULL);
852 
853 	/*
854 	 * The first canputnext failed.  Try again except this time with the
855 	 * lock held, so that we can check the state of the stream to see if
856 	 * it is closing.  If either of these conditions evaluate to true
857 	 * then send the meesage.
858 	 */
859 	mutex_enter(&rmp->rm_lock);
860 	if (canputnext(q) || (rmp->rm_state & RM_CLOSING)) {
861 		mutex_exit(&rmp->rm_lock);
862 		putnext(q, mp);
863 	} else {
864 		/*
865 		 * canputnext failed again and the stream is not closing.
866 		 * Place the message on the queue and let the service
867 		 * procedure handle the message.
868 		 */
869 		mutex_exit(&rmp->rm_lock);
870 		(void) putq(q, mp);
871 	}
872 }
873 
874 static void
875 rpcmodwput_other(queue_t *q, mblk_t *mp)
876 {
877 	struct rpcm	*rmp;
878 	struct iocblk	*iocp;
879 
880 	rmp = (struct rpcm *)q->q_ptr;
881 	ASSERT(rmp != NULL);
882 
883 	switch (mp->b_datap->db_type) {
884 		case M_IOCTL:
885 			iocp = (struct iocblk *)mp->b_rptr;
886 			ASSERT(iocp != NULL);
887 			switch (iocp->ioc_cmd) {
888 				case RPC_CLIENT:
889 				case RPC_SERVER:
890 					mutex_enter(&rmp->rm_lock);
891 					rmp->rm_type = iocp->ioc_cmd;
892 					mutex_exit(&rmp->rm_lock);
893 					mp->b_datap->db_type = M_IOCACK;
894 					qreply(q, mp);
895 					return;
896 				default:
897 				/*
898 				 * pass the ioctl downstream and hope someone
899 				 * down there knows how to handle it.
900 				 */
901 					putnext(q, mp);
902 					return;
903 			}
904 		default:
905 			break;
906 	}
907 	/*
908 	 * This is something we definitely do not know how to handle, just
909 	 * pass the message downstream
910 	 */
911 	putnext(q, mp);
912 }
913 
914 /*
915  * Module write service procedure. This is called by downstream modules
916  * for back enabling during flow control.
917  */
918 void
919 rpcmodwsrv(queue_t *q)
920 {
921 	struct rpcm	*rmp;
922 	mblk_t		*mp = NULL;
923 
924 	rmp = (struct rpcm *)q->q_ptr;
925 	ASSERT(rmp != NULL);
926 
927 	/*
928 	 * Get messages that may be queued and send them down stream
929 	 */
930 	while ((mp = getq(q)) != NULL) {
931 		/*
932 		 * Optimize the service procedure for the server-side, by
933 		 * avoiding a call to canputnext().
934 		 */
935 		if (rmp->rm_type == RPC_SERVER || canputnext(q)) {
936 			putnext(q, mp);
937 			continue;
938 		}
939 		(void) putbq(q, mp);
940 		return;
941 	}
942 }
943 
944 void
945 rpcmod_hold(queue_t *q)
946 {
947 	struct rpcm *rmp = (struct rpcm *)q->q_ptr;
948 
949 	mutex_enter(&rmp->rm_lock);
950 	rmp->rm_ref++;
951 	mutex_exit(&rmp->rm_lock);
952 }
953 
954 void
955 rpcmod_release(queue_t *q, mblk_t *bp,
956     /* LINTED E_FUNC_ARG_UNUSED */
957     bool_t enable __unused)
958 {
959 	struct rpcm *rmp;
960 
961 	/*
962 	 * For now, just free the message.
963 	 */
964 	if (bp)
965 		freemsg(bp);
966 	rmp = (struct rpcm *)q->q_ptr;
967 
968 	mutex_enter(&rmp->rm_lock);
969 	rmp->rm_ref--;
970 
971 	if (rmp->rm_ref == 0 && (rmp->rm_state & RM_CLOSING)) {
972 		cv_broadcast(&rmp->rm_cwait);
973 	}
974 
975 	mutex_exit(&rmp->rm_lock);
976 }
977 
978 /*
979  * This part of rpcmod is pushed on a connection-oriented transport for use
980  * by RPC.  It serves to bypass the Stream head, implements
981  * the record marking protocol, and dispatches incoming RPC messages.
982  */
983 
984 /* Default idle timer values */
985 #define	MIR_CLNT_IDLE_TIMEOUT	(5 * (60 * 1000L))	/* 5 minutes */
986 #define	MIR_SVC_IDLE_TIMEOUT	(6 * (60 * 1000L))	/* 6 minutes */
987 #define	MIR_SVC_ORDREL_TIMEOUT	(10 * (60 * 1000L))	/* 10 minutes */
988 #define	MIR_LASTFRAG	0x80000000	/* Record marker */
989 
990 #define	MIR_SVC_QUIESCED(mir)	\
991 	(mir->mir_ref_cnt == 0 && mir->mir_inrservice == 0)
992 
993 #define	MIR_CLEAR_INRSRV(mir_ptr)	{	\
994 	(mir_ptr)->mir_inrservice = 0;	\
995 	if ((mir_ptr)->mir_type == RPC_SERVER &&	\
996 		(mir_ptr)->mir_closing)	\
997 		cv_signal(&(mir_ptr)->mir_condvar);	\
998 }
999 
1000 /*
1001  * Don't block service procedure (and mir_close) if
1002  * we are in the process of closing.
1003  */
1004 #define	MIR_WCANPUTNEXT(mir_ptr, write_q)	\
1005 	(canputnext(write_q) || ((mir_ptr)->mir_svc_no_more_msgs == 1))
1006 
1007 static int	mir_clnt_dup_request(queue_t *q, mblk_t *mp);
1008 static void	mir_rput_proto(queue_t *q, mblk_t *mp);
1009 static int	mir_svc_policy_notify(queue_t *q, int event);
1010 static void	mir_svc_start(queue_t *wq);
1011 static void	mir_svc_idle_start(queue_t *, mir_t *);
1012 static void	mir_svc_idle_stop(queue_t *, mir_t *);
1013 static void	mir_svc_start_close(queue_t *, mir_t *);
1014 static void	mir_clnt_idle_do_stop(queue_t *);
1015 static void	mir_clnt_idle_stop(queue_t *, mir_t *);
1016 static void	mir_clnt_idle_start(queue_t *, mir_t *);
1017 static void	mir_wput(queue_t *q, mblk_t *mp);
1018 static void	mir_wput_other(queue_t *q, mblk_t *mp);
1019 static void	mir_wsrv(queue_t *q);
1020 static	void	mir_disconnect(queue_t *, mir_t *ir);
1021 static	int	mir_check_len(queue_t *, mblk_t *);
1022 static	void	mir_timer(void *);
1023 
1024 extern void	(*mir_start)(queue_t *);
1025 extern void	(*clnt_stop_idle)(queue_t *);
1026 
1027 clock_t	clnt_idle_timeout = MIR_CLNT_IDLE_TIMEOUT;
1028 clock_t	svc_idle_timeout = MIR_SVC_IDLE_TIMEOUT;
1029 
1030 /*
1031  * Timeout for subsequent notifications of idle connection.  This is
1032  * typically used to clean up after a wedged orderly release.
1033  */
1034 clock_t	svc_ordrel_timeout = MIR_SVC_ORDREL_TIMEOUT; /* milliseconds */
1035 
1036 extern	uint_t	*clnt_max_msg_sizep;
1037 extern	uint_t	*svc_max_msg_sizep;
1038 uint_t	clnt_max_msg_size = RPC_MAXDATASIZE;
1039 uint_t	svc_max_msg_size = RPC_MAXDATASIZE;
1040 uint_t	mir_krpc_cell_null;
1041 
1042 static void
1043 mir_timer_stop(mir_t *mir)
1044 {
1045 	timeout_id_t tid;
1046 
1047 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
1048 
1049 	/*
1050 	 * Since the mir_mutex lock needs to be released to call
1051 	 * untimeout(), we need to make sure that no other thread
1052 	 * can start/stop the timer (changing mir_timer_id) during
1053 	 * that time.  The mir_timer_call bit and the mir_timer_cv
1054 	 * condition variable are used to synchronize this.  Setting
1055 	 * mir_timer_call also tells mir_timer() (refer to the comments
1056 	 * in mir_timer()) that it does not need to do anything.
1057 	 */
1058 	while (mir->mir_timer_call)
1059 		cv_wait(&mir->mir_timer_cv, &mir->mir_mutex);
1060 	mir->mir_timer_call = B_TRUE;
1061 
1062 	if ((tid = mir->mir_timer_id) != 0) {
1063 		mir->mir_timer_id = 0;
1064 		mutex_exit(&mir->mir_mutex);
1065 		(void) untimeout(tid);
1066 		mutex_enter(&mir->mir_mutex);
1067 	}
1068 	mir->mir_timer_call = B_FALSE;
1069 	cv_broadcast(&mir->mir_timer_cv);
1070 }
1071 
1072 static void
1073 mir_timer_start(queue_t *q, mir_t *mir, clock_t intrvl)
1074 {
1075 	timeout_id_t tid;
1076 
1077 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
1078 
1079 	while (mir->mir_timer_call)
1080 		cv_wait(&mir->mir_timer_cv, &mir->mir_mutex);
1081 	mir->mir_timer_call = B_TRUE;
1082 
1083 	if ((tid = mir->mir_timer_id) != 0) {
1084 		mutex_exit(&mir->mir_mutex);
1085 		(void) untimeout(tid);
1086 		mutex_enter(&mir->mir_mutex);
1087 	}
1088 	/* Only start the timer when it is not closing. */
1089 	if (!mir->mir_closing) {
1090 		mir->mir_timer_id = timeout(mir_timer, q,
1091 		    MSEC_TO_TICK(intrvl));
1092 	}
1093 	mir->mir_timer_call = B_FALSE;
1094 	cv_broadcast(&mir->mir_timer_cv);
1095 }
1096 
1097 static int
1098 mir_clnt_dup_request(queue_t *q, mblk_t *mp)
1099 {
1100 	mblk_t  *mp1;
1101 	uint32_t  new_xid;
1102 	uint32_t  old_xid;
1103 
1104 	ASSERT(MUTEX_HELD(&((mir_t *)q->q_ptr)->mir_mutex));
1105 	new_xid = BE32_TO_U32(&mp->b_rptr[4]);
1106 	/*
1107 	 * This loop is a bit tacky -- it walks the STREAMS list of
1108 	 * flow-controlled messages.
