xref: /titanic_41/usr/src/cmd/lvm/rpc.metad/metad_svc_subr.c (revision 354d1447ce995f3923a8f53d41c49fd3e6543282)
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, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #pragma ident	"%Z%%M%	%I%	%E% SMI"
28 
29 #include "metad_local.h"
30 #include <metad.h>
31 #include <sys/lvm/md_mddb.h>
32 #include <sdssc.h>
33 #include <sys/lvm/md_mirror.h>
34 #include <syslog.h>
35 
36 #include <sys/sysevent/eventdefs.h>
37 #include <sys/sysevent/svm.h>
38 #include <thread.h>
39 
40 #define	MDDOORS		"/usr/lib/lvm/mddoors"
41 
42 /*
43  * rpc.metad daemon
44  *
45  * The rpc.metad deamon supports two versions of the svm rpc calls - version 1
46  * and version 2. The over-the-wire structures sent as part of these rpc calls
47  * are also versioned - version 1 and version 2 exist. It must be noted that
48  * the version 2 structures have sub-versions or revisions as well. The
49  * revisions in the version 2 structures allow for flexiblility in changing
50  * over the wire structures without creating a new version of the svm rpc
51  * calls. No changes may be made to the version 1 routines or structures used
52  * by these routines.
53  *
54  * If, for example, the version 2 mdrpc_devinfo_args over the wire structure
55  * (mdrpc_devinfo_2_args*) is changed then the structure change must be
56  * accompanied by the following:
57  *
58  * Header file changes:
59  * . May need to introduce a new structure revision MD_METAD_ARGS_REV_X, where
60  *   X is the revision number.
61  * . Create mdrpc_devinfo_2_args_rX, where X is the new revision of the
62  *   structure.
63  * . Add a switch statement in mdrpc_devinfo_2_args.
64  *
65  * rpc.metad changes:
66  * . Check for the structure revision in the appropriate mdrpc_devinfo_svc
67  *   routine (mdrpc_devinfo_2_svc).
68  *
69  * libmeta changes:
70  * . In the libmeta code that makes the mdrpc_devinfo rpc call, the arguments
71  *   being passed as part of this call (namely mdrpc_devinfo_Y_args) must have
72  *   the revision field and associated structure populated correctly.
73  */
74 
75 static	md_setkey_t	*my_svc_sk = NULL;
76 
77 /*
78  * Add namespace entry to local mddb for using given sideno, key
79  * and names.
80  */
81 static int
82 add_sideno_sidenm(
83 	mdsidenames_t	*sidenms,
84 	mdkey_t		local_key,
85 	side_t		sideno,
86 	md_set_desc	*sd,		/* Only used with Version 2 */
87 	md_error_t	*ep
88 )
89 {
90 	mdsidenames_t	*sn;
91 	mdsetname_t	*local_sp;
92 	char		*nm;
93 
94 	if ((local_sp = metasetname(MD_LOCAL_NAME, ep)) == NULL)
95 		return (-1);
96 
97 	for (sn = sidenms; sn != NULL; sn = sn->next)
98 		if (sn->sideno == sideno)
99 			break;
100 
101 	assert(sn != NULL);
102 
103 
104 	/*
105 	 * SKEW will be used on the traditional diskset despite of the
106 	 * rpc version.  SKEW is not used on the multinode diskset
107 	 */
108 	if (MD_MNSET_DESC(sd)) {
109 		nm = meta_getnmbykey(MD_LOCAL_SET, sideno, local_key, ep);
110 	} else {
111 		nm = meta_getnmbykey(MD_LOCAL_SET, sideno+SKEW, local_key, ep);
112 	}
113 
114 	if (nm == NULL) {
115 		if (! mdisok(ep)) {
116 			if (! mdissyserror(ep, ENOENT))
117 				return (-1);
118 			mdclrerror(ep);
119 		}
120 
121 		/*
122 		 * Ignore returned key from add_name, only care about errs
123 		 *
124 		 * SKEW is used for a regular diskset since sideno could
125 		 * have a value of 0 in that diskset type.  add_name is
126 		 * writing to the local mddb and a sideno of 0 in the
127 		 * local mddb is reserved for non-diskset names.
128 		 * SKEW is added to the sideno in the local mddb so that
129 		 * the sideno for the diskset will never be 0.
130 		 *
131 		 * In a MNdiskset, the sideno will never be 0 (by design).
132 		 * So, no SKEW is needed when writing to the local mddb.
133 		 */
134 		if (MD_MNSET_DESC(sd)) {
135 			if (add_name(local_sp, sideno, local_key,
136 			    sn->dname, sn->mnum, sn->cname, ep) == -1)
137 				return (-1);
138 		} else {
139 			if (add_name(local_sp, sideno+SKEW, local_key,
140 			    sn->dname, sn->mnum, sn->cname, ep) == -1)
141 				return (-1);
142 		}
143 	} else
144 		Free(nm);
145 
146 	return (0);
147 }
148 
149 /*
150  * Delete sidename entry from local set using key and sideno.
151  */
152 static int
153 del_sideno_sidenm(
154 	mdkey_t		sidekey,
155 	side_t		sideno,
156 	md_error_t	*ep
157 )
158 {
159 	mdsetname_t	*local_sp;
160 
161 	if ((local_sp = metasetname(MD_LOCAL_NAME, ep)) == NULL)
162 		return (-1);
163 
164 	if (del_name(local_sp, sideno, sidekey, ep) == -1)
165 		mdclrerror(ep); /* ignore errs */
166 
167 	return (0);
168 }
169 
170 
171 /*
172  * Add namespace entries to local mddb for drives in drive list in
173  * set descriptor.
174  *
175  * If a MNset and if this host is being added to the set (this host
176  * is in the node_v list), add a namespace entry for the name of
177  * each drive using this host's sideno.
178  *
179  * If not a MNset, add namespace entries for all the new hosts being
180  * added to this set (list in node_v).
181  */
182 static void
183 add_drv_sidenms(
184 	char		*hostname,
185 	mdsetname_t	*sp,
186 	md_set_desc	*sd,
187 	int		node_c,
188 	char		**node_v,
189 	md_error_t	*ep
190 )
191 {
192 	mdsetname_t	*my_sp;
193 	md_drive_desc	*dd, *my_dd, *p, *q;
194 	mddrivename_t	*dn, *my_dn;
195 	int		i;
196 	side_t		sideno = 0, mysideno = 0;
197 	ddi_devid_t	devid_remote = NULL;
198 	ddi_devid_t	devid_local = NULL;
199 	int		devid_same = -1;
200 	int		using_devid = 0;
201 	md_mnnode_desc	*nd;
202 
203 	assert(sd->sd_drvs != NULL);
204 	dd = sd->sd_drvs;
205 
206 	if (dd->dd_dnp == NULL)
207 		return;
208 
209 	if ((my_sp = metasetname(sp->setname, ep)) == NULL)
210 		return;
211 	metaflushsetname(my_sp);
212 
213 	/* If a MN diskset */
214 	if (MD_MNSET_DESC(sd)) {
215 		/* Find sideno associated with RPC client. */
216 		nd = sd->sd_nodelist;
217 		while (nd) {
218 
219 			if (strcmp(nd->nd_nodename, hostname) == 0) {
220 				sideno = nd->nd_nodeid;
221 			}
222 
223 			/* While looping, find my side num as well */
224 			if (strcmp(nd->nd_nodename, mynode()) == 0) {
225 				mysideno = nd->nd_nodeid;
226 			}
227 
228 			if ((sideno) && (mysideno)) {
229 				break;
230 			}
231 			nd = nd->nd_next;
232 		}
233 
234 		if (!sideno) {
235 			(void) mddserror(ep, MDE_DS_HOSTNOSIDE,
236 			    sp->setno, hostname, NULL, sp->setname);
237 			return;
238 		}
239 	} else {
240 		/*
241 		 * if not a MN diskset
242 		 * do action for traditional diskset.
243 		 * despite of the rpc version
244 		 */
245 		for (sideno = 0; sideno < MD_MAXSIDES; sideno++) {
246 			/* Skip empty slots */
247 			if (sd->sd_nodes[sideno][0] == '\0')
248 				continue;
249 
250 			if (strcmp(hostname, sd->sd_nodes[sideno]) == 0)
251 				break;
252 		}
253 
254 		if (sideno == MD_MAXSIDES) {
255 			(void) mddserror(ep, MDE_DS_HOSTNOSIDE, sp->setno,
256 			    hostname, NULL, sp->setname);
257 			return;
258 		}
259 	}
260 	if ((my_dd = metaget_drivedesc_sideno(my_sp, sideno, MD_BASICNAME_OK,
261 	    ep)) == NULL) {
262 		if (! mdisok(ep))
263 			return;
264 		/* we are supposed to have drives!!!! */
265 		assert(0);
266 	}
267 
268 	/*
269 	 * The system is either all devid or all
270 	 * non-devid so we look at the first item
271 	 * in the list to determine if we're using devids or not.
272 	 * We also check to make sure it's not a multi-node diskset.
273 	 * If it is, we don't use devid's.
274 	 *
275 	 * For did disks, the dd_dnp->devid is a valid pointer which
276 	 * points to a '' string of devid.  We need to check this
277 	 * before set the using_devid.
278 	 */
279 	if ((dd->dd_dnp->devid != NULL) && (dd->dd_dnp->devid[0] != '\0') &&
280 	    (!(MD_MNSET_DESC(sd))))
281 		using_devid = 1;
282 
283 	/*
284 	 * We have to match-up the dd that were passed
285 	 * across the wire to the dd we have in this daemon.
286 	 * That way we can pick up the new sidenames that were
287 	 * passed to us and match them up with the local namespace key.
288 	 * Only we have the key, this cannot be passed in.