1109 	 */
1110 	if ((mp1 = q->q_first) != NULL) {
1111 		do {
1112 			old_xid = BE32_TO_U32(&mp1->b_rptr[4]);
1113 			if (new_xid == old_xid)
1114 				return (1);
1115 		} while ((mp1 = mp1->b_next) != NULL);
1116 	}
1117 	return (0);
1118 }
1119 
1120 static int
1121 mir_close(queue_t *q)
1122 {
1123 	mir_t	*mir = q->q_ptr;
1124 	mblk_t	*mp;
1125 	bool_t queue_cleaned = FALSE;
1126 
1127 	RPCLOG(32, "rpcmod: mir_close of q 0x%p\n", (void *)q);
1128 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
1129 	mutex_enter(&mir->mir_mutex);
1130 	if ((mp = mir->mir_head_mp) != NULL) {
1131 		mir->mir_head_mp = NULL;
1132 		mir->mir_tail_mp = NULL;
1133 		freemsg(mp);
1134 	}
1135 	/*
1136 	 * Set mir_closing so we get notified when MIR_SVC_QUIESCED()
1137 	 * is TRUE.  And mir_timer_start() won't start the timer again.
1138 	 */
1139 	mir->mir_closing = B_TRUE;
1140 	mir_timer_stop(mir);
1141 
1142 	if (mir->mir_type == RPC_SERVER) {
1143 		flushq(q, FLUSHDATA);	/* Ditch anything waiting on read q */
1144 
1145 		/*
1146 		 * This will prevent more requests from arriving and
1147 		 * will force rpcmod to ignore flow control.
1148 		 */
1149 		mir_svc_start_close(WR(q), mir);
1150 
1151 		while ((!MIR_SVC_QUIESCED(mir)) || mir->mir_inwservice == 1) {
1152 
1153 			if (mir->mir_ref_cnt && !mir->mir_inrservice &&
1154 			    (queue_cleaned == FALSE)) {
1155 				/*
1156 				 * call into SVC to clean the queue
1157 				 */
1158 				mutex_exit(&mir->mir_mutex);
1159 				svc_queueclean(q);
1160 				queue_cleaned = TRUE;
1161 				mutex_enter(&mir->mir_mutex);
1162 				continue;
1163 			}
1164 
1165 			/*
1166 			 * Bugid 1253810 - Force the write service
1167 			 * procedure to send its messages, regardless
1168 			 * whether the downstream  module is ready
1169 			 * to accept data.
1170 			 */
1171 			if (mir->mir_inwservice == 1)
1172 				qenable(WR(q));
1173 
1174 			cv_wait(&mir->mir_condvar, &mir->mir_mutex);
1175 		}
1176 
1177 		mutex_exit(&mir->mir_mutex);
1178 		qprocsoff(q);
1179 
1180 		/* Notify kRPC that this stream is going away. */
1181 		svc_queueclose(q);
1182 	} else {
1183 		mutex_exit(&mir->mir_mutex);
1184 		qprocsoff(q);
1185 	}
1186 
1187 	mutex_destroy(&mir->mir_mutex);
1188 	cv_destroy(&mir->mir_condvar);
1189 	cv_destroy(&mir->mir_timer_cv);
1190 	kmem_free(mir, sizeof (mir_t));
1191 	return (0);
1192 }
1193 
1194 /*
1195  * This is server side only (RPC_SERVER).
1196  *
1197  * Exit idle mode.
1198  */
1199 static void
1200 mir_svc_idle_stop(queue_t *q, mir_t *mir)
1201 {
1202 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
1203 	ASSERT((q->q_flag & QREADR) == 0);
1204 	ASSERT(mir->mir_type == RPC_SERVER);
1205 	RPCLOG(16, "rpcmod: mir_svc_idle_stop of q 0x%p\n", (void *)q);
1206 
1207 	mir_timer_stop(mir);
1208 }
1209 
1210 /*
1211  * This is server side only (RPC_SERVER).
1212  *
1213  * Start idle processing, which will include setting idle timer if the
1214  * stream is not being closed.
1215  */
1216 static void
1217 mir_svc_idle_start(queue_t *q, mir_t *mir)
1218 {
1219 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
1220 	ASSERT((q->q_flag & QREADR) == 0);
1221 	ASSERT(mir->mir_type == RPC_SERVER);
1222 	RPCLOG(16, "rpcmod: mir_svc_idle_start q 0x%p\n", (void *)q);
1223 
1224 	/*
1225 	 * Don't re-start idle timer if we are closing queues.
1226 	 */
1227 	if (mir->mir_closing) {
1228 		RPCLOG(16, "mir_svc_idle_start - closing: 0x%p\n",
1229 		    (void *)q);
1230 
1231 		/*
1232 		 * We will call mir_svc_idle_start() whenever MIR_SVC_QUIESCED()
1233 		 * is true.  When it is true, and we are in the process of
1234 		 * closing the stream, signal any thread waiting in
1235 		 * mir_close().
1236 		 */
1237 		if (mir->mir_inwservice == 0)
1238 			cv_signal(&mir->mir_condvar);
1239 
1240 	} else {
1241 		RPCLOG(16, "mir_svc_idle_start - reset %s timer\n",
1242 		    mir->mir_ordrel_pending ? "ordrel" : "normal");
1243 		/*
1244 		 * Normal condition, start the idle timer.  If an orderly
1245 		 * release has been sent, set the timeout to wait for the
1246 		 * client to close its side of the connection.  Otherwise,
1247 		 * use the normal idle timeout.
1248 		 */
1249 		mir_timer_start(q, mir, mir->mir_ordrel_pending ?
1250 		    svc_ordrel_timeout : mir->mir_idle_timeout);
1251 	}
1252 }
1253 
1254 /* ARGSUSED */
1255 static int
1256 mir_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
1257 {
1258 	mir_t	*mir;
1259 
1260 	RPCLOG(32, "rpcmod: mir_open of q 0x%p\n", (void *)q);
1261 	/* Set variables used directly by kRPC. */
1262 	if (!mir_start)
1263 		mir_start = mir_svc_start;
1264 	if (!clnt_stop_idle)
1265 		clnt_stop_idle = mir_clnt_idle_do_stop;
1266 	if (!clnt_max_msg_sizep)
1267 		clnt_max_msg_sizep = &clnt_max_msg_size;
1268 	if (!svc_max_msg_sizep)
1269 		svc_max_msg_sizep = &svc_max_msg_size;
1270 
1271 	/* Allocate a zero'ed out mir structure for this stream. */
1272 	mir = kmem_zalloc(sizeof (mir_t), KM_SLEEP);
1273 
1274 	/*
1275 	 * We set hold inbound here so that incoming messages will
1276 	 * be held on the read-side queue until the stream is completely
1277 	 * initialized with a RPC_CLIENT or RPC_SERVER ioctl.  During
1278 	 * the ioctl processing, the flag is cleared and any messages that
1279 	 * arrived between the open and the ioctl are delivered to kRPC.
1280 	 *
1281 	 * Early data should never arrive on a client stream since
1282 	 * servers only respond to our requests and we do not send any.
1283 	 * until after the stream is initialized.  Early data is
1284 	 * very common on a server stream where the client will start
1285 	 * sending data as soon as the connection is made (and this
1286 	 * is especially true with TCP where the protocol accepts the
1287 	 * connection before nfsd or kRPC is notified about it).
1288 	 */
1289 
1290 	mir->mir_hold_inbound = 1;
1291 
1292 	/*
1293 	 * Start the record marker looking for a 4-byte header.  When
1294 	 * this length is negative, it indicates that rpcmod is looking
1295 	 * for bytes to consume for the record marker header.  When it
1296 	 * is positive, it holds the number of bytes that have arrived
1297 	 * for the current fragment and are being held in mir_header_mp.
1298 	 */
1299 
1300 	mir->mir_frag_len = -(int32_t)sizeof (uint32_t);
1301 
1302 	mir->mir_zoneid = rpc_zoneid();
1303 	mutex_init(&mir->mir_mutex, NULL, MUTEX_DEFAULT, NULL);
1304 	cv_init(&mir->mir_condvar, NULL, CV_DRIVER, NULL);
1305 	cv_init(&mir->mir_timer_cv, NULL, CV_DRIVER, NULL);
1306 
1307 	q->q_ptr = (char *)mir;
1308 	WR(q)->q_ptr = (char *)mir;
1309 
1310 	/*
1311 	 * We noenable the read-side queue because we don't want it
1312 	 * automatically enabled by putq.  We enable it explicitly
1313 	 * in mir_wsrv when appropriate. (See additional comments on
1314 	 * flow control at the beginning of mir_rsrv.)
1315 	 */
1316 	noenable(q);
1317 
1318 	qprocson(q);
1319 	return (0);
1320 }
1321 
1322 /*
1323  * Read-side put routine for both the client and server side.  Does the
1324  * record marking for incoming RPC messages, and when complete, dispatches
1325  * the message to either the client or server.