289 	 */
290 	for (p = dd; p != NULL; p = p->dd_next) {
291 		dn = p->dd_dnp;
292 		devid_remote = NULL;
293 
294 		if (dn->devid != NULL && (strlen(dn->devid) != 0) &&
295 		    using_devid) {
296 			/*
297 			 * We have a devid so use it
298 			 */
299 			(void) devid_str_decode(dn->devid, &devid_remote, NULL);
300 		}
301 
302 		/* check to make sure using_devid agrees with reality... */
303 		if ((using_devid == 1) && (devid_remote == NULL)) {
304 			/* something went really wrong. Can't process */
305 			(void) mddserror(ep, MDE_DS_INVALIDDEVID, sp->setno,
306 			    hostname, dn->cname, sp->setname);
307 			return;
308 		}
309 
310 		for (q = my_dd; q != NULL; q = q->dd_next) {
311 			my_dn = q->dd_dnp;
312 			devid_same = -1;
313 
314 			if (my_dn->devid != NULL && using_devid) {
315 				if (devid_str_decode(my_dn->devid,
316 				    &devid_local, NULL) == 0) {
317 					devid_same = devid_compare(devid_remote,
318 					    devid_local);
319 					devid_free(devid_local);
320 				}
321 			}
322 
323 			if (using_devid && devid_same == 0) {
324 				break;
325 			}
326 
327 			if (!using_devid &&
328 			    strcmp(my_dn->cname, dn->cname) == 0)
329 				break;
330 		}
331 
332 		if (devid_remote) {
333 			devid_free(devid_remote);
334 		}
335 		assert(q != NULL);
336 		assert(my_dn->side_names_key != MD_KEYWILD);
337 
338 		if (MD_MNSET_DESC(sd)) {
339 			/*
340 			 * Add the side names to the local db
341 			 * for this node only.
342 			 */
343 			if (add_sideno_sidenm(dn->side_names,
344 			    my_dn->side_names_key, mysideno, sd, ep))
345 				return;
346 			/*
347 			 * Sidenames for this drive were added
348 			 * to this host during the routine adddrvs.
349 			 * The sidenames that were added are the
350 			 * names associated with this drive on
351 			 * each of the hosts that were previously
352 			 * in the set.
353 			 * When the sidename for this drive on
354 			 * this host is added, the sidename
355 			 * from the host executing the command
356 			 * (not this host) is sent to this host.
357 			 * This host finds the originating host's
358 			 * sidename and can then determine this
359 			 * host's sidename.
360 			 * The sidenames from the other hosts serve
361 			 * only as temporary sidenames until this
362 			 * host's sidename can be added.
363 			 * In order to conserve space in the
364 			 * local mddb, the code now deletes the
365 			 * temporary sidenames added during adddrvs.
366 			 * When finished, only the sidename for this
367 			 * node should be left.
368 			 * Ignore any errors during this process since
369 			 * a failure to delete the extraneous
370 			 * sidenames shouldn't cause this routine
371 			 * to fail (in case that sidename didn't exist).
372 			 */
373 			nd = sd->sd_nodelist;
374 			while (nd) {
375 				if (nd->nd_nodeid != mysideno) {
376 					if (del_sideno_sidenm(
377 					    dn->side_names_key,
378 					    nd->nd_nodeid, ep) == -1)
379 						mdclrerror(ep);
380 				}
381 				nd = nd->nd_next;
382 			}
383 		} else {
384 			for (i = 0; i < MD_MAXSIDES; i++) {
385 				/* Skip empty slots */
386 				if (sd->sd_nodes[i][0] == '\0')
387 					continue;
388 
389 				/* Skip nodes not being added */
390 				if (! strinlst(sd->sd_nodes[i],
391 					node_c, node_v))
392 					continue;
393 
394 				/* Add the per side names to local db */
395 				if (add_sideno_sidenm(dn->side_names,
396 				    my_dn->side_names_key, i, sd, ep))
397 					return;
398 			}
399 		}
400 	}
401 }
402 
403 /* ARGSUSED */
404 bool_t
405 mdrpc_flush_internal_common(mdrpc_null_args *args, mdrpc_generic_res *res,
406     struct svc_req *rqstp)
407 {
408 	md_error_t	*ep = &res->status;
409 	int		err,
410 			op_mode = W_OK;
411 
412 	memset(res, 0, sizeof (*res));
413 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
414 		return (FALSE);
415 	else if (err != 0)
416 		return (TRUE);
417 
418 	metaflushnames(1);
419 
420 	err = svc_fini(ep);
421 
422 	return (TRUE);
423 }
424 
425 bool_t
426 mdrpc_flush_internal_1_svc(mdrpc_null_args *args, mdrpc_generic_res *res,
427     struct svc_req *rqstp)
428 {
429 	return (mdrpc_flush_internal_common(args, res, rqstp));
430 }
431 
432 bool_t
433 mdrpc_flush_internal_2_svc(mdrpc_null_args *args, mdrpc_generic_res *res,
434     struct svc_req *rqstp)
435 {
436 	return (mdrpc_flush_internal_common(args, res, rqstp));
437 }
438 
439 /*
440  * add 1 or more namespace entries per drive record.
441  * (into the local namespace)
442  */
443 bool_t
444 mdrpc_add_drv_sidenms_common(
445 	mdrpc_drv_sidenm_2_args_r1	*args,
446 	mdrpc_generic_res		*res,
447 	struct svc_req			*rqstp		/* RPC stuff */
448 )
449 {
450 	md_error_t		*ep = &res->status;
451 	int			err;
452 	int			op_mode = W_OK;
453 
454 	/* setup, check permissions */
455 	(void) memset(res, 0, sizeof (*res));
456 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
457 		return (FALSE);
458 	else if (err != 0)
459 		return (TRUE);
460 
461 	if (check_set_lock(op_mode, args->cl_sk, ep))
462 		return (TRUE);
463 
464 	/* doit */
465 	add_drv_sidenms(args->hostname, args->sp, args->sd,
466 	    args->node_v.node_v_len, args->node_v.node_v_val, ep);
467 
468 	err = svc_fini(ep);
469 
470 	return (TRUE);
471 }
472 
473 /*
474  * version 1 of the remote procedure. This procedure is called if the
475  * client is running in version 1. We first convert version 1 arguments
476  * into version 2 arguments and then call the common remote procedure.
477  */
478 bool_t
479 mdrpc_add_drv_sidenms_1_svc(
480 	mdrpc_drv_sidenm_args	*args,
481 	mdrpc_generic_res	*res,
482 	struct svc_req		*rqstp		/* RPC stuff */
483 )
484 {
485 	bool_t				retval;
486 	mdrpc_drv_sidenm_2_args_r1	v2_args;
487 	int				i, j;
488 
489 	/* allocate memory */
490 	v2_args.sd = Zalloc(sizeof (md_set_desc));
491 	alloc_newdrvdesc(args->sd->sd_drvs, &v2_args.sd->sd_drvs);
492 
493 	/* build args */
494 	v2_args.hostname = args->hostname;
495 	v2_args.cl_sk = args->cl_sk;
496 	v2_args.sp = args->sp;
497 	/* set descriptor */
498 	v2_args.sd->sd_ctime = args->sd->sd_ctime;
499 	v2_args.sd->sd_genid = args->sd->sd_genid;
500 	v2_args.sd->sd_setno = args->sd->sd_setno;
501 	v2_args.sd->sd_flags = args->sd->sd_flags;
502 	for (i = 0; i < MD_MAXSIDES; i++) {
503 		v2_args.sd->sd_isown[i] = args->sd->sd_isown[i];
504 
505 		for (j = 0; j < MD_MAX_NODENAME_PLUS_1; j++)
506 			v2_args.sd->sd_nodes[i][j] =
507 			    args->sd->sd_nodes[i][j];
508 	}
509 	v2_args.sd->sd_med = args->sd->sd_med;
510 	/* convert v1 args to v2 (revision 1) args */
511 	meta_conv_drvdesc_old2new(args->sd->sd_drvs, v2_args.sd->sd_drvs);
512 	v2_args.node_v.node_v_len = args->node_v.node_v_len;
513 	v2_args.node_v.node_v_val = args->node_v.node_v_val;
514 
515 	retval = mdrpc_add_drv_sidenms_common(&v2_args, res, rqstp);
516 
517 	free(v2_args.sd);
518 	free_newdrvdesc(v2_args.sd->sd_drvs);
519 
520 	return (retval);
521 }
522 
523 bool_t
524 mdrpc_add_drv_sidenms_2_svc(
525 	mdrpc_drv_sidenm_2_args	*args,
526 	mdrpc_generic_res	*res,
527 	struct svc_req		*rqstp		/* RPC stuff */
528 )
529 {
530 	switch (args->rev) {
531 	    case MD_METAD_ARGS_REV_1:
532 		return (mdrpc_add_drv_sidenms_common(
533 		    &args->mdrpc_drv_sidenm_2_args_u.rev1, res, rqstp));
534 	    default:
535 		return (FALSE);
536 	}
537 }
538 
539 static int
540 add_sidenamelist(
541 	mddrivename_t	*dn,
542 	side_t		thisside,
543 	md_set_record	*sr, 		/* used by RPC version 2 */
544 	md_error_t	*ep
545 )
546 {
547 	mdsidenames_t	*sn;
548 	mdkey_t		key;
549 	int		err;
550 	mdsetname_t	*local_sp;
551 	md_mnset_record	*mnsr;
552 	md_mnnode_record *nr;
553 	uint_t		nodeid = 0;
554 
555 	if ((local_sp = metasetname(MD_LOCAL_NAME, ep)) == NULL)
556 		return (-1);
557 
558 	key = MD_KEYWILD;
559 
560 	/*
561 	 * If a multi-node diskset, find nodeid associated with this node.
562 	 */
563 	if (MD_MNSET_REC(sr)) {
564 		mnsr = (struct md_mnset_record *)sr;
565 		nr = mnsr->sr_nodechain;
566 		while (nr) {
567 			if (strcmp(nr->nr_nodename, mynode()) == 0) {
568 				break;
569 			}
570 			nr = nr->nr_next;
571 		}
572 		/*
573 		 * If node is found, then a new drive is being added to
574 		 * a MN set of which this node is a member.
575 		 * If node is not found, then this host is being added to
576 		 * a MN set that has drives associated with it.
577 		 */
578 		if (nr)
579 			nodeid = nr->nr_nodeid;
580 	}
581 	for (sn = dn->side_names; sn != NULL; sn = sn->next) {
582 		if (MD_MNSET_REC(sr)) {
583 			/*
584 			 * In multi-node disksets, only add side information
585 			 * to the local mddb about this node.
586 			 * If the sideno for this node is found, then
587 			 * a new drive is being added to a MN set of
588 			 * which this node is a member.
589 			 * If the sideno for this node is not found, then
590 			 * this host is being added to a MNset that
591 			 * has drives associated with it.  In this case,
592 			 * need to add the sidename associated with the
593 			 * rpc client, but since we don't know which node
594 			 * is the client, then add temp entries for all sides.
595 			 * Later, the sidename for this node will be set
596 			 * via add_drv_sidenms and then the temp
597 			 * sidenames can be removed.