1326  */
1327 static void
1328 mir_rput(queue_t *q, mblk_t *mp)
1329 {
1330 	int	excess;
1331 	int32_t	frag_len, frag_header;
1332 	mblk_t	*cont_mp, *head_mp, *tail_mp, *mp1;
1333 	mir_t	*mir = q->q_ptr;
1334 	boolean_t stop_timer = B_FALSE;
1335 
1336 	ASSERT(mir != NULL);
1337 
1338 	/*
1339 	 * If the stream has not been set up as a RPC_CLIENT or RPC_SERVER
1340 	 * with the corresponding ioctl, then don't accept
1341 	 * any inbound data.  This should never happen for streams
1342 	 * created by nfsd or client-side kRPC because they are careful
1343 	 * to set the mode of the stream before doing anything else.
1344 	 */
1345 	if (mir->mir_type == 0) {
1346 		freemsg(mp);
1347 		return;
1348 	}
1349 
1350 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
1351 
1352 	switch (mp->b_datap->db_type) {
1353 	case M_DATA:
1354 		break;
1355 	case M_PROTO:
1356 	case M_PCPROTO:
1357 		if (MBLKL(mp) < sizeof (t_scalar_t)) {
1358 			RPCLOG(1, "mir_rput: runt TPI message (%d bytes)\n",
1359 			    (int)MBLKL(mp));
1360 			freemsg(mp);
1361 			return;
1362 		}
1363 		if (((union T_primitives *)mp->b_rptr)->type != T_DATA_IND) {
1364 			mir_rput_proto(q, mp);
1365 			return;
1366 		}
1367 
1368 		/* Throw away the T_DATA_IND block and continue with data. */
1369 		mp1 = mp;
1370 		mp = mp->b_cont;
1371 		freeb(mp1);
1372 		break;
1373 	case M_SETOPTS:
1374 		/*
1375 		 * If a module on the stream is trying set the Stream head's
1376 		 * high water mark, then set our hiwater to the requested
1377 		 * value.  We are the "stream head" for all inbound
1378 		 * data messages since messages are passed directly to kRPC.
1379 		 */
1380 		if (MBLKL(mp) >= sizeof (struct stroptions)) {
1381 			struct stroptions	*stropts;
1382 
1383 			stropts = (struct stroptions *)mp->b_rptr;
1384 			if ((stropts->so_flags & SO_HIWAT) &&
1385 			    !(stropts->so_flags & SO_BAND)) {
1386 				(void) strqset(q, QHIWAT, 0, stropts->so_hiwat);
1387 			}
1388 		}
1389 		putnext(q, mp);
1390 		return;
1391 	case M_FLUSH:
1392 		RPCLOG(32, "mir_rput: ignoring M_FLUSH %x ", *mp->b_rptr);
1393 		RPCLOG(32, "on q 0x%p\n", (void *)q);
1394 		putnext(q, mp);
1395 		return;
1396 	default:
1397 		putnext(q, mp);
1398 		return;
1399 	}
1400 
1401 	mutex_enter(&mir->mir_mutex);
1402 
1403 	/*
1404 	 * If this connection is closing, don't accept any new messages.
1405 	 */
1406 	if (mir->mir_svc_no_more_msgs) {
1407 		ASSERT(mir->mir_type == RPC_SERVER);
1408 		mutex_exit(&mir->mir_mutex);
1409 		freemsg(mp);
1410 		return;
1411 	}
1412 
1413 	/* Get local copies for quicker access. */
1414 	frag_len = mir->mir_frag_len;
1415 	frag_header = mir->mir_frag_header;
1416 	head_mp = mir->mir_head_mp;
1417 	tail_mp = mir->mir_tail_mp;
1418 
1419 	/* Loop, processing each message block in the mp chain separately. */
1420 	do {
1421 		cont_mp = mp->b_cont;
1422 		mp->b_cont = NULL;
1423 
1424 		/*
1425 		 * Drop zero-length mblks to prevent unbounded kernel memory
1426 		 * consumption.
1427 		 */
1428 		if (MBLKL(mp) == 0) {
1429 			freeb(mp);
1430 			continue;
1431 		}
1432 
1433 		/*
1434 		 * If frag_len is negative, we're still in the process of
1435 		 * building frag_header -- try to complete it with this mblk.
1436 		 */
1437 		while (frag_len < 0 && mp->b_rptr < mp->b_wptr) {
1438 			frag_len++;
1439 			frag_header <<= 8;
1440 			frag_header += *mp->b_rptr++;
1441 		}
1442 
1443 		if (MBLKL(mp) == 0 && frag_len < 0) {
1444 			/*
1445 			 * We consumed this mblk while trying to complete the
1446 			 * fragment header.  Free it and move on.
1447 			 */
1448 			freeb(mp);
1449 			continue;
1450 		}
1451 
1452 		ASSERT(frag_len >= 0);
1453 
1454 		/*
1455 		 * Now frag_header has the number of bytes in this fragment
1456 		 * and we're just waiting to collect them all.  Chain our
1457 		 * latest mblk onto the list and see if we now have enough
1458 		 * bytes to complete the fragment.
1459 		 */
1460 		if (head_mp == NULL) {
1461 			ASSERT(tail_mp == NULL);
1462 			head_mp = tail_mp = mp;
1463 		} else {
1464 			tail_mp->b_cont = mp;
1465 			tail_mp = mp;
1466 		}
1467 
1468 		frag_len += MBLKL(mp);
1469 		excess = frag_len - (frag_header & ~MIR_LASTFRAG);
1470 		if (excess < 0) {
1471 			/*
1472 			 * We still haven't received enough data to complete
1473 			 * the fragment, so continue on to the next mblk.
1474 			 */
1475 			continue;
1476 		}
1477 
1478 		/*
1479 		 * We've got a complete fragment.  If there are excess bytes,
1480 		 * then they're part of the next fragment's header (of either
1481 		 * this RPC message or the next RPC message).  Split that part
1482 		 * into its own mblk so that we can safely freeb() it when
1483 		 * building frag_header above.
1484 		 */
1485 		if (excess > 0) {
1486 			if ((mp1 = dupb(mp)) == NULL &&
1487 			    (mp1 = copyb(mp)) == NULL) {
1488 				freemsg(head_mp);
1489 				freemsg(cont_mp);
1490 				RPCLOG0(1, "mir_rput: dupb/copyb failed\n");
1491 				mir->mir_frag_header = 0;
1492 				mir->mir_frag_len = -(int32_t)sizeof (uint32_t);
1493 				mir->mir_head_mp = NULL;
1494 				mir->mir_tail_mp = NULL;
1495 				mir_disconnect(q, mir);	/* drops mir_mutex */
1496 				return;
1497 			}
1498 
1499 			/*
1500 			 * Relink the message chain so that the next mblk is
1501 			 * the next fragment header, followed by the rest of
1502 			 * the message chain.
1503 			 */
1504 			mp1->b_cont = cont_mp;
1505 			cont_mp = mp1;
1506 
1507 			/*
1508 			 * Data in the new mblk begins at the next fragment,
1509 			 * and data in the old mblk ends at the next fragment.
1510 			 */
1511 			mp1->b_rptr = mp1->b_wptr - excess;
1512 			mp->b_wptr -= excess;
1513 		}
1514 
1515 		/*
1516 		 * Reset frag_len and frag_header for the next fragment.
1517 		 */
1518 		frag_len = -(int32_t)sizeof (uint32_t);
1519 		if (!(frag_header & MIR_LASTFRAG)) {
1520 			/*
1521 			 * The current fragment is complete, but more
1522 			 * fragments need to be processed before we can
1523 			 * pass along the RPC message headed at head_mp.
1524 			 */
1525 			frag_header = 0;
1526 			continue;
1527 		}
1528 		frag_header = 0;
1529 
1530 		/*
1531 		 * We've got a complete RPC message; pass it to the
1532 		 * appropriate consumer.
1533 		 */
1534 		switch (mir->mir_type) {
1535 		case RPC_CLIENT:
1536 			if (clnt_dispatch_notify(head_mp, mir->mir_zoneid)) {
1537 				/*
1538 				 * Mark this stream as active.  This marker
1539 				 * is used in mir_timer().
1540 				 */
1541 				mir->mir_clntreq = 1;
1542 				mir->mir_use_timestamp = ddi_get_lbolt();
1543 			} else {
1544 				freemsg(head_mp);
1545 			}
1546 			break;
1547 
1548 		case RPC_SERVER:
1549 			/*
1550 			 * Check for flow control before passing the
1551 			 * message to kRPC.
1552 			 */
1553 			if (!mir->mir_hold_inbound) {
1554 				if (mir->mir_krpc_cell) {
1555 
1556 					if (mir_check_len(q, head_mp))
1557 						return;
1558 
1559 					if (q->q_first == NULL &&
1560 					    svc_queuereq(q, head_mp, TRUE)) {
1561 						/*
1562 						 * If the reference count is 0
1563 						 * (not including this
1564 						 * request), then the stream is
1565 						 * transitioning from idle to
1566 						 * non-idle.  In this case, we
1567 						 * cancel the idle timer.
1568 						 */
1569 						if (mir->mir_ref_cnt++ == 0)
1570 							stop_timer = B_TRUE;
1571 					} else {
1572 						(void) putq(q, head_mp);
1573 						mir->mir_inrservice = B_TRUE;
1574 					}
1575 				} else {
1576 					/*
1577 					 * Count # of times this happens. Should
1578 					 * be never, but experience shows
1579 					 * otherwise.
1580 					 */
1581 					mir_krpc_cell_null++;
1582 					freemsg(head_mp);
1583 				}
1584 			} else {
1585 				/*
1586 				 * If the outbound side of the stream is
1587 				 * flow controlled, then hold this message
1588 				 * until client catches up. mir_hold_inbound
1589 				 * is set in mir_wput and cleared in mir_wsrv.