598 			 */
599 			if (nodeid == sn->sideno) {
600 				if ((err = add_name(local_sp, sn->sideno, key,
601 				    sn->dname, sn->mnum, sn->cname, ep)) == -1)
602 					return (-1);
603 				key = (mdkey_t)err;
604 				break;
605 			}
606 		} else {
607 			/*
608 			 * When a sidename is added into the namespace the local
609 			 * side information for the name is added first of all.
610 			 * When the first sidename is created this causes the
611 			 * devid of the disk to be recorded in the namespace, if
612 			 * the non-local side information is added first then
613 			 * there is the possibility of getting the wrong devid
614 			 * because there is no guarantee that the dev_t (mnum in
615 			 * this instance) is the same across all the nodes in
616 			 * the set. So the only way to make sure that the
617 			 * correct dev_t is used is to force the adding in of
618 			 * the local sidename record first of all. This same
619 			 * issue affects add_key_name().
620 			 */
621 			if (sn->sideno != thisside)
622 				continue;
623 			if ((err = add_name(local_sp, sn->sideno+SKEW, key,
624 			    sn->dname, sn->mnum, sn->cname, ep)) == -1)
625 				return (-1);
626 			key = (mdkey_t)err;
627 			break;
628 		}
629 	}
630 
631 	/*
632 	 * Now the other sides for non-MN set
633 	 */
634 	if (!MD_MNSET_REC(sr)) {
635 		for (sn = dn->side_names; sn != NULL; sn = sn->next) {
636 			if (sn->sideno == thisside)
637 				continue;
638 			if ((err = add_name(local_sp, sn->sideno+SKEW, key,
639 				sn->dname, sn->mnum, sn->cname, ep)) == -1)
640 				return (-1);
641 			key = (mdkey_t)err;
642 		}
643 	}
644 
645 	/* Temporarily add all sides. */
646 	if ((key == MD_KEYWILD) && (MD_MNSET_REC(sr))) {
647 		for (sn = dn->side_names; sn != NULL; sn = sn->next) {
648 			sn = dn->side_names;
649 			if (sn) {
650 				if ((err = add_name(local_sp, sn->sideno, key,
651 				    sn->dname, sn->mnum, sn->cname, ep)) == -1)
652 						return (-1);
653 				key = (mdkey_t)err;
654 			}
655 		}
656 	}
657 
658 	dn->side_names_key = key;
659 	return (0);
660 }
661 
662 static void
663 adddrvs(
664 	char 		*setname,
665 	md_drive_desc	*dd,
666 	md_timeval32_t	timestamp,
667 	ulong_t		genid,
668 	md_error_t	*ep
669 )
670 {
671 	mddb_userreq_t	req;
672 	md_drive_record	*dr;
673 	md_set_record	*sr;
674 	md_drive_desc	*p;
675 	mddrivename_t	*dn;
676 	mdname_t	*np;
677 	md_dev64_t	dev;
678 	md_error_t	xep = mdnullerror;
679 	int		i;
680 
681 	if ((sr = getsetbyname(setname, ep)) == NULL)
682 		return;
683 
684 	if (MD_MNSET_REC(sr))
685 		i = 0;
686 	else {
687 		/* get thisside */
688 		for (i = 0; i < MD_MAXSIDES; i++) {
689 			if (sr->sr_nodes[i][0] == '\0')
690 				continue;
691 			if (strcmp(mynode(), sr->sr_nodes[i]) == 0)
692 				break;
693 		}
694 
695 		if (i == MD_MAXSIDES) {
696 			/* so find the first free slot! */
697 			for (i = 0; i < MD_MAXSIDES; i++) {
698 				if (sr->sr_nodes[i][0] == '\0')
699 					break;
700 			}
701 		}
702 	}
703 
704 	for (p = dd; p != NULL; p = p->dd_next) {
705 		uint_t	rep_slice;
706 
707 		dn = p->dd_dnp;
708 
709 		/* Add the per side names to the local db */
710 		if (add_sidenamelist(dn, (side_t)i,  sr, ep)) {
711 				free_sr(sr);
712 				return;
713 		}
714 
715 		/* Create the drive record */
716 		(void) memset(&req, 0, sizeof (req));
717 		METAD_SETUP_DR(MD_DB_CREATE, 0);
718 		req.ur_size = sizeof (*dr);
719 		if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
720 			(void) mdstealerror(ep, &req.ur_mde);
721 			free_sr(sr);
722 			return;
723 		}
724 
725 		/* Fill in the drive record values */
726 		dr = Zalloc(sizeof (*dr));
727 		dr->dr_selfid = req.ur_recid;
728 		dr->dr_dbcnt = p->dd_dbcnt;
729 		dr->dr_dbsize = p->dd_dbsize;
730 		dr->dr_key = dn->side_names_key;
731 
732 		dr->dr_ctime = timestamp;
733 		dr->dr_genid = genid;
734 		dr->dr_revision = MD_DRIVE_RECORD_REVISION;
735 		dr->dr_flags = MD_DR_ADD;
736 
737 		/* Link the drive records and fill in in-core data */
738 		dr_cache_add(sr, dr);
739 
740 		dev = NODEV64;
741 		if ((meta_replicaslice(dn, &rep_slice, &xep) == 0) &&
742 		    ((np = metaslicename(dn, rep_slice, &xep)) != NULL))
743 			dev = np->dev;
744 		else
745 			mdclrerror(&xep);
746 
747 		SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_REMOVE, SVM_TAG_DRIVE,
748 		    MD_LOCAL_SET, dev);
749 		SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD, SVM_TAG_DRIVE,
750 		    sr->sr_setno, dev);
751 	}
752 
753 	/* Commit all the records atomically */
754 	commitset(sr, TRUE, ep);
755 	free_sr(sr);
756 }
757 
758 /*
759  * add 1 or more drive records to a set.
760  */
761 bool_t
762 mdrpc_adddrvs_common(
763 	mdrpc_drives_2_args_r1	*args,
764 	mdrpc_generic_res	*res,
765 	struct svc_req		*rqstp		/* RPC stuff */
766 )
767 {
768 	md_error_t		*ep = &res->status;
769 	int			err;
770 	int			op_mode = W_OK;
771 
772 	/* setup, check permissions */
773 	(void) memset(res, 0, sizeof (*res));
774 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
775 		return (FALSE);
776 	else if (err != 0)
777 		return (TRUE);
778 
779 	if (check_set_lock(op_mode, args->cl_sk, ep))
780 		return (TRUE);
781 
782 	/* doit */
783 	adddrvs(args->sp->setname, args->drivedescs, args->timestamp,
784 	    args->genid, ep);
785 
786 	err = svc_fini(ep);
787 
788 	return (TRUE);
789 }
790 
791 /*
792  * version 1 of the remote procedure. This procedure is called if the
793  * client is running in version 1. We first convert version 1 arguments
794  * into version 2 arguments and then call the common remote procedure.
795  */
796 bool_t
797 mdrpc_adddrvs_1_svc(
798 	mdrpc_drives_args	*args,
799 	mdrpc_generic_res	*res,
800 	struct svc_req		*rqstp		/* RPC stuff */
801 )
802 {
803 	bool_t			retval;
804 	mdrpc_drives_2_args_r1	v2_args;
805 
806 	/* allocate memory */
807 	alloc_newdrvdesc(args->drivedescs, &v2_args.drivedescs);
808 
809 	/* build args */
810 	v2_args.cl_sk = args->cl_sk;
811 	v2_args.sp = args->sp;
812 	/* convert v1 args to v2 (revision 1) args */
813 	meta_conv_drvdesc_old2new(args->drivedescs, v2_args.drivedescs);
814 	v2_args.timestamp = args->timestamp;
815 	v2_args.genid = args->genid;
816 
817 	retval = mdrpc_adddrvs_common(&v2_args, res, rqstp);
818 
819 	free_newdrvdesc(v2_args.drivedescs);
820 
821 	return (retval);
822 }
823 
824 bool_t
825 mdrpc_adddrvs_2_svc(
826 	mdrpc_drives_2_args	*args,
827 	mdrpc_generic_res	*res,
828 	struct svc_req		*rqstp		/* RPC stuff */
829 )
830 {
831 	switch (args->rev) {
832 	    case MD_METAD_ARGS_REV_1:
833 		return (mdrpc_adddrvs_common(
834 		    &args->mdrpc_drives_2_args_u.rev1, res, rqstp));
835 	    default:
836 		return (FALSE);
837 	}
838 }
839 
840 static void
841 addhosts(
842 	char		*setname,
843 	int		node_c,
844 	char		**node_v,
845 	int		version,	/* RPC version of calling routine */
846 	md_error_t	*ep
847 )
848 {
849 	mddb_userreq_t		req;
850 	md_set_record		*sr;
851 	int			i, j;
852 	md_mnset_record		*mnsr;
853 	md_mnnode_record	*nr;
854 	mddb_set_node_params_t	snp;
855 	int			nodecnt;
856 	mndiskset_membershiplist_t *nl, *nl2;
857 
858 	if ((sr = getsetbyname(setname, ep)) == NULL)
859 		return;
860 
861 	/* Do MN operation if rpc version supports it and if a MN set */
862 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
863 		mnsr = (md_mnset_record *)sr;
864 		/*
865 		 * Verify nodes are in membership list on THIS node.
866 		 * Initiating node has verified that nodes are in membership
867 		 * list on the initiating node.
868 		 * Get membershiplist from API routine.  If there's
869 		 * an error, fail to add hosts and pass back error.
870 		 */
871 		if (meta_read_nodelist(&nodecnt, &nl, ep) == -1) {
872 			free_sr(sr);
873 			return;
874 		}
875 		/* Verify that all nodes are in member list */
876 		for (i = 0; i < node_c; i++) {
877 			/*
878 			 * If node in list isn't a member of the membership,
879 			 * just return error.
880 			 */
881 			if (meta_is_member(node_v[i], NULL, nl) == 0) {
882 				meta_free_nodelist(nl);
883 				(void) mddserror(ep, MDE_DS_NOTINMEMBERLIST,
884 				    sr->sr_setno, node_v[i], NULL, setname);
885 				free_sr(sr);
886 				return;
887 			}
888 		}
889 	}
890 
891 	for (i = 0; i < node_c; i++) {
892 		/* Do MN operation if rpc version supports it and if a MN set */
893 		if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
894 			mnsr = (md_mnset_record *)sr;
895 			/* Create the node record */
896 			(void) memset(&req, 0, sizeof (req));
897 			METAD_SETUP_NR(MD_DB_CREATE, 0);
898 			req.ur_size = sizeof (*nr);
899 			if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL)
900 			    != 0) {
901 				(void) mdstealerror(ep, &req.ur_mde);
902 				meta_free_nodelist(nl);
903 				free_sr(sr);
904 				return;
905 			}
906 
907 			nr = Zalloc(sizeof (*nr));
908 			nr->nr_revision = MD_MNNODE_RECORD_REVISION;
909 			nr->nr_selfid = req.ur_recid;
910 			nr->nr_ctime = sr->sr_ctime;
911 			nr->nr_genid = sr->sr_genid;
912 			nr->nr_flags = MD_MN_NODE_ADD;
913 			nl2 = nl;
914 			while (nl2) {
915 				if (strcmp(nl2->msl_node_name, node_v[i])
916 				    == 0) {
917 					nr->nr_nodeid = nl2->msl_node_id;
918 					break;
919 				}
920 				nl2 = nl2->next;
921 			}
922 
923 			(void) strcpy(nr->nr_nodename, node_v[i]);
924 
925 			/*
926 			 * When a node is added to a MN diskset, set the
927 			 * nodeid of this node in the md_set structure
928 			 * in the kernel.
929 			 */
930 			if (strcmp(nr->nr_nodename, mynode()) == 0) {
931 				(void) memset(&snp, 0, sizeof (snp));
932 				snp.sn_nodeid = nr->nr_nodeid;
933 				snp.sn_setno = mnsr->sr_setno;
934 				if (metaioctl(MD_MN_SET_NODEID, &snp,
935 				    &snp.sn_mde, NULL) != 0) {
936 					(void) mdstealerror(ep, &snp.sn_mde);
937 					meta_free_nodelist(nl);
938 					free_sr(sr);
939 					return;
940 				}
941 			}
942 
943 			/* Link the node records and fill in in-core data */
944 			mnnr_cache_add(mnsr, nr);
945 
946 			SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD, SVM_TAG_HOST,
947 				mnsr->sr_setno, nr->nr_nodeid);
948 		} else {
949 			for (j = 0; j < MD_MAXSIDES; j++) {
950 				if (sr->sr_nodes[j][0] != '\0')
951 					continue;
952 				(void) strcpy(sr->sr_nodes[j], node_v[i]);
953 				SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD,
954 				    SVM_TAG_HOST, sr->sr_setno, j);
955 				break;
956 			}
957 		}
958 	}
959 	/* Do MN operation if rpc version supports it and if a MN set */
960 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
961 		meta_free_nodelist(nl);
962 	}
963 
964 	(void) memset(&req, '\0', sizeof (req));
965 
966 	METAD_SETUP_SR(MD_DB_SETDATA, sr->sr_selfid)
967 
968 	/* Do MN operation if rpc version supports it and if a MN set */
969 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
970 		req.ur_size = sizeof (*mnsr);
971 	} else {
972 		req.ur_size = sizeof (*sr);
973 	}
974 	req.ur_data = (uintptr_t)sr;
975 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
976 		(void) mdstealerror(ep, &req.ur_mde);
977 		free_sr(sr);
978 		return;
979 	}
980 
981 	commitset(sr, TRUE, ep);
982 
983 	free_sr(sr);
984 }
985 
986 /*
987  * add 1 or more hosts to a set.
988  */
989 bool_t
990 mdrpc_addhosts_common(
991 	mdrpc_host_args		*args,
992 	mdrpc_generic_res	*res,
993 	struct svc_req		*rqstp,		/* RPC stuff */
994 	int			version		/* RPC version */
995 )
996 {
997 	md_error_t		*ep = &res->status;
998 	int			err;
999 	int			op_mode = W_OK;
1000 
1001 	/* setup, check permissions */
1002 	(void) memset(res, 0, sizeof (*res));
1003 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
1004 		return (FALSE);
1005 	else if (err != 0)
1006 		return (TRUE);
1007 
1008 	if (check_set_lock(op_mode, args->cl_sk, ep))
1009 		return (TRUE);
1010 
1011 	/* doit */
1012 	addhosts(args->sp->setname, args->hosts.hosts_len,
1013 	    args->hosts.hosts_val, version, ep);
1014 
1015 	err = svc_fini(ep);
1016 
1017 	return (TRUE);
1018 }
1019 
1020 bool_t
1021 mdrpc_addhosts_1_svc(
1022 	mdrpc_host_args		*args,
1023 	mdrpc_generic_res	*res,
1024 	struct svc_req		*rqstp		/* RPC stuff */
1025 )
1026 {
1027 	/* Pass RPC version (METAD_VERSION) to common routine */
1028 	return (mdrpc_addhosts_common(args, res, rqstp, METAD_VERSION));
1029 }
1030 
1031 bool_t
1032 mdrpc_addhosts_2_svc(
1033 	mdrpc_host_2_args	*args,
1034 	mdrpc_generic_res	*res,
1035 	struct svc_req		*rqstp		/* RPC stuff */
1036 )
1037 {
1038 	switch (args->rev) {
1039 	    case MD_METAD_ARGS_REV_1:
1040 		/* Pass RPC version (METAD_VERSION_DEVID) to common routine */
1041 		return (mdrpc_addhosts_common(
1042 		    &args->mdrpc_host_2_args_u.rev1, res,
1043 		    rqstp, METAD_VERSION_DEVID));
1044 	    default:
1045 		return (FALSE);
1046 	}
1047 }
1048 
1049 static void
1050 createset(
1051 	mdsetname_t		*sp,
1052 	md_node_nm_arr_t	nodes,
1053 	md_timeval32_t		timestamp,
1054 	ulong_t			genid,
1055 	md_error_t		*ep
1056 )
1057 {
1058 	mddb_userreq_t		req;
1059 	md_set_record		*sr;
1060 	int			i;
1061 
1062 	(void) memset(&req, 0, sizeof (req));
1063 	METAD_SETUP_SR(MD_DB_CREATE, 0);
1064 	req.ur_size = sizeof (*sr);
1065 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
1066 		(void) mdstealerror(ep, &req.ur_mde);
1067 		return;
1068 	}
1069 
1070 	sr = Zalloc(sizeof (*sr));
1071 
1072 	sr->sr_selfid = req.ur_recid;
1073 	sr->sr_setno = sp->setno;
1074 	(void) strcpy(sr->sr_setname, sp->setname);
1075 
1076 	SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_CREATE, SVM_TAG_SET, sp->setno,
1077 	    NODEV64);
1078 
1079 	(void) meta_smf_enable(META_SMF_DISKSET, NULL);
1080 
1081 	for (i = 0; i < MD_MAXSIDES; i++) {
1082 		(void) strcpy(sr->sr_nodes[i], nodes[i]);
1083 		/* Skip empty slots */
1084 		if (sr->sr_nodes[i][0] == '\0')
1085 			continue;
1086 		SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD, SVM_TAG_HOST, sp->setno,
1087 		    i);
1088 	}
1089 
1090 	sr->sr_ctime = timestamp;
1091 	sr->sr_genid = genid;
1092 	sr->sr_revision = MD_SET_RECORD_REVISION;
1093 	sr->sr_flags |= MD_SR_ADD;
1094 
1095 	sr->sr_mhiargs = defmhiargs;
1096 
1097 	sr_cache_add(sr);
1098 
1099 	commitset(sr, TRUE, ep);
1100 }
1101 
1102 static void
1103 mncreateset(
1104 	mdsetname_t		*sp,
1105 	md_mnnode_desc		*nodelist,
1106 	md_timeval32_t		timestamp,
1107 	ulong_t			genid,
1108 	md_node_nm_t		master_nodenm,
1109 	int			master_nodeid,
1110 	md_error_t		*ep
1111 )
1112 {
1113 	mddb_userreq_t			req;
1114 	md_mnset_record			*mnsr;
1115 	md_mnnode_record		*nr;
1116 	md_mnnode_desc			*nd;
1117 	mddb_set_node_params_t		snp;
1118 	int				nodecnt;
1119 	mndiskset_membershiplist_t	*nl;
1120 
1121 	/*
1122 	 * Validate that nodes in set being created are in the
1123 	 * membership list on THIS node.
1124 	 * Initiating node has verified that nodes are in membership
1125 	 * list on the initiating node.
1126 	 * Get membershiplist from API routine.  If there's
1127 	 * an error, fail to add set and pass back error.
1128 	 */
1129 	if (meta_read_nodelist(&nodecnt, &nl, ep) == -1) {
1130 		return;
1131 	}
1132 	/* Verify that all nodes are in member list */
1133 	nd = nodelist;
1134 	while (nd) {
1135 		/*
1136 		 * If node in list isn't a member of the membership,
1137 		 * just return error.
1138 		 */
1139 		if (meta_is_member(nd->nd_nodename, 0, nl) == 0) {
1140 			meta_free_nodelist(nl);
1141 			(void) mddserror(ep, MDE_DS_NOTINMEMBERLIST,
1142 			    sp->setno, nd->nd_nodename, NULL, sp->setname);
1143 			return;
1144 		}
1145 		nd = nd->nd_next;
1146 	}
1147 	meta_free_nodelist(nl);
1148 
1149 	(void) memset(&req, 0, sizeof (req));
1150 	METAD_SETUP_SR(MD_DB_CREATE, 0);
1151 	req.ur_size = sizeof (*mnsr);
1152 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
1153 		(void) mdstealerror(ep, &req.ur_mde);
1154 		return;
1155 	}
1156 
1157 	mnsr = Zalloc(sizeof (*mnsr));
1158 	mnsr->sr_selfid = req.ur_recid;
1159 	mnsr->sr_setno = sp->setno;
1160 	(void) strlcpy(mnsr->sr_setname, sp->setname, MD_MAX_SETNAME);
1161 
1162 	SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_CREATE, SVM_TAG_SET, sp->setno,
1163 	    NODEV64);
1164 
1165 	(void) meta_smf_enable(META_SMF_DISKSET | META_SMF_MN_DISKSET, NULL);
1166 
1167 	nd = nodelist;
1168 	while (nd) {
1169 		/* Create the node record */
1170 		(void) memset(&req, 0, sizeof (req));
1171 		METAD_SETUP_NR(MD_DB_CREATE, 0);
1172 		req.ur_size = sizeof (*nr);
1173 		if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
1174 			/* Frees mnsr and any alloc'd node records */
1175 			free_sr((struct md_set_record *)mnsr);
1176 			(void) mdstealerror(ep, &req.ur_mde);
1177 			return;
1178 		}
1179 
1180 		nr = Zalloc(sizeof (*nr));
1181 		nr->nr_revision = MD_MNNODE_RECORD_REVISION;
1182 		nr->nr_selfid = req.ur_recid;
1183 		nr->nr_ctime = timestamp;
1184 		nr->nr_genid = genid;
1185 		nr->nr_nodeid = nd->nd_nodeid;
1186 		nr->nr_flags = nd->nd_flags;
1187 		(void) strlcpy(nr->nr_nodename, nd->nd_nodename,
1188 			MD_MAX_NODENAME);
1189 
1190 		/* Link the node records and fill in in-core data */
1191 		mnnr_cache_add(mnsr, nr);
1192 
1193 		SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD, SVM_TAG_HOST, sp->setno,
1194 		    nr->nr_nodeid);
1195 
1196 		nd = nd->nd_next;
1197 	}
1198 
1199 	/*
1200 	 * For backward compatibility, fill in mynode name
1201 	 * as the only name in the sr_nodes array.  This
1202 	 * allows the pre-MNdiskset code to see that there
1203 	 * is a node in this diskset.  This will keep the
1204 	 * pre-MNdiskset code from removing this set.