1590 				 */
1591 				(void) putq(q, head_mp);
1592 				mir->mir_inrservice = B_TRUE;
1593 			}
1594 			break;
1595 		default:
1596 			RPCLOG(1, "mir_rput: unknown mir_type %d\n",
1597 			    mir->mir_type);
1598 			freemsg(head_mp);
1599 			break;
1600 		}
1601 
1602 		/*
1603 		 * Reset the chain since we're starting on a new RPC message.
1604 		 */
1605 		head_mp = tail_mp = NULL;
1606 	} while ((mp = cont_mp) != NULL);
1607 
1608 	/*
1609 	 * Sanity check the message length; if it's too large mir_check_len()
1610 	 * will shutdown the connection, drop mir_mutex, and return non-zero.
1611 	 */
1612 	if (head_mp != NULL && mir->mir_setup_complete &&
1613 	    mir_check_len(q, head_mp))
1614 		return;
1615 
1616 	/* Save our local copies back in the mir structure. */
1617 	mir->mir_frag_header = frag_header;
1618 	mir->mir_frag_len = frag_len;
1619 	mir->mir_head_mp = head_mp;
1620 	mir->mir_tail_mp = tail_mp;
1621 
1622 	/*
1623 	 * The timer is stopped after the whole message chain is processed.
1624 	 * The reason is that stopping the timer releases the mir_mutex
1625 	 * lock temporarily.  This means that the request can be serviced
1626 	 * while we are still processing the message chain.  This is not
1627 	 * good.  So we stop the timer here instead.
1628 	 *
1629 	 * Note that if the timer fires before we stop it, it will not
1630 	 * do any harm as MIR_SVC_QUIESCED() is false and mir_timer()
1631 	 * will just return.
1632 	 */
1633 	if (stop_timer) {
1634 		RPCLOG(16, "mir_rput: stopping idle timer on 0x%p because "
1635 		    "ref cnt going to non zero\n", (void *)WR(q));
1636 		mir_svc_idle_stop(WR(q), mir);
1637 	}
1638 	mutex_exit(&mir->mir_mutex);
1639 }
1640 
1641 static void
1642 mir_rput_proto(queue_t *q, mblk_t *mp)
1643 {
1644 	mir_t	*mir = (mir_t *)q->q_ptr;
1645 	uint32_t	type;
1646 	uint32_t reason = 0;
1647 
1648 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
1649 
1650 	type = ((union T_primitives *)mp->b_rptr)->type;
1651 	switch (mir->mir_type) {
1652 	case RPC_CLIENT:
1653 		switch (type) {
1654 		case T_DISCON_IND:
1655 			reason = ((struct T_discon_ind *)
1656 			    (mp->b_rptr))->DISCON_reason;
1657 			/*FALLTHROUGH*/
1658 		case T_ORDREL_IND:
1659 			mutex_enter(&mir->mir_mutex);
1660 			if (mir->mir_head_mp) {
1661 				freemsg(mir->mir_head_mp);
1662 				mir->mir_head_mp = (mblk_t *)0;
1663 				mir->mir_tail_mp = (mblk_t *)0;
1664 			}
1665 			/*
1666 			 * We are disconnecting, but not necessarily
1667 			 * closing. By not closing, we will fail to
1668 			 * pick up a possibly changed global timeout value,
1669 			 * unless we store it now.
1670 			 */
1671 			mir->mir_idle_timeout = clnt_idle_timeout;
1672 			mir_clnt_idle_stop(WR(q), mir);
1673 
1674 			/*
1675 			 * Even though we are unconnected, we still
1676 			 * leave the idle timer going on the client. The
1677 			 * reason for is that if we've disconnected due
1678 			 * to a server-side disconnect, reset, or connection
1679 			 * timeout, there is a possibility the client may
1680 			 * retry the RPC request. This retry needs to done on
1681 			 * the same bound address for the server to interpret
1682 			 * it as such. However, we don't want
1683 			 * to wait forever for that possibility. If the
1684 			 * end-point stays unconnected for mir_idle_timeout
1685 			 * units of time, then that is a signal to the
1686 			 * connection manager to give up waiting for the
1687 			 * application (eg. NFS) to send a retry.
1688 			 */
1689 			mir_clnt_idle_start(WR(q), mir);
1690 			mutex_exit(&mir->mir_mutex);
1691 			clnt_dispatch_notifyall(WR(q), type, reason);
1692 			freemsg(mp);
1693 			return;
1694 		case T_ERROR_ACK:
1695 		{
1696 			struct T_error_ack	*terror;
1697 
1698 			terror = (struct T_error_ack *)mp->b_rptr;
1699 			RPCLOG(1, "mir_rput_proto T_ERROR_ACK for queue 0x%p",
1700 			    (void *)q);
1701 			RPCLOG(1, " ERROR_prim: %s,",
1702 			    rpc_tpiprim2name(terror->ERROR_prim));
1703 			RPCLOG(1, " TLI_error: %s,",
1704 			    rpc_tpierr2name(terror->TLI_error));
1705 			RPCLOG(1, " UNIX_error: %d\n", terror->UNIX_error);
1706 			if (terror->ERROR_prim == T_DISCON_REQ)  {
1707 				clnt_dispatch_notifyall(WR(q), type, reason);
1708 				freemsg(mp);
1709 				return;
1710 			} else {
1711 				if (clnt_dispatch_notifyconn(WR(q), mp))
1712 					return;
1713 			}
1714 			break;
1715 		}
1716 		case T_OK_ACK:
1717 		{
1718 			struct T_ok_ack	*tok = (struct T_ok_ack *)mp->b_rptr;
1719 
1720 			if (tok->CORRECT_prim == T_DISCON_REQ) {
1721 				clnt_dispatch_notifyall(WR(q), type, reason);
1722 				freemsg(mp);
1723 				return;
1724 			} else {
1725 				if (clnt_dispatch_notifyconn(WR(q), mp))
1726 					return;
1727 			}
1728 			break;
1729 		}
1730 		case T_CONN_CON:
1731 		case T_INFO_ACK:
1732 		case T_OPTMGMT_ACK:
1733 			if (clnt_dispatch_notifyconn(WR(q), mp))
1734 				return;
1735 			break;
1736 		case T_BIND_ACK:
1737 			break;
1738 		default:
1739 			RPCLOG(1, "mir_rput: unexpected message %d "
1740 			    "for kRPC client\n",
1741 			    ((union T_primitives *)mp->b_rptr)->type);
1742 			break;
1743 		}
1744 		break;
1745 
1746 	case RPC_SERVER:
1747 		switch (type) {
1748 		case T_BIND_ACK:
1749 		{
1750 			struct T_bind_ack	*tbind;
1751 
1752 			/*
1753 			 * If this is a listening stream, then shut
1754 			 * off the idle timer.
1755 			 */
1756 			tbind = (struct T_bind_ack *)mp->b_rptr;
1757 			if (tbind->CONIND_number > 0) {
1758 				mutex_enter(&mir->mir_mutex);
1759 				mir_svc_idle_stop(WR(q), mir);
1760 
1761 				/*
1762 				 * mark this as a listen endpoint
1763 				 * for special handling.
1764 				 */
1765 
1766 				mir->mir_listen_stream = 1;
1767 				mutex_exit(&mir->mir_mutex);
1768 			}
1769 			break;
1770 		}
1771 		case T_DISCON_IND:
1772 		case T_ORDREL_IND:
1773 			RPCLOG(16, "mir_rput_proto: got %s indication\n",
1774 			    type == T_DISCON_IND ? "disconnect"
1775 			    : "orderly release");
1776 
1777 			/*
1778 			 * For listen endpoint just pass
1779 			 * on the message.
1780 			 */
1781 
1782 			if (mir->mir_listen_stream)
1783 				break;
1784 
1785 			mutex_enter(&mir->mir_mutex);
1786 
1787 			/*
1788 			 * If client wants to break off connection, record
1789 			 * that fact.
1790 			 */
1791 			mir_svc_start_close(WR(q), mir);
1792 
1793 			/*
1794 			 * If we are idle, then send the orderly release
1795 			 * or disconnect indication to nfsd.
1796 			 */
1797 			if (MIR_SVC_QUIESCED(mir)) {
1798 				mutex_exit(&mir->mir_mutex);
1799 				break;
1800 			}
1801 
1802 			RPCLOG(16, "mir_rput_proto: not idle, so "
1803 			    "disconnect/ord rel indication not passed "
1804 			    "upstream on 0x%p\n", (void *)q);
1805 
1806 			/*
1807 			 * Hold the indication until we get idle
1808 			 * If there already is an indication stored,
1809 			 * replace it if the new one is a disconnect. The
1810 			 * reasoning is that disconnection takes less time
1811 			 * to process, and once a client decides to
1812 			 * disconnect, we should do that.
1813 			 */
1814 			if (mir->mir_svc_pend_mp) {
1815 				if (type == T_DISCON_IND) {
1816 					RPCLOG(16, "mir_rput_proto: replacing"
1817 					    " held disconnect/ord rel"
1818 					    " indication with disconnect on"
1819 					    " 0x%p\n", (void *)q);
1820 
1821 					freemsg(mir->mir_svc_pend_mp);
1822 					mir->mir_svc_pend_mp = mp;
1823 				} else {
1824 					RPCLOG(16, "mir_rput_proto: already "
1825 					    "held a disconnect/ord rel "
1826 					    "indication. freeing ord rel "
1827 					    "ind on 0x%p\n", (void *)q);
1828 					freemsg(mp);
1829 				}
1830 			} else
1831 				mir->mir_svc_pend_mp = mp;
1832 
1833 			mutex_exit(&mir->mir_mutex);
1834 			return;
1835 
1836 		default:
1837 			/* nfsd handles server-side non-data messages. */
1838 			break;
1839 		}
1840 		break;
1841 
1842 	default:
1843 		break;
1844 	}
1845 
1846 	putnext(q, mp);
1847 }
1848 
1849 /*
1850  * The server-side read queues are used to hold inbound messages while
1851  * outbound flow control is exerted.  When outbound flow control is
1852  * relieved, mir_wsrv qenables the read-side queue.  Read-side queues
1853  * are not enabled by STREAMS and are explicitly noenable'ed in mir_open.