1205 	 */
1206 	(void) strlcpy(mnsr->sr_nodes_bw_compat[0], mynode(), MD_MAX_NODENAME);
1207 
1208 	mnsr->sr_ctime = timestamp;
1209 	mnsr->sr_genid = genid;
1210 	mnsr->sr_revision = MD_SET_RECORD_REVISION;
1211 	mnsr->sr_flags |= MD_SR_ADD;
1212 
1213 	mnsr->sr_flags |= MD_SR_MN;
1214 	strcpy(mnsr->sr_master_nodenm, master_nodenm);
1215 	mnsr->sr_master_nodeid = master_nodeid;
1216 
1217 	mnsr->sr_mhiargs = defmhiargs;
1218 
1219 	sr_cache_add((struct md_set_record *)mnsr);
1220 
1221 	commitset((struct md_set_record *)mnsr, TRUE, ep);
1222 
1223 	/*
1224 	 * When a set is created for the first time, the nodelist
1225 	 * will contain this node.
1226 	 * When a node is just being added to a set, the nodelist
1227 	 * will not contain this node.  This node is added to the
1228 	 * set structure with a later call to addhosts.
1229 	 *
1230 	 * So, if the nodelist contains an entry for this node
1231 	 * then set the nodeid of this node in the md_set kernel
1232 	 * data structure.
1233 	 */
1234 	nd = nodelist;
1235 	while (nd) {
1236 		if (strcmp(nd->nd_nodename, mynode()) == 0) {
1237 			break;
1238 		}
1239 		nd = nd->nd_next;
1240 	}
1241 	if (nd) {
1242 		(void) memset(&snp, 0, sizeof (snp));
1243 		snp.sn_nodeid = nd->nd_nodeid;
1244 		snp.sn_setno = sp->setno;
1245 		if (metaioctl(MD_MN_SET_NODEID, &snp, &snp.sn_mde, NULL) != 0) {
1246 			(void) mdstealerror(ep, &snp.sn_mde);
1247 			return;
1248 		}
1249 	}
1250 }
1251 
1252 /*
1253  * create a set on a host
1254  */
1255 bool_t
1256 mdrpc_createset_common(
1257 	mdrpc_createset_args	*args,
1258 	mdrpc_generic_res	*res,
1259 	struct svc_req		*rqstp		/* RPC stuff */
1260 )
1261 {
1262 	md_error_t		*ep = &res->status;
1263 	char			stringbuf1[MAXPATHLEN];
1264 	char			stringbuf2[MAXPATHLEN];
1265 	int			err;
1266 	int			op_mode = W_OK;
1267 
1268 	/* setup, check permissions */
1269 	(void) memset(res, 0, sizeof (*res));
1270 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
1271 		return (FALSE);
1272 	else if (err != 0)
1273 		return (TRUE);
1274 
1275 	if (check_set_lock(op_mode, args->cl_sk, ep))
1276 		return (TRUE);
1277 
1278 	/* create the arguments for the symlink() and unlink() calls */
1279 	(void) snprintf(stringbuf2, sizeof (stringbuf2), "/dev/md/%s",
1280 	    args->sp->setname);
1281 	(void) snprintf(stringbuf1, sizeof (stringbuf1), "shared/%d",
1282 	    args->sp->setno);
1283 
1284 	/*
1285 	 * Since we already verified that the setname was OK, make sure to
1286 	 * cleanup before proceeding.
1287 	 */
1288 	if (unlink(stringbuf2) == -1) {
1289 		if (errno != ENOENT) {
1290 			(void) mdsyserror(ep, errno, stringbuf2);
1291 			return (TRUE);
1292 		}
1293 	}
1294 
1295 	/* create the set */
1296 	createset(args->sp, args->nodes, args->timestamp, args->genid, ep);
1297 
1298 	if (! mdisok(ep))
1299 		return (TRUE);
1300 
1301 	/* create the symlink */
1302 	if (symlink(stringbuf1, stringbuf2) == -1)
1303 		(void) mdsyserror(ep, errno, stringbuf2);
1304 
1305 	err = svc_fini(ep);
1306 
1307 	return (TRUE);
1308 }
1309 
1310 bool_t
1311 mdrpc_createset_1_svc(
1312 	mdrpc_createset_args	*args,
1313 	mdrpc_generic_res	*res,
1314 	struct svc_req		*rqstp		/* RPC stuff */
1315 )
1316 {
1317 	return (mdrpc_createset_common(args, res, rqstp));
1318 }
1319 
1320 bool_t
1321 mdrpc_createset_2_svc(
1322 	mdrpc_createset_2_args	*args,
1323 	mdrpc_generic_res	*res,
1324 	struct svc_req		*rqstp		/* RPC stuff */
1325 )
1326 {
1327 	switch (args->rev) {
1328 	    case MD_METAD_ARGS_REV_1:
1329 		return (mdrpc_createset_common(
1330 		    &args->mdrpc_createset_2_args_u.rev1, res, rqstp));
1331 	    default:
1332 		return (FALSE);
1333 	}
1334 }
1335 
1336 bool_t
1337 mdrpc_mncreateset_common(
1338 	mdrpc_mncreateset_args	*args,
1339 	mdrpc_generic_res	*res,
1340 	struct svc_req		*rqstp		/* RPC stuff */
1341 )
1342 {
1343 	md_error_t		*ep = &res->status;
1344 	char			stringbuf1[MAXPATHLEN];
1345 	char			stringbuf2[MAXPATHLEN];
1346 	int			err;
1347 	int			op_mode = W_OK;
1348 
1349 	/* setup, check permissions */
1350 	(void) memset(res, 0, sizeof (*res));
1351 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
1352 		return (FALSE);
1353 	else if (err != 0)
1354 		return (TRUE);
1355 
1356 	if (check_set_lock(op_mode, args->cl_sk, ep))
1357 		return (TRUE);
1358 
1359 	/* create the arguments for the symlink() and unlink() calls */
1360 	(void) snprintf(stringbuf2, sizeof (stringbuf2), "/dev/md/%s",
1361 	    args->sp->setname);
1362 	(void) snprintf(stringbuf1, sizeof (stringbuf1), "shared/%d",
1363 	    args->sp->setno);
1364 
1365 	/*
1366 	 * Since we already verified that the setname was OK, make sure to
1367 	 * cleanup before proceeding.
1368 	 */
1369 	if (unlink(stringbuf2) == -1) {
1370 		if (errno != ENOENT) {
1371 			(void) mdsyserror(ep, errno, stringbuf2);
1372 			return (TRUE);
1373 		}
1374 	}
1375 
1376 	/* create the set */
1377 	mncreateset(args->sp, args->nodelist, args->timestamp, args->genid,
1378 		    args->master_nodenm, args->master_nodeid, ep);
1379 
1380 	if (! mdisok(ep)) {
1381 		return (TRUE);
1382 	}
1383 
1384 	/* create the symlink */
1385 	if (symlink(stringbuf1, stringbuf2) == -1)
1386 		(void) mdsyserror(ep, errno, stringbuf2);
1387 
1388 	err = svc_fini(ep);
1389 
1390 	return (TRUE);
1391 }
1392 
1393 bool_t
1394 mdrpc_mncreateset_2_svc(
1395 	mdrpc_mncreateset_2_args	*args,
1396 	mdrpc_generic_res	*res,
1397 	struct svc_req		*rqstp		/* RPC stuff */
1398 )
1399 {
1400 	switch (args->rev) {
1401 	    case MD_METAD_ARGS_REV_1:
1402 		return (mdrpc_mncreateset_common(
1403 		    &args->mdrpc_mncreateset_2_args_u.rev1, res, rqstp));
1404 	    default:
1405 		return (FALSE);
1406 	}
1407 }
1408 
1409 static void
1410 del_drv_sidenms(
1411 	mdsetname_t	*sp,
1412 	int		version,	/* RPC version of calling routine */
1413 	md_error_t	*ep
1414 )
1415 {
1416 	md_set_record	*sr;
1417 	md_drive_desc	*dd, *p;
1418 	mddrivename_t	*dn;
1419 	mdsetname_t	*local_sp;
1420 	int		i;
1421 	int		rb_mode = 0;
1422 
1423 	if ((local_sp = metasetname(MD_LOCAL_NAME, ep)) == NULL)
1424 		return;
1425 
1426 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
1427 		return;
1428 
1429 	/* Do MN operation if rpc version supports it and if a MN set */
1430 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
1431 		/*
1432 		 * In the multi-node diskset, there are no diskset
1433 		 * entries in the local set for other nodes, so there's
1434 		 * nothing to do.