1854  */
1855 static void
1856 mir_rsrv(queue_t *q)
1857 {
1858 	mir_t	*mir;
1859 	mblk_t	*mp;
1860 	boolean_t stop_timer = B_FALSE;
1861 
1862 	mir = (mir_t *)q->q_ptr;
1863 	mutex_enter(&mir->mir_mutex);
1864 
1865 	mp = NULL;
1866 	switch (mir->mir_type) {
1867 	case RPC_SERVER:
1868 		if (mir->mir_ref_cnt == 0)
1869 			mir->mir_hold_inbound = 0;
1870 		if (mir->mir_hold_inbound)
1871 			break;
1872 
1873 		while (mp = getq(q)) {
1874 			if (mir->mir_krpc_cell &&
1875 			    (mir->mir_svc_no_more_msgs == 0)) {
1876 
1877 				if (mir_check_len(q, mp))
1878 					return;
1879 
1880 				if (svc_queuereq(q, mp, TRUE)) {
1881 					/*
1882 					 * If we were idle, turn off idle timer
1883 					 * since we aren't idle any more.
1884 					 */
1885 					if (mir->mir_ref_cnt++ == 0)
1886 						stop_timer = B_TRUE;
1887 				} else {
1888 					(void) putbq(q, mp);
1889 					break;
1890 				}
1891 			} else {
1892 				/*
1893 				 * Count # of times this happens. Should be
1894 				 * never, but experience shows otherwise.
1895 				 */
1896 				if (mir->mir_krpc_cell == NULL)
1897 					mir_krpc_cell_null++;
1898 				freemsg(mp);
1899 			}
1900 		}
1901 		break;
1902 	case RPC_CLIENT:
1903 		break;
1904 	default:
1905 		RPCLOG(1, "mir_rsrv: unexpected mir_type %d\n", mir->mir_type);
1906 
1907 		if (q->q_first == NULL)
1908 			MIR_CLEAR_INRSRV(mir);
1909 
1910 		mutex_exit(&mir->mir_mutex);
1911 
1912 		return;
1913 	}
1914 
1915 	/*
1916 	 * The timer is stopped after all the messages are processed.
1917 	 * The reason is that stopping the timer releases the mir_mutex
1918 	 * lock temporarily.  This means that the request can be serviced
1919 	 * while we are still processing the message queue.  This is not
1920 	 * good.  So we stop the timer here instead.
1921 	 */
1922 	if (stop_timer)  {
1923 		RPCLOG(16, "mir_rsrv stopping idle timer on 0x%p because ref "
1924 		    "cnt going to non zero\n", (void *)WR(q));
1925 		mir_svc_idle_stop(WR(q), mir);
1926 	}
1927 
1928 	if (q->q_first == NULL) {
1929 		mblk_t	*cmp = NULL;
1930 
1931 		MIR_CLEAR_INRSRV(mir);
1932 
1933 		if (mir->mir_type == RPC_SERVER && MIR_SVC_QUIESCED(mir)) {
1934 			cmp = mir->mir_svc_pend_mp;
1935 			mir->mir_svc_pend_mp = NULL;
1936 		}
1937 
1938 		mutex_exit(&mir->mir_mutex);
1939 
1940 		if (cmp != NULL) {
1941 			RPCLOG(16, "mir_rsrv: line %d: sending a held "
1942 			    "disconnect/ord rel indication upstream\n",
1943 			    __LINE__);
1944 			putnext(q, cmp);
1945 		}
1946 
1947 		return;
1948 	}
1949 	mutex_exit(&mir->mir_mutex);
1950 }
1951 
1952 static int mir_svc_policy_fails;
1953 
1954 /*
1955  * Called to send an event code to nfsd/lockd so that it initiates
1956  * connection close.
1957  */
1958 static int
1959 mir_svc_policy_notify(queue_t *q, int event)
1960 {
1961 	mblk_t	*mp;
1962 #ifdef DEBUG
1963 	mir_t *mir = (mir_t *)q->q_ptr;
1964 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
1965 #endif
1966 	ASSERT(q->q_flag & QREADR);
1967 
1968 	/*
1969 	 * Create an M_DATA message with the event code and pass it to the
1970 	 * Stream head (nfsd or whoever created the stream will consume it).
1971 	 */
1972 	mp = allocb(sizeof (int), BPRI_HI);
1973 
1974 	if (!mp) {
1975 
1976 		mir_svc_policy_fails++;
1977 		RPCLOG(16, "mir_svc_policy_notify: could not allocate event "
1978 		    "%d\n", event);
1979 		return (ENOMEM);
1980 	}
1981 
1982 	U32_TO_BE32(event, mp->b_rptr);
1983 	mp->b_wptr = mp->b_rptr + sizeof (int);
1984 	putnext(q, mp);
1985 	return (0);
1986 }
1987 
1988 /*
1989  * Server side: start the close phase. We want to get this rpcmod slot in an
1990  * idle state before mir_close() is called.
1991  */
1992 static void
1993 mir_svc_start_close(queue_t *wq, mir_t *mir)
1994 {
1995 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
1996 	ASSERT((wq->q_flag & QREADR) == 0);
1997 	ASSERT(mir->mir_type == RPC_SERVER);
1998 
1999 	/*
2000 	 * Do not accept any more messages.
2001 	 */
2002 	mir->mir_svc_no_more_msgs = 1;
2003 
2004 	/*
2005 	 * Next two statements will make the read service procedure
2006 	 * free everything stuck in the streams read queue.
2007 	 * It's not necessary because enabling the write queue will
2008 	 * have the same effect, but why not speed the process along?
2009 	 */
2010 	mir->mir_hold_inbound = 0;
2011 	qenable(RD(wq));
2012 
2013 	/*
2014 	 * Meanwhile force the write service procedure to send the
2015 	 * responses downstream, regardless of flow control.
2016 	 */
2017 	qenable(wq);
2018 }
2019 
2020 void
2021 mir_svc_hold(queue_t *wq)
2022 {
2023 	mir_t *mir = (mir_t *)wq->q_ptr;
2024 
2025 	mutex_enter(&mir->mir_mutex);
2026 	mir->mir_ref_cnt++;
2027 	mutex_exit(&mir->mir_mutex);
2028 }
2029 
2030 /*
2031  * This routine is called directly by kRPC after a request is completed,
2032  * whether a reply was sent or the request was dropped.
2033  */
2034 void
2035 mir_svc_release(queue_t *wq, mblk_t *mp, bool_t enable)
2036 {
2037 	mir_t   *mir = (mir_t *)wq->q_ptr;
2038 	mblk_t	*cmp = NULL;
2039 
2040 	ASSERT((wq->q_flag & QREADR) == 0);
2041 	if (mp)
2042 		freemsg(mp);
2043 
2044 	if (enable)
2045 		qenable(RD(wq));
2046 
2047 	mutex_enter(&mir->mir_mutex);
2048 
2049 	/*
2050 	 * Start idle processing if this is the last reference.
2051 	 */
2052 	if ((mir->mir_ref_cnt == 1) && (mir->mir_inrservice == 0)) {
2053 		cmp = mir->mir_svc_pend_mp;
2054 		mir->mir_svc_pend_mp = NULL;
2055 	}
2056 
2057 	if (cmp) {
2058 		RPCLOG(16, "mir_svc_release: sending a held "
2059 		    "disconnect/ord rel indication upstream on queue 0x%p\n",
2060 		    (void *)RD(wq));
2061 
2062 		mutex_exit(&mir->mir_mutex);
2063 
2064 		putnext(RD(wq), cmp);
2065 
2066 		mutex_enter(&mir->mir_mutex);
2067 	}
2068 
2069 	/*
2070 	 * Start idle processing if this is the last reference.
2071 	 */
2072 	if (mir->mir_ref_cnt == 1 && mir->mir_inrservice == 0) {
2073 
2074 		RPCLOG(16, "mir_svc_release starting idle timer on 0x%p "
2075 		    "because ref cnt is zero\n", (void *) wq);
2076 
2077 		mir_svc_idle_start(wq, mir);
2078 	}
2079 
2080 	mir->mir_ref_cnt--;
2081 	ASSERT(mir->mir_ref_cnt >= 0);
2082 
2083 	/*
2084 	 * Wake up the thread waiting to close.
2085 	 */
2086 
2087 	if ((mir->mir_ref_cnt == 0) && mir->mir_closing)
2088 		cv_signal(&mir->mir_condvar);
2089 
2090 	mutex_exit(&mir->mir_mutex);
2091 }
2092 
2093 /*
2094  * This routine is called by server-side kRPC when it is ready to
2095  * handle inbound messages on the stream.
2096  */
2097 static void
2098 mir_svc_start(queue_t *wq)
2099 {
2100 	mir_t   *mir = (mir_t *)wq->q_ptr;
2101 
2102 	/*
2103 	 * no longer need to take the mir_mutex because the
2104 	 * mir_setup_complete field has been moved out of
2105 	 * the binary field protected by the mir_mutex.
2106 	 */
2107 
2108 	mir->mir_setup_complete = 1;
2109 	qenable(RD(wq));
2110 }
2111 
2112 /*
2113  * client side wrapper for stopping timer with normal idle timeout.
2114  */
2115 static void
2116 mir_clnt_idle_stop(queue_t *wq, mir_t *mir)
2117 {
2118 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
2119 	ASSERT((wq->q_flag & QREADR) == 0);
2120 	ASSERT(mir->mir_type == RPC_CLIENT);
2121 
2122 	mir_timer_stop(mir);
2123 }
2124 
2125 /*
2126  * client side wrapper for stopping timer with normal idle timeout.
2127  */
2128 static void
2129 mir_clnt_idle_start(queue_t *wq, mir_t *mir)
2130 {
2131 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
2132 	ASSERT((wq->q_flag & QREADR) == 0);
2133 	ASSERT(mir->mir_type == RPC_CLIENT);
2134 
2135 	mir_timer_start(wq, mir, mir->mir_idle_timeout);
2136 }
2137 
2138 /*
2139  * client side only. Forces rpcmod to stop sending T_ORDREL_REQs on
2140  * end-points that aren't connected.
2141  */
2142 static void
2143 mir_clnt_idle_do_stop(queue_t *wq)
2144 {
2145 	mir_t   *mir = (mir_t *)wq->q_ptr;
2146 
2147 	RPCLOG(1, "mir_clnt_idle_do_stop: wq 0x%p\n", (void *)wq);
2148 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
2149 	mutex_enter(&mir->mir_mutex);
2150 	mir_clnt_idle_stop(wq, mir);
2151 	mutex_exit(&mir->mir_mutex);
2152 }
2153 
2154 /*
2155  * Timer handler.  It handles idle timeout and memory shortage problem.
2156  */
2157 static void
2158 mir_timer(void *arg)
2159 {
2160 	queue_t *wq = (queue_t *)arg;
2161 	mir_t *mir = (mir_t *)wq->q_ptr;
2162 	boolean_t notify;
2163 	clock_t now;
2164 
2165 	mutex_enter(&mir->mir_mutex);
2166 
2167 	/*
2168 	 * mir_timer_call is set only when either mir_timer_[start|stop]
2169 	 * is progressing.  And mir_timer() can only be run while they
2170 	 * are progressing if the timer is being stopped.  So just
2171 	 * return.
2172 	 */
2173 	if (mir->mir_timer_call) {
2174 		mutex_exit(&mir->mir_mutex);
2175 		return;
2176 	}
2177 	mir->mir_timer_id = 0;
2178 
2179 	switch (mir->mir_type) {
2180 	case RPC_CLIENT:
2181 
2182 		/*
2183 		 * For clients, the timer fires at clnt_idle_timeout
2184 		 * intervals.  If the activity marker (mir_clntreq) is
2185 		 * zero, then the stream has been idle since the last
2186 		 * timer event and we notify kRPC.  If mir_clntreq is
2187 		 * non-zero, then the stream is active and we just
2188 		 * restart the timer for another interval.  mir_clntreq
2189 		 * is set to 1 in mir_wput for every request passed
2190 		 * downstream.
2191 		 *
2192 		 * If this was a memory shortage timer reset the idle
2193 		 * timeout regardless; the mir_clntreq will not be a
2194 		 * valid indicator.
2195 		 *
2196 		 * The timer is initially started in mir_wput during
2197 		 * RPC_CLIENT ioctl processing.
2198 		 *
2199 		 * The timer interval can be changed for individual
2200 		 * streams with the ND variable "mir_idle_timeout".
2201 		 */
2202 		now = ddi_get_lbolt();
2203 		if (mir->mir_clntreq > 0 && mir->mir_use_timestamp +
2204 		    MSEC_TO_TICK(mir->mir_idle_timeout) - now >= 0) {
2205 			clock_t tout;
2206 
2207 			tout = mir->mir_idle_timeout -
2208 			    TICK_TO_MSEC(now - mir->mir_use_timestamp);
2209 			if (tout < 0)
2210 				tout = 1000;
2211 #if 0
2212 			printf("mir_timer[%d < %d + %d]: reset client timer "
2213 			    "to %d (ms)\n", TICK_TO_MSEC(now),
2214 			    TICK_TO_MSEC(mir->mir_use_timestamp),
2215 			    mir->mir_idle_timeout, tout);
2216 #endif
2217 			mir->mir_clntreq = 0;
2218 			mir_timer_start(wq, mir, tout);
2219 			mutex_exit(&mir->mir_mutex);
2220 			return;
2221 		}
2222 #if 0
2223 printf("mir_timer[%d]: doing client timeout\n", now / hz);
2224 #endif
2225 		/*
2226 		 * We are disconnecting, but not necessarily
2227 		 * closing. By not closing, we will fail to
2228 		 * pick up a possibly changed global timeout value,
2229 		 * unless we store it now.
2230 		 */
2231 		mir->mir_idle_timeout = clnt_idle_timeout;
2232 		mir_clnt_idle_start(wq, mir);
2233 
2234 		mutex_exit(&mir->mir_mutex);
2235 		/*
2236 		 * We pass T_ORDREL_REQ as an integer value
2237 		 * to kRPC as the indication that the stream
2238 		 * is idle.  This is not a T_ORDREL_REQ message,
2239 		 * it is just a convenient value since we call
2240 		 * the same kRPC routine for T_ORDREL_INDs and
2241 		 * T_DISCON_INDs.
2242 		 */
2243 		clnt_dispatch_notifyall(wq, T_ORDREL_REQ, 0);
2244 		return;
2245 
2246 	case RPC_SERVER:
2247 
2248 		/*
2249 		 * For servers, the timer is only running when the stream
2250 		 * is really idle or memory is short.  The timer is started
2251 		 * by mir_wput when mir_type is set to RPC_SERVER and
2252 		 * by mir_svc_idle_start whenever the stream goes idle
2253 		 * (mir_ref_cnt == 0).  The timer is cancelled in
2254 		 * mir_rput whenever a new inbound request is passed to kRPC
2255 		 * and the stream was previously idle.
2256 		 *
2257 		 * The timer interval can be changed for individual
2258 		 * streams with the ND variable "mir_idle_timeout".
2259 		 *
2260 		 * If the stream is not idle do nothing.
2261 		 */
2262 		if (!MIR_SVC_QUIESCED(mir)) {
2263 			mutex_exit(&mir->mir_mutex);
2264 			return;
2265 		}
2266 
2267 		notify = !mir->mir_inrservice;
2268 		mutex_exit(&mir->mir_mutex);
2269 
2270 		/*
2271 		 * If there is no packet queued up in read queue, the stream
2272 		 * is really idle so notify nfsd to close it.
2273 		 */
2274 		if (notify) {
2275 			RPCLOG(16, "mir_timer: telling stream head listener "
2276 			    "to close stream (0x%p)\n", (void *) RD(wq));
2277 			(void) mir_svc_policy_notify(RD(wq), 1);
2278 		}
2279 		return;
2280 	default:
2281 		RPCLOG(1, "mir_timer: unexpected mir_type %d\n",
2282 		    mir->mir_type);
2283 		mutex_exit(&mir->mir_mutex);
2284 		return;
2285 	}
2286 }
2287 
2288 /*
2289  * Called by the RPC package to send either a call or a return, or a
2290  * transport connection request.  Adds the record marking header.
2291  */
2292 static void
2293 mir_wput(queue_t *q, mblk_t *mp)
2294 {
2295 	uint_t	frag_header;
2296 	mir_t	*mir = (mir_t *)q->q_ptr;
2297 	uchar_t	*rptr = mp->b_rptr;
2298 
2299 	if (!mir) {
2300 		freemsg(mp);
2301 		return;
2302 	}
2303 
2304 	if (mp->b_datap->db_type != M_DATA) {
2305 		mir_wput_other(q, mp);
2306 		return;
2307 	}
2308 
2309 	if (mir->mir_ordrel_pending == 1) {
2310 		freemsg(mp);
2311 		RPCLOG(16, "mir_wput wq 0x%p: got data after T_ORDREL_REQ\n",
2312 		    (void *)q);
2313 		return;
2314 	}
2315 
2316 	frag_header = (uint_t)DLEN(mp);
2317 	frag_header |= MIR_LASTFRAG;
2318 
2319 	/* Stick in the 4 byte record marking header. */
2320 	if ((rptr - mp->b_datap->db_base) < sizeof (uint32_t) ||
2321 	    !IS_P2ALIGNED(mp->b_rptr, sizeof (uint32_t))) {
2322 		/*
2323 		 * Since we know that M_DATA messages are created exclusively
2324 		 * by kRPC, we expect that kRPC will leave room for our header
2325 		 * and 4 byte align which is normal for XDR.
2326 		 * If kRPC (or someone else) does not cooperate, then we
2327 		 * just throw away the message.
2328 		 */
2329 		RPCLOG(1, "mir_wput: kRPC did not leave space for record "
2330 		    "fragment header (%d bytes left)\n",
2331 		    (int)(rptr - mp->b_datap->db_base));
2332 		freemsg(mp);
2333 		return;
2334 	}
2335 	rptr -= sizeof (uint32_t);
2336 	*(uint32_t *)rptr = htonl(frag_header);
2337 	mp->b_rptr = rptr;
2338 
2339 	mutex_enter(&mir->mir_mutex);
2340 	if (mir->mir_type == RPC_CLIENT) {
2341 		/*
2342 		 * For the client, set mir_clntreq to indicate that the
2343 		 * connection is active.
2344 		 */
2345 		mir->mir_clntreq = 1;
2346 		mir->mir_use_timestamp = ddi_get_lbolt();
2347 	}
2348 
2349 	/*
2350 	 * If we haven't already queued some data and the downstream module
2351 	 * can accept more data, send it on, otherwise we queue the message
2352 	 * and take other actions depending on mir_type.
2353 	 */
2354 	if (!mir->mir_inwservice && MIR_WCANPUTNEXT(mir, q)) {
2355 		mutex_exit(&mir->mir_mutex);
2356 
2357 		/*
2358 		 * Now we pass the RPC message downstream.