1435 		 */
1436 		free_sr(sr);
1437 		return;
1438 	}
1439 
1440 	if ((dd = metaget_drivedesc(sp, (MD_BASICNAME_OK | PRINT_FAST),
1441 	    ep)) == NULL) {
1442 		if (! mdisdserror(ep, MDE_DS_HOSTNOSIDE)) {
1443 			metaflushsetname(sp);
1444 			if (! mdisok(ep)) {
1445 				free_sr(sr);
1446 				return;
1447 			}
1448 			/* we are supposed to have drives!!!! */
1449 			assert(0);
1450 		}
1451 		rb_mode = 1;
1452 		mdclrerror(ep);
1453 		for (i = 0; i < MD_MAXSIDES; i++) {
1454 			/* Skip empty sides of the diskset */
1455 			if (sr->sr_nodes[i][0] == '\0')
1456 				continue;
1457 			dd = metaget_drivedesc_sideno(sp, i,
1458 			    (MD_BASICNAME_OK | PRINT_FAST), ep);
1459 			/* Got dd, get out of loop */
1460 			if (dd != NULL)
1461 				break;
1462 
1463 			/* some error occurred, get out of loop */
1464 			if (! mdisok(ep))
1465 				break;
1466 		}
1467 		/*
1468 		 * At this point, we have one of three possibilities:
1469 		 *	1) dd != NULL (we have found drives using an alternate
1470 		 *	   side.)
1471 		 *	2) dd == NULL (no drives) && mdisok(ep) : assert(0)
1472 		 *	3) dd == NULL (no drives) && ! mdisok(ep) : return
1473 		 *	   error information to caller.
1474 		 */
1475 		if (dd == NULL) {
1476 			metaflushsetname(sp);
1477 			if (! mdisok(ep)) {
1478 				free_sr(sr);
1479 				return;
1480 			}
1481 			/* we are supposed to have drives!!!! */
1482 			assert(0);
1483 		}
1484 	}
1485 
1486 	/*
1487 	 * Let's run through each drive descriptor, and delete the
1488 	 * sidename for all sides that are not in the sr_nodes array.
1489 	 * We will ignore errors, cause the empty side may not
1490 	 * have had any names to begin with.
1491 	 */
1492 	for (p = dd; p != NULL; p = p->dd_next) {
1493 		dn = p->dd_dnp;
1494 
1495 		for (i = 0; i < MD_MAXSIDES; i++) {
1496 			/* Skip existing sides of the diskset */
1497 			if (!rb_mode && sr->sr_nodes[i][0] != '\0')
1498 				continue;
1499 			/* An empty side, delete the sidename */
1500 			if (del_name(local_sp, i+SKEW,
1501 			    dn->side_names_key, ep)) {
1502 				if (!mdissyserror(ep, ENOENT)) {
1503 					free_sr(sr);
1504 					return;
1505 				}
1506 				mdclrerror(ep);
1507 			}
1508 		}
1509 	}
1510 	free_sr(sr);
1511 	metaflushsetname(sp);
1512 }
1513 
1514 /*
1515  * delete 1 or more sidenames per drive desc, from the local namespace
1516  */
1517 bool_t
1518 mdrpc_del_drv_sidenms_common(
1519 	mdrpc_sp_args		*args,
1520 	mdrpc_generic_res	*res,
1521 	struct svc_req		*rqstp,		/* RPC stuff */
1522 	int			version		/* RPC version */
1523 )
1524 {
1525 	md_error_t		*ep = &res->status;
1526 	int			err;
1527 	int			op_mode = W_OK;
1528 
1529 	/* setup, check permissions */
1530 	(void) memset(res, 0, sizeof (*res));
1531 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
1532 		return (FALSE);
1533 	else if (err != 0)
1534 		return (TRUE);
1535 
1536 	if (check_set_lock(op_mode, args->cl_sk, ep))
1537 		return (TRUE);
1538 
1539 	/* doit */
1540 	del_drv_sidenms(args->sp, version, ep);
1541 
1542 	err = svc_fini(ep);
1543 
1544 	return (TRUE);
1545 }
1546 
1547 bool_t
1548 mdrpc_del_drv_sidenms_1_svc(
1549 	mdrpc_sp_args		*args,
1550 	mdrpc_generic_res	*res,
1551 	struct svc_req		*rqstp		/* RPC stuff */
1552 )
1553 {
1554 	/* Pass RPC version (METAD_VERSION) to common routine */
1555 	return (mdrpc_del_drv_sidenms_common(args, res, rqstp, METAD_VERSION));
1556 }
1557 
1558 bool_t
1559 mdrpc_del_drv_sidenms_2_svc(
1560 	mdrpc_sp_2_args		*args,
1561 	mdrpc_generic_res	*res,
1562 	struct svc_req		*rqstp		/* RPC stuff */
1563 )
1564 {
1565 	switch (args->rev) {
1566 	    case MD_METAD_ARGS_REV_1:
1567 		/* Pass RPC version (METAD_VERSION_DEVID) to common routine */
1568 		return (mdrpc_del_drv_sidenms_common(
1569 		    &args->mdrpc_sp_2_args_u.rev1, res,
1570 		    rqstp, METAD_VERSION_DEVID));
1571 	    default:
1572 		return (FALSE);
1573 	}
1574 }
1575 
1576 static int
1577 del_sidenamelist(
1578 	md_set_record	*sr,
1579 	mddrivename_t	*dn,
1580 	md_error_t	*ep
1581 )
1582 {
1583 	mdsidenames_t	*sn;
1584 	mdsetname_t	*local_sp;
1585 	md_mnset_record	*mnsr;
1586 	md_mnnode_record *nr;
1587 
1588 	if ((local_sp = metasetname(MD_LOCAL_NAME, ep)) == NULL)
1589 		return (-1);
1590 
1591 	for (sn = dn->side_names; sn != NULL; sn = sn->next)
1592 		if (MD_MNSET_REC(sr)) {
1593 			mnsr = (struct md_mnset_record *)sr;
1594 			/*
1595 			 * Only delete side name entries for this node
1596 			 * on a multi-node diskset.
1597 			 */
1598 			nr = mnsr->sr_nodechain;
1599 			while (nr) {
1600 				if (nr->nr_nodeid == sn->sideno) {
1601 					if (del_name(local_sp, sn->sideno,
1602 					    dn->side_names_key, ep) == -1)
1603 						mdclrerror(ep); /* ignore err */
1604 					break;
1605 				}
1606 				nr = nr->nr_next;
1607 			}
1608 		} else {
1609 			if (del_name(local_sp, sn->sideno+SKEW,
1610 			    dn->side_names_key, ep) == -1)
1611 				mdclrerror(ep);	/* ignore errors */
1612 		}
1613 
1614 	dn->side_names_key = MD_KEYBAD;
1615 	return (0);
1616 }
1617 
1618 static void
1619 deldrvs(
1620 	char		*setname,
1621 	md_drive_desc	*dd,
1622 	md_error_t	*ep
1623 )
1624 {
1625 	mdsetname_t	*sp;
1626 	md_set_record	*sr;
1627 	md_drive_record	*dr;
1628 	mddb_userreq_t	req;
1629 	md_drive_desc	*p;
1630 	mddrivename_t	*dn, *dn1;
1631 	side_t		sideno;
1632 	int		i;
1633 	int		rb_mode = 0;
1634 	mdname_t	*np;
1635 	md_dev64_t	dev;
1636 	md_error_t	xep = mdnullerror;
1637 	ddi_devid_t	devid_remote = NULL;
1638 	ddi_devid_t	devid_local = NULL;
1639 	int		devid_same = -1;
1640 	int		using_devid = 0;
1641 	md_mnnode_record	*nr;
1642 	md_mnset_record		*mnsr;
1643 
1644 	if ((sp = metasetname(setname, ep)) == NULL)
1645 		return;
1646 
1647 	metaflushsetname(sp);
1648 
1649 	if ((sideno = getmyside(sp, ep)) == MD_SIDEWILD) {
1650 		if (! mdisdserror(ep, MDE_DS_HOSTNOSIDE))
1651 			return;
1652 		mdclrerror(ep);
1653 		/*
1654 		 * The set record is incomplete, so we need to make note
1655 		 * here so that we can do some special handling later.
1656 		 */
1657 		rb_mode = 1;
1658 	}
1659 
1660 	if ((sr = getsetbyname(setname, ep)) == NULL)
1661 		return;
1662 
1663 	if (dd->dd_dnp == NULL)
1664 		return;
1665 
1666 	/*
1667 	 * The system is either all devid or all
1668 	 * non-devid so we determine this by looking
1669 	 * at the first item in the list.
1670 	 *
1671 	 * For did disks, the dd_dnp->devid is a valid pointer which
1672 	 * points to a '' string of devid.  We need to check this
1673 	 * before set the using_devid.
1674 	 */
1675 	if ((dd->dd_dnp->devid != NULL) && (dd->dd_dnp->devid[0] != '\0') &&
1676 	    (!(MD_MNSET_REC(sr))))
1677 		using_devid = 1;
1678 
1679 	for (p = dd; p != NULL; p = p->dd_next) {
1680 		dn = p->dd_dnp;
1681 		devid_remote = NULL;
1682 
1683 		if (dn->devid != NULL && (strlen(dn->devid) != 0) &&
1684 		    using_devid) {
1685 			/*
1686 			 * We have a devid so use it
1687 			 */
1688 			(void) devid_str_decode(dn->devid, &devid_remote, NULL);
1689 		}
1690 
1691 		/* check to make sure using_devid agrees with reality... */
1692 		if ((using_devid == 1) && (devid_remote == NULL)) {
1693 			/* something went really wrong. Can't process */
1694 			(void) mddserror(ep, MDE_DS_INVALIDDEVID, sp->setno,
1695 			    mynode(), dn->cname, sp->setname);
1696 			return;
1697 		}
1698 
1699 		for (dr = sr->sr_drivechain; dr; dr = dr->dr_next) {
1700 			devid_same = -1;
1701 
1702 			if (! rb_mode) {
1703 				dn1 = metadrivename_withdrkey(sp, sideno,
1704 				    dr->dr_key, MD_BASICNAME_OK, ep);
1705 				if (dn1 == NULL) {
1706 					free_sr(sr);
1707 					if (devid_remote)
1708 						devid_free(devid_remote);
1709 					return;
1710 				}
1711 			} else {
1712 				/*
1713 				 * Handle special case here where sidenames
1714 				 * from other hosts for this drive may be
1715 				 * in the local mddb, but there is no
1716 				 * sidename entry for this host for this drive.
1717 				 * This could have happened if the node
1718 				 * panic'd between the 2 operations when
1719 				 * adding this node to the set.
1720 				 * So, delete all sidename entries for this
1721 				 * drive.