2359 		 */
2360 		putnext(q, mp);
2361 		return;
2362 	}
2363 
2364 	switch (mir->mir_type) {
2365 	case RPC_CLIENT:
2366 		/*
2367 		 * Check for a previous duplicate request on the
2368 		 * queue.  If there is one, then we throw away
2369 		 * the current message and let the previous one
2370 		 * go through.  If we can't find a duplicate, then
2371 		 * send this one.  This tap dance is an effort
2372 		 * to reduce traffic and processing requirements
2373 		 * under load conditions.
2374 		 */
2375 		if (mir_clnt_dup_request(q, mp)) {
2376 			mutex_exit(&mir->mir_mutex);
2377 			freemsg(mp);
2378 			return;
2379 		}
2380 		break;
2381 	case RPC_SERVER:
2382 		/*
2383 		 * Set mir_hold_inbound so that new inbound RPC
2384 		 * messages will be held until the client catches
2385 		 * up on the earlier replies.  This flag is cleared
2386 		 * in mir_wsrv after flow control is relieved;
2387 		 * the read-side queue is also enabled at that time.
2388 		 */
2389 		mir->mir_hold_inbound = 1;
2390 		break;
2391 	default:
2392 		RPCLOG(1, "mir_wput: unexpected mir_type %d\n", mir->mir_type);
2393 		break;
2394 	}
2395 	mir->mir_inwservice = 1;
2396 	(void) putq(q, mp);
2397 	mutex_exit(&mir->mir_mutex);
2398 }
2399 
2400 static void
2401 mir_wput_other(queue_t *q, mblk_t *mp)
2402 {
2403 	mir_t	*mir = (mir_t *)q->q_ptr;
2404 	struct iocblk	*iocp;
2405 	uchar_t	*rptr = mp->b_rptr;
2406 	bool_t	flush_in_svc = FALSE;
2407 
2408 	ASSERT(MUTEX_NOT_HELD(&mir->mir_mutex));
2409 	switch (mp->b_datap->db_type) {
2410 	case M_IOCTL:
2411 		iocp = (struct iocblk *)rptr;
2412 		switch (iocp->ioc_cmd) {
2413 		case RPC_CLIENT:
2414 			mutex_enter(&mir->mir_mutex);
2415 			if (mir->mir_type != 0 &&
2416 			    mir->mir_type != iocp->ioc_cmd) {
2417 ioc_eperm:
2418 				mutex_exit(&mir->mir_mutex);
2419 				iocp->ioc_error = EPERM;
2420 				iocp->ioc_count = 0;
2421 				mp->b_datap->db_type = M_IOCACK;
2422 				qreply(q, mp);
2423 				return;
2424 			}
2425 
2426 			mir->mir_type = iocp->ioc_cmd;
2427 
2428 			/*
2429 			 * Clear mir_hold_inbound which was set to 1 by
2430 			 * mir_open.  This flag is not used on client
2431 			 * streams.
2432 			 */
2433 			mir->mir_hold_inbound = 0;
2434 			mir->mir_max_msg_sizep = &clnt_max_msg_size;
2435 
2436 			/*
2437 			 * Start the idle timer.  See mir_timer() for more
2438 			 * information on how client timers work.
2439 			 */
2440 			mir->mir_idle_timeout = clnt_idle_timeout;
2441 			mir_clnt_idle_start(q, mir);
2442 			mutex_exit(&mir->mir_mutex);
2443 
2444 			mp->b_datap->db_type = M_IOCACK;
2445 			qreply(q, mp);
2446 			return;
2447 		case RPC_SERVER:
2448 			mutex_enter(&mir->mir_mutex);
2449 			if (mir->mir_type != 0 &&
2450 			    mir->mir_type != iocp->ioc_cmd)
2451 				goto ioc_eperm;
2452 
2453 			/*
2454 			 * We don't clear mir_hold_inbound here because
2455 			 * mir_hold_inbound is used in the flow control
2456 			 * model. If we cleared it here, then we'd commit
2457 			 * a small violation to the model where the transport
2458 			 * might immediately block downstream flow.
2459 			 */
2460 
2461 			mir->mir_type = iocp->ioc_cmd;
2462 			mir->mir_max_msg_sizep = &svc_max_msg_size;
2463 
2464 			/*
2465 			 * Start the idle timer.  See mir_timer() for more
2466 			 * information on how server timers work.
2467 			 *
2468 			 * Note that it is important to start the idle timer
2469 			 * here so that connections time out even if we
2470 			 * never receive any data on them.
2471 			 */
2472 			mir->mir_idle_timeout = svc_idle_timeout;
2473 			RPCLOG(16, "mir_wput_other starting idle timer on 0x%p "
2474 			    "because we got RPC_SERVER ioctl\n", (void *)q);
2475 			mir_svc_idle_start(q, mir);
2476 			mutex_exit(&mir->mir_mutex);
2477 
2478 			mp->b_datap->db_type = M_IOCACK;
2479 			qreply(q, mp);
2480 			return;
2481 		default:
2482 			break;
2483 		}
2484 		break;
2485 
2486 	case M_PROTO:
2487 		if (mir->mir_type == RPC_CLIENT) {
2488 			/*
2489 			 * We are likely being called from the context of a
2490 			 * service procedure. So we need to enqueue. However
2491 			 * enqueing may put our message behind data messages.
2492 			 * So flush the data first.
2493 			 */
2494 			flush_in_svc = TRUE;
2495 		}
2496 		if ((mp->b_wptr - rptr) < sizeof (uint32_t) ||
2497 		    !IS_P2ALIGNED(rptr, sizeof (uint32_t)))
2498 			break;
2499 
2500 		switch (((union T_primitives *)rptr)->type) {
2501 		case T_DATA_REQ:
2502 			/* Don't pass T_DATA_REQ messages downstream. */
2503 			freemsg(mp);
2504 			return;
2505 		case T_ORDREL_REQ:
2506 			RPCLOG(8, "mir_wput_other wq 0x%p: got T_ORDREL_REQ\n",
2507 			    (void *)q);
2508 			mutex_enter(&mir->mir_mutex);
2509 			if (mir->mir_type != RPC_SERVER) {
2510 				/*
2511 				 * We are likely being called from
2512 				 * clnt_dispatch_notifyall(). Sending
2513 				 * a T_ORDREL_REQ will result in
2514 				 * a some kind of _IND message being sent,
2515 				 * will be another call to
2516 				 * clnt_dispatch_notifyall(). To keep the stack
2517 				 * lean, queue this message.
2518 				 */
2519 				mir->mir_inwservice = 1;
2520 				(void) putq(q, mp);
2521 				mutex_exit(&mir->mir_mutex);
2522 				return;
2523 			}
2524 
2525 			/*
2526 			 * Mark the structure such that we don't accept any
2527 			 * more requests from client. We could defer this
2528 			 * until we actually send the orderly release
2529 			 * request downstream, but all that does is delay
2530 			 * the closing of this stream.
2531 			 */
2532 			RPCLOG(16, "mir_wput_other wq 0x%p: got T_ORDREL_REQ "
2533 			    " so calling mir_svc_start_close\n", (void *)q);
2534 
2535 			mir_svc_start_close(q, mir);
2536 
2537 			/*
2538 			 * If we have sent down a T_ORDREL_REQ, don't send
2539 			 * any more.
2540 			 */
2541 			if (mir->mir_ordrel_pending) {
2542 				freemsg(mp);
2543 				mutex_exit(&mir->mir_mutex);
2544 				return;
2545 			}
2546 
2547 			/*
2548 			 * If the stream is not idle, then we hold the
2549 			 * orderly release until it becomes idle.  This
2550 			 * ensures that kRPC will be able to reply to
2551 			 * all requests that we have passed to it.
2552 			 *
2553 			 * We also queue the request if there is data already
2554 			 * queued, because we cannot allow the T_ORDREL_REQ
2555 			 * to go before data. When we had a separate reply
2556 			 * count, this was not a problem, because the
2557 			 * reply count was reconciled when mir_wsrv()
2558 			 * completed.
2559 			 */
2560 			if (!MIR_SVC_QUIESCED(mir) ||
2561 			    mir->mir_inwservice == 1) {
2562 				mir->mir_inwservice = 1;
2563 				(void) putq(q, mp);
2564 
2565 				RPCLOG(16, "mir_wput_other: queuing "
2566 				    "T_ORDREL_REQ on 0x%p\n", (void *)q);
2567 
2568 				mutex_exit(&mir->mir_mutex);
2569 				return;
2570 			}
2571 
2572 			/*
2573 			 * Mark the structure so that we know we sent
2574 			 * an orderly release request, and reset the idle timer.
2575 			 */
2576 			mir->mir_ordrel_pending = 1;
2577 
2578 			RPCLOG(16, "mir_wput_other: calling mir_svc_idle_start"
2579 			    " on 0x%p because we got T_ORDREL_REQ\n",
2580 			    (void *)q);
2581 
2582 			mir_svc_idle_start(q, mir);
2583 			mutex_exit(&mir->mir_mutex);
2584 
2585 			/*
2586 			 * When we break, we will putnext the T_ORDREL_REQ.
2587 			 */
2588 			break;
2589 
2590 		case T_CONN_REQ:
2591 			mutex_enter(&mir->mir_mutex);
2592 			if (mir->mir_head_mp != NULL) {
2593 				freemsg(mir->mir_head_mp);
2594 				mir->mir_head_mp = NULL;
2595 				mir->mir_tail_mp = NULL;
2596 			}
2597 			mir->mir_frag_len = -(int32_t)sizeof (uint32_t);
2598 			/*
2599 			 * Restart timer in case mir_clnt_idle_do_stop() was
2600 			 * called.
2601 			 */
2602 			mir->mir_idle_timeout = clnt_idle_timeout;
2603 			mir_clnt_idle_stop(q, mir);
2604 			mir_clnt_idle_start(q, mir);
2605 			mutex_exit(&mir->mir_mutex);
2606 			break;
2607 
2608 		default:
2609 			/*
2610 			 * T_DISCON_REQ is one of the interesting default
2611 			 * cases here. Ideally, an M_FLUSH is done before
2612 			 * T_DISCON_REQ is done. However, that is somewhat
2613 			 * cumbersome for clnt_cots.c to do. So we queue
2614 			 * T_DISCON_REQ, and let the service procedure
2615 			 * flush all M_DATA.
2616 			 */
2617 			break;
2618 		}
2619 		/* fallthru */;
2620 	default:
2621 		if (mp->b_datap->db_type >= QPCTL) {
2622 			if (mp->b_datap->db_type == M_FLUSH) {
2623 				if (mir->mir_type == RPC_CLIENT &&
2624 				    *mp->b_rptr & FLUSHW) {
2625 					RPCLOG(32, "mir_wput_other: flushing "
2626 					    "wq 0x%p\n", (void *)q);
2627 					if (*mp->b_rptr & FLUSHBAND) {
2628 						flushband(q, *(mp->b_rptr + 1),
2629 						    FLUSHDATA);
2630 					} else {
2631 						flushq(q, FLUSHDATA);
2632 					}
2633 				} else {
2634 					RPCLOG(32, "mir_wput_other: ignoring "
2635 					    "M_FLUSH on wq 0x%p\n", (void *)q);
2636 				}
2637 			}
2638 			break;
2639 		}
2640 
2641 		mutex_enter(&mir->mir_mutex);
2642 		if (mir->mir_inwservice == 0 && MIR_WCANPUTNEXT(mir, q)) {
2643 			mutex_exit(&mir->mir_mutex);
2644 			break;
2645 		}
2646 		mir->mir_inwservice = 1;
2647 		mir->mir_inwflushdata = flush_in_svc;
2648 		(void) putq(q, mp);
2649 		mutex_exit(&mir->mir_mutex);
2650 		qenable(q);
2651 
2652 		return;
2653 	}
2654 	putnext(q, mp);
2655 }
2656 
2657 static void
2658 mir_wsrv(queue_t *q)
2659 {
2660 	mblk_t	*mp;
2661 	mir_t	*mir;
2662 	bool_t flushdata;
2663 
2664 	mir = (mir_t *)q->q_ptr;
2665 	mutex_enter(&mir->mir_mutex);
2666 
2667 	flushdata = mir->mir_inwflushdata;
2668 	mir->mir_inwflushdata = 0;
2669 
2670 	while (mp = getq(q)) {
2671 		if (mp->b_datap->db_type == M_DATA) {
2672 			/*
2673 			 * Do not send any more data if we have sent
2674 			 * a T_ORDREL_REQ.
2675 			 */
2676 			if (flushdata || mir->mir_ordrel_pending == 1) {
2677 				freemsg(mp);
2678 				continue;
2679 			}
2680 
2681 			/*
2682 			 * Make sure that the stream can really handle more
2683 			 * data.
2684 			 */
2685 			if (!MIR_WCANPUTNEXT(mir, q)) {
2686 				(void) putbq(q, mp);
2687 				mutex_exit(&mir->mir_mutex);
2688 				return;
2689 			}
2690 
2691 			/*
2692 			 * Now we pass the RPC message downstream.
2693 			 */
2694 			mutex_exit(&mir->mir_mutex);
2695 			putnext(q, mp);
2696 			mutex_enter(&mir->mir_mutex);
2697 			continue;
2698 		}
2699 
2700 		/*
2701 		 * This is not an RPC message, pass it downstream
2702 		 * (ignoring flow control) if the server side is not sending a
2703 		 * T_ORDREL_REQ downstream.
2704 		 */
2705 		if (mir->mir_type != RPC_SERVER ||
2706 		    ((union T_primitives *)mp->b_rptr)->type !=
2707 		    T_ORDREL_REQ) {
2708 			mutex_exit(&mir->mir_mutex);
2709 			putnext(q, mp);
2710 			mutex_enter(&mir->mir_mutex);
2711 			continue;
2712 		}
2713 
2714 		if (mir->mir_ordrel_pending == 1) {
2715 			/*
2716 			 * Don't send two T_ORDRELs
2717 			 */
2718 			freemsg(mp);
2719 			continue;
2720 		}
2721 
2722 		/*
2723 		 * Mark the structure so that we know we sent an orderly
2724 		 * release request.  We will check to see slot is idle at the
2725 		 * end of this routine, and if so, reset the idle timer to
2726 		 * handle orderly release timeouts.
2727 		 */
2728 		mir->mir_ordrel_pending = 1;
2729 		RPCLOG(16, "mir_wsrv: sending ordrel req on q 0x%p\n",
2730 		    (void *)q);
2731 		/*
2732 		 * Send the orderly release downstream. If there are other
2733 		 * pending replies we won't be able to send them.  However,
2734 		 * the only reason we should send the orderly release is if
2735 		 * we were idle, or if an unusual event occurred.
2736 		 */
2737 		mutex_exit(&mir->mir_mutex);
2738 		putnext(q, mp);
2739 		mutex_enter(&mir->mir_mutex);
2740 	}
2741 
2742 	if (q->q_first == NULL)
2743 		/*
2744 		 * If we call mir_svc_idle_start() below, then
2745 		 * clearing mir_inwservice here will also result in
2746 		 * any thread waiting in mir_close() to be signaled.
2747 		 */
2748 		mir->mir_inwservice = 0;
2749 
2750 	if (mir->mir_type != RPC_SERVER) {
2751 		mutex_exit(&mir->mir_mutex);
2752 		return;
2753 	}
2754 
2755 	/*
2756 	 * If idle we call mir_svc_idle_start to start the timer (or wakeup
2757 	 * a close). Also make sure not to start the idle timer on the
2758 	 * listener stream. This can cause nfsd to send an orderly release
2759 	 * command on the listener stream.
2760 	 */
2761 	if (MIR_SVC_QUIESCED(mir) && !(mir->mir_listen_stream)) {
2762 		RPCLOG(16, "mir_wsrv: calling mir_svc_idle_start on 0x%p "
2763 		    "because mir slot is idle\n", (void *)q);
2764 		mir_svc_idle_start(q, mir);
2765 	}
2766 
2767 	/*
2768 	 * If outbound flow control has been relieved, then allow new
2769 	 * inbound requests to be processed.
2770 	 */
2771 	if (mir->mir_hold_inbound) {
2772 		mir->mir_hold_inbound = 0;
2773 		qenable(RD(q));
2774 	}
2775 	mutex_exit(&mir->mir_mutex);
2776 }
2777 
2778 static void
2779 mir_disconnect(queue_t *q, mir_t *mir)
2780 {
2781 	ASSERT(MUTEX_HELD(&mir->mir_mutex));
2782 
2783 	switch (mir->mir_type) {
2784 	case RPC_CLIENT:
2785 		/*
2786 		 * We are disconnecting, but not necessarily
2787 		 * closing. By not closing, we will fail to
2788 		 * pick up a possibly changed global timeout value,
2789 		 * unless we store it now.
2790 		 */
2791 		mir->mir_idle_timeout = clnt_idle_timeout;
2792 		mir_clnt_idle_start(WR(q), mir);
2793 		mutex_exit(&mir->mir_mutex);
2794 
2795 		/*
2796 		 * T_DISCON_REQ is passed to kRPC as an integer value
2797 		 * (this is not a TPI message).  It is used as a
2798 		 * convenient value to indicate a sanity check
2799 		 * failure -- the same kRPC routine is also called
2800 		 * for T_DISCON_INDs and T_ORDREL_INDs.
2801 		 */
2802 		clnt_dispatch_notifyall(WR(q), T_DISCON_REQ, 0);
2803 		break;
2804 
2805 	case RPC_SERVER:
2806 		mir->mir_svc_no_more_msgs = 1;
2807 		mir_svc_idle_stop(WR(q), mir);
2808 		mutex_exit(&mir->mir_mutex);
2809 		RPCLOG(16, "mir_disconnect: telling "
2810 		    "stream head listener to disconnect stream "
2811 		    "(0x%p)\n", (void *) q);
2812 		(void) mir_svc_policy_notify(q, 2);
2813 		break;
2814 
2815 	default:
2816 		mutex_exit(&mir->mir_mutex);
2817 		break;
2818 	}
2819 }
2820 
2821 /*
2822  * Sanity check the message length, and if it's too large, shutdown the
2823  * connection.  Returns 1 if the connection is shutdown; 0 otherwise.
2824  */
2825 static int
2826 mir_check_len(queue_t *q, mblk_t *head_mp)
2827 {
2828 	mir_t *mir = q->q_ptr;
2829 	uint_t maxsize = 0;
2830 	size_t msg_len = msgdsize(head_mp);
2831 
2832 	if (mir->mir_max_msg_sizep != NULL)
2833 		maxsize = *mir->mir_max_msg_sizep;
2834 
2835 	if (maxsize == 0 || msg_len <= maxsize)
2836 		return (0);
2837 
2838 	freemsg(head_mp);
2839 	mir->mir_head_mp = NULL;
2840 	mir->mir_tail_mp = NULL;
2841 	mir->mir_frag_header = 0;
2842 	mir->mir_frag_len = -(int32_t)sizeof (uint32_t);
2843 	if (mir->mir_type != RPC_SERVER || mir->mir_setup_complete) {
2844 		cmn_err(CE_NOTE,
2845 		    "kRPC: record fragment from %s of size(%lu) exceeds "
2846 		    "maximum (%u). Disconnecting",
2847 		    (mir->mir_type == RPC_CLIENT) ? "server" :
2848 		    (mir->mir_type == RPC_SERVER) ? "client" :
2849 		    "test tool", msg_len, maxsize);
2850 	}
2851 
2852 	mir_disconnect(q, mir);
2853 	return (1);
2854 }
2855