1722 				 */
1723 				if (MD_MNSET_REC(sr)) {
1724 					mnsr = (struct md_mnset_record *)sr;
1725 					nr = mnsr->sr_nodechain;
1726 					while (nr) {
1727 						/* We delete all dr sides */
1728 						dn1 = metadrivename_withdrkey(
1729 						    sp, nr->nr_nodeid,
1730 						    dr->dr_key,
1731 						    MD_BASICNAME_OK, ep);
1732 
1733 						/* if we do, get out of loop */
1734 						if (dn1 != NULL)
1735 							break;
1736 
1737 						/* save error for later */
1738 						(void) mdstealerror(&xep, ep);
1739 
1740 						mdclrerror(ep);
1741 
1742 						nr = nr->nr_next;
1743 					}
1744 				} else {
1745 					/*
1746 					 * Handle special case here
1747 					 * for traditional diskset
1748 					 */
1749 					for (i = 0; i < MD_MAXSIDES; i++) {
1750 						/* We delete all dr sides */
1751 						dn1 = metadrivename_withdrkey(
1752 						    sp, i, dr->dr_key,
1753 						    MD_BASICNAME_OK, ep);
1754 
1755 						/* if we do, get out of loop */
1756 						if (dn1 != NULL)
1757 							break;
1758 
1759 						/* save error for later */
1760 						(void) mdstealerror(&xep, ep);
1761 
1762 						mdclrerror(ep);
1763 					}
1764 				}
1765 
1766 				if (dn1 == NULL) {
1767 					(void) mdstealerror(ep, &xep);
1768 					free_sr(sr);
1769 					if (devid_remote)
1770 						devid_free(devid_remote);
1771 					return;
1772 				}
1773 
1774 				if (!using_devid)
1775 					mdclrerror(ep);
1776 			}
1777 
1778 			if (dn1->devid != NULL && using_devid) {
1779 				if (devid_str_decode(dn1->devid, &devid_local,
1780 				    NULL) == 0) {
1781 					devid_same = devid_compare(devid_remote,
1782 					    devid_local);
1783 					devid_free(devid_local);
1784 				}
1785 			}
1786 
1787 			/*
1788 			 * Has the required disk been found - either the devids
1789 			 * match if devid are being used or the actual name of
1790 			 * the disk matches.
1791 			 */
1792 			if ((using_devid && devid_same == 0) ||
1793 			    (!using_devid &&
1794 			    strcmp(dn->cname, dn1->cname) == 0)) {
1795 				uint_t	rep_slice;
1796 
1797 				dev = NODEV64;
1798 				np = NULL;
1799 				if (meta_replicaslice(dn1,
1800 				    &rep_slice, &xep) == 0) {
1801 					np = metaslicename(dn1,
1802 					    rep_slice, &xep);
1803 				}
1804 
1805 				if (np != NULL)
1806 					dev = np->dev;
1807 				else
1808 					mdclrerror(&xep);
1809 				break;
1810 			}
1811 		}
1812 
1813 		if (dr) {
1814 			(void) memset(&req, 0, sizeof (req));
1815 			METAD_SETUP_DR(MD_DB_DELETE, dr->dr_selfid)
1816 			if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL)
1817 			    != 0) {
1818 				(void) mdstealerror(ep, &req.ur_mde);
1819 				if (devid_remote)
1820 					devid_free(devid_remote);
1821 				free_sr(sr);
1822 				return;
1823 			}
1824 
1825 			dr_cache_del(sr, dr->dr_selfid);
1826 
1827 			if (del_sidenamelist(sr, dn1, ep) == -1) {
1828 				goto out;
1829 			}
1830 
1831 			SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_REMOVE, SVM_TAG_DRIVE,
1832 			    sr->sr_setno, dev);
1833 			SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_ADD, SVM_TAG_DRIVE,
1834 			    MD_LOCAL_SET, dev);
1835 
1836 			continue;
1837 		}
1838 
1839 		if (devid_remote)
1840 			devid_free(devid_remote);
1841 	}
1842 
1843 out:
1844 	commitset(sr, TRUE, ep);
1845 
1846 	free_sr(sr);
1847 }
1848 
1849 /*
1850  * delete 1 or more drive records from a host.
1851  */
1852 bool_t
1853 mdrpc_deldrvs_common(
1854 	mdrpc_drives_2_args_r1	*args,
1855 	mdrpc_generic_res	*res,
1856 	struct svc_req		*rqstp		/* RPC stuff */
1857 )
1858 {
1859 	md_error_t		*ep = &res->status;
1860 	int			err;
1861 	int			op_mode = W_OK;
1862 
1863 	/* setup, check permissions */
1864 	(void) memset(res, 0, sizeof (*res));
1865 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
1866 		return (FALSE);
1867 	else if (err != 0)
1868 		return (TRUE);
1869 
1870 	if (check_set_lock(op_mode, args->cl_sk, ep))
1871 		return (TRUE);
1872 
1873 	/* doit */
1874 	deldrvs(args->sp->setname, args->drivedescs, ep);
1875 
1876 	err = svc_fini(ep);
1877 
1878 	return (TRUE);
1879 }
1880 
1881 /*
1882  * version 1 of the remote procedure. This procedure is called if the
1883  * client is running in version 1. We first convert version 1 arguments
1884  * into version 2 arguments and then call the common remote procedure.
1885  */
1886 bool_t
1887 mdrpc_deldrvs_1_svc(
1888 	mdrpc_drives_args	*args,
1889 	mdrpc_generic_res	*res,
1890 	struct svc_req		*rqstp		/* RPC stuff */
1891 )
1892 {
1893 	bool_t			retval;
1894 	mdrpc_drives_2_args_r1	v2_args;
1895 
1896 	/* allocate memory */
1897 	alloc_newdrvdesc(args->drivedescs, &v2_args.drivedescs);
1898 
1899 	/* build args */
1900 	v2_args.cl_sk = args->cl_sk;
1901 	v2_args.sp = args->sp;
1902 	/* convert v1 args to v2 (revision 1) args */
1903 	meta_conv_drvdesc_old2new(args->drivedescs, v2_args.drivedescs);
1904 	v2_args.timestamp = args->timestamp;
1905 	v2_args.genid = args->genid;
1906 
1907 	retval = mdrpc_deldrvs_common(&v2_args, res, rqstp);
1908 
1909 	free_newdrvdesc(v2_args.drivedescs);
1910 
1911 	return (retval);
1912 }
1913 
1914 bool_t
1915 mdrpc_deldrvs_2_svc(
1916 	mdrpc_drives_2_args	*args,
1917 	mdrpc_generic_res	*res,
1918 	struct svc_req		*rqstp		/* RPC stuff */
1919 )
1920 {
1921 	switch (args->rev) {
1922 	    case MD_METAD_ARGS_REV_1:
1923 		return (mdrpc_deldrvs_common(
1924 		    &args->mdrpc_drives_2_args_u.rev1, res, rqstp));
1925 	    default:
1926 		return (FALSE);
1927 	}
1928 }
1929 
1930 static void
1931 delhosts(
1932 	char		*setname,
1933 	int		node_c,
1934 	char		**node_v,
1935 	int		version,	/* RPC version of calling routine */
1936 	md_error_t	*ep
1937 )
1938 {
1939 	mddb_userreq_t		req;
1940 	md_set_record		*sr;
1941 	int			i, j;
1942 	md_mnset_record		*mnsr;
1943 	md_mnnode_record	*nr;
1944 
1945 	if ((sr = getsetbyname(setname, ep)) == NULL)
1946 		return;
1947 
1948 	for (i = 0; i < node_c; i++) {
1949 		/* Do MN operation if rpc version supports it and if a MN set */
1950 		if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
1951 			mnsr = (struct md_mnset_record *)sr;
1952 			nr = mnsr->sr_nodechain;
1953 			while (nr) {
1954 				if (strcmp(nr->nr_nodename, node_v[i]) == 0) {
1955 					SE_NOTIFY(EC_SVM_CONFIG,
1956 						ESC_SVM_REMOVE, SVM_TAG_HOST,
1957 						sr->sr_setno, nr->nr_nodeid);
1958 					(void) memset(&req, '\0', sizeof (req));
1959 					METAD_SETUP_NR(MD_DB_DELETE,
1960 					    nr->nr_selfid);
1961 					if (metaioctl(MD_DB_USERREQ, &req,
1962 					    &req.ur_mde, NULL) != 0) {
1963 						(void) mdstealerror(ep,
1964 						    &req.ur_mde);
1965 						free_sr(sr);
1966 						return;
1967 					}
1968 					mnnr_cache_del(mnsr, nr->nr_selfid);
1969 					break;
1970 				}
1971 				nr = nr->nr_next;
1972 			}
1973 		} else {
1974 			for (j = 0; j < MD_MAXSIDES; j++) {
1975 				if (sr->sr_nodes[j][0] == '\0')
1976 					continue;
1977 				if (strcmp(sr->sr_nodes[j], node_v[i]) != 0)
1978 					continue;
1979 				(void) memset(sr->sr_nodes[j], '\0',
1980 				    sizeof (sr->sr_nodes[j]));
1981 				SE_NOTIFY(EC_SVM_CONFIG, ESC_SVM_REMOVE,
1982 				    SVM_TAG_HOST, sr->sr_setno, j);
1983 				break;
1984 			}
1985 		}
1986 	}
1987 
1988 	(void) memset(&req, '\0', sizeof (req));
1989 	METAD_SETUP_SR(MD_DB_SETDATA, sr->sr_selfid)
1990 	/* Do MN operation if rpc version supports it and if a MN set */
1991 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
1992 		req.ur_size = sizeof (*mnsr);
1993 	} else {
1994 		req.ur_size = sizeof (*sr);
1995 	}
1996 	req.ur_data = (uintptr_t)sr;
1997 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
1998 		(void) mdstealerror(ep, &req.ur_mde);
1999 		free_sr(sr);
2000 		return;
2001 	}
2002 
2003 	commitset(sr, TRUE, ep);
2004 	free_sr(sr);
2005 }
2006 
2007 /*
2008  * delete 1 or more a hosts from a set.
2009  */
2010 bool_t
2011 mdrpc_delhosts_common(
2012 	mdrpc_host_args		*args,
2013 	mdrpc_generic_res	*res,
2014 	struct svc_req		*rqstp,		/* RPC stuff */
2015 	int			version		/* RPC version */
2016 )
2017 {
2018 	md_error_t		*ep = &res->status;
2019 	int			err;
2020 	int			op_mode = W_OK;
2021 
2022 	/* setup, check permissions */
2023 	(void) memset(res, 0, sizeof (*res));
2024 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2025 		return (FALSE);
2026 	else if (err != 0)
2027 		return (TRUE);
2028 
2029 	if (check_set_lock(op_mode, args->cl_sk, ep))
2030 		return (TRUE);
2031 
2032 	/* doit */
2033 	delhosts(args->sp->setname, args->hosts.hosts_len,
2034 	    args->hosts.hosts_val, version, ep);
2035 
2036 	err = svc_fini(ep);
2037 
2038 	return (TRUE);
2039 }
2040 
2041 bool_t
2042 mdrpc_delhosts_1_svc(
2043 	mdrpc_host_args		*args,
2044 	mdrpc_generic_res	*res,
2045 	struct svc_req		*rqstp		/* RPC stuff */
2046 )
2047 {
2048 	/* Pass RPC version (METAD_VERSION) to common routine */
2049 	return (mdrpc_delhosts_common(args, res, rqstp, METAD_VERSION));
2050 }
2051 
2052 bool_t
2053 mdrpc_delhosts_2_svc(
2054 	mdrpc_host_2_args	*args,
2055 	mdrpc_generic_res	*res,
2056 	struct svc_req		*rqstp		/* RPC stuff */
2057 )
2058 {
2059 	switch (args->rev) {
2060 	    case MD_METAD_ARGS_REV_1:
2061 		/* Pass RPC version (METAD_VERSION_DEVID) to common routine */
2062 		return (mdrpc_delhosts_common(
2063 		    &args->mdrpc_host_2_args_u.rev1, res,
2064 		    rqstp, METAD_VERSION_DEVID));
2065 	    default:
2066 		return (FALSE);
2067 	}
2068 }
2069 
2070 /*
2071  * delete a set.
2072  */
2073 bool_t
2074 mdrpc_delset_common(
2075 	mdrpc_sp_args		*args,
2076 	mdrpc_generic_res	*res,
2077 	struct svc_req		*rqstp		/* RPC stuff */
2078 )
2079 {
2080 	md_error_t		*ep = &res->status;
2081 	int			err;
2082 	int			op_mode = W_OK;
2083 
2084 	/* setup, check permissions */
2085 	(void) memset(res, 0, sizeof (*res));
2086 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2087 		return (FALSE);
2088 	else if (err != 0)
2089 		return (TRUE);
2090 
2091 	if (check_set_lock(op_mode, args->cl_sk, ep))
2092 		return (TRUE);
2093 
2094 	/* doit */
2095 	s_delset(args->sp->setname, ep);
2096 
2097 	err = svc_fini(ep);
2098 
2099 	return (TRUE);
2100 }
2101 
2102 bool_t
2103 mdrpc_delset_1_svc(
2104 	mdrpc_sp_args		*args,
2105 	mdrpc_generic_res	*res,
2106 	struct svc_req		*rqstp		/* RPC stuff */
2107 )
2108 {
2109 	return (mdrpc_delset_common(args, res, rqstp));
2110 }
2111 
2112 bool_t
2113 mdrpc_delset_2_svc(
2114 	mdrpc_sp_2_args		*args,
2115 	mdrpc_generic_res	*res,
2116 	struct svc_req		*rqstp		/* RPC stuff */
2117 )
2118 {
2119 	switch (args->rev) {
2120 	    case MD_METAD_ARGS_REV_1:
2121 		return (mdrpc_delset_common(
2122 		    &args->mdrpc_sp_2_args_u.rev1, res, rqstp));
2123 	    default:
2124 		return (FALSE);
2125 	}
2126 }
2127 
2128 /*
2129  * return device info
2130  */
2131 static void
2132 devinfo(
2133 	mdsetname_t		*sp,
2134 	mddrivename_t		*dp,
2135 	mdrpc_devinfo_2_res 	*res,
2136 	md_error_t		*ep
2137 )
2138 {
2139 	mdname_t		*np, *real_np;
2140 
2141 	if ((np = metaslicename(dp, MD_SLICE0, ep)) == NULL)
2142 		return;
2143 
2144 	if ((real_np = metaname(&sp, np->bname, ep)) == NULL)
2145 		return;
2146 
2147 	res->dev = real_np->dev;
2148 	(void) getdevstamp(dp, (long *)&res->vtime, ep);
2149 	res->enc_devid = meta_get_devid(np->rname);
2150 }
2151 
2152 bool_t
2153 mdrpc_devinfo_common(
2154 	mdrpc_devinfo_2_args_r1	*args,
2155 	mdrpc_devinfo_2_res 	*res,
2156 	struct svc_req		*rqstp			/* RPC stuff */
2157 )
2158 {
2159 	int			slice;
2160 	mdname_t		*np;
2161 	mddrivename_t		*dnp = args->drivenamep;
2162 	md_error_t		*ep = &res->status;
2163 	int			err;
2164 	int			op_mode = R_OK;
2165 
2166 	/* setup, check permissions */
2167 	(void) memset(res, 0, sizeof (*res));
2168 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2169 		return (FALSE);
2170 	else if (err != 0)
2171 		return (TRUE);
2172 
2173 	if (check_set_lock(op_mode, NULL, ep))
2174 		return (TRUE);
2175 
2176 	/*
2177 	 * fix all the drivenamep's in the mdname_t's to
2178 	 * point to the right place.
2179 	 */
2180 	for (slice = 0; (slice < dnp->parts.parts_len); ++slice) {
2181 		if ((np = metaslicename(dnp, slice, ep)) == NULL)
2182 			return (TRUE);
2183 		np->drivenamep = dnp;
2184 	}
2185 
2186 	/* doit */
2187 	devinfo(args->sp, dnp, res, ep);
2188 
2189 	err = svc_fini(ep);
2190 
2191 	return (TRUE);
2192 }
2193 
2194 /*
2195  * version 1 of the remote procedure. This procedure is called if the
2196  * client is running in version 1. We first convert version 1 arguments
2197  * into version 2 arguments and then call the common remote procedure.
2198  */
2199 bool_t
2200 mdrpc_devinfo_1_svc(
2201 	mdrpc_devinfo_args	*args,
2202 	mdrpc_devinfo_res 	*res,
2203 	struct svc_req		*rqstp			/* RPC stuff */
2204 )
2205 {
2206 	bool_t			retval;
2207 	mdrpc_devinfo_2_args_r1	v2_args;
2208 	mdrpc_devinfo_2_res	v2_res;
2209 
2210 	/* allocate memory */
2211 	v2_args.drivenamep = Zalloc(sizeof (mddrivename_t));
2212 	v2_args.drivenamep->parts.parts_val =
2213 	    Zalloc(sizeof (mdname_t) * args->drivenamep->parts.parts_len);
2214 
2215 	/* convert v1 args to v2 (revision 1) args */
2216 	meta_conv_drvname_old2new(args->drivenamep, v2_args.drivenamep);
2217 	retval = mdrpc_devinfo_common(&v2_args, &v2_res, rqstp);
2218 
2219 	/*
2220 	 * Fill in the result appropriately.
2221 	 * Since dev_t's for version 2 are 64-bit,
2222 	 * we need to convert them to 32-bit for version 1.
2223 	 */
2224 	res->dev = meta_cmpldev(v2_res.dev);
2225 	res->vtime = v2_res.vtime;
2226 	res->status = v2_res.status;
2227 
2228 	free(v2_args.drivenamep);
2229 	free(v2_args.drivenamep->parts.parts_val);
2230 
2231 	return (retval);
2232 }
2233 
2234 bool_t
2235 mdrpc_devinfo_2_svc(
2236 	mdrpc_devinfo_2_args	*args,
2237 	mdrpc_devinfo_2_res 	*res,
2238 	struct svc_req		*rqstp			/* RPC stuff */
2239 )
2240 {
2241 	switch (args->rev) {
2242 	    case MD_METAD_ARGS_REV_1:
2243 		return (mdrpc_devinfo_common(
2244 		    &args->mdrpc_devinfo_2_args_u.rev1, res, rqstp));
2245 	    default:
2246 		return (FALSE);
2247 	}
2248 }
2249 
2250 /*
2251  * return device id
2252  */
2253 static void
2254 mdrpc_get_devid(
2255 	mdsetname_t		*sp,
2256 	mddrivename_t		*dp,
2257 	mdrpc_devid_res 	*res,
2258 	md_error_t		*ep
2259 )
2260 {
2261 	mdname_t	*np;
2262 
2263 	if ((np = metaslicename(dp, MD_SLICE0, ep)) == NULL)
2264 		return;
2265 
2266 	if (metaname(&sp, np->bname, ep) == NULL)
2267 		return;
2268 
2269 	res->enc_devid = meta_get_devid(np->rname);
2270 }
2271 
2272 bool_t
2273 mdrpc_devid_2_svc(
2274 	mdrpc_devid_2_args	*args,
2275 	mdrpc_devid_res 	*res,
2276 	struct svc_req		*rqstp			/* RPC stuff */
2277 )
2278 {
2279 	int			slice;
2280 	mdname_t		*np;
2281 	mddrivename_t		*dnp;
2282 	md_error_t		*ep = &res->status;
2283 	int			err;
2284 	int			op_mode = R_OK;
2285 
2286 	switch (args->rev) {
2287 	    case MD_METAD_ARGS_REV_1:
2288 		dnp = (&(args->mdrpc_devid_2_args_u.rev1))->drivenamep;
2289 		break;
2290 	    default:
2291 		return (FALSE);
2292 	}
2293 
2294 	/* setup, check permissions */
2295 	(void) memset(res, 0, sizeof (*res));
2296 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2297 		return (FALSE);
2298 	else if (err != 0)
2299 		return (TRUE);
2300 
2301 	if (check_set_lock(op_mode, NULL, ep))
2302 		return (TRUE);
2303 
2304 	/*
2305 	 * fix all the drivenamep's in the mdname_t's to
2306 	 * point to the right place.
2307 	 */
2308 	for (slice = 0; (slice < dnp->parts.parts_len); ++slice) {
2309 		if ((np = metaslicename(dnp, slice, ep)) == NULL)
2310 			return (TRUE);
2311 		np->drivenamep = dnp;
2312 	}
2313 
2314 	/* doit */
2315 	mdrpc_get_devid((&(args->mdrpc_devid_2_args_u.rev1))->sp, dnp, res, ep);
2316 
2317 	err = svc_fini(ep);
2318 
2319 	return (TRUE);
2320 }
2321 
2322 /*
2323  * This routine should not be called for a multi-node diskset.
2324  *
2325  * The devid support is disabled for MN diskset so this routine
2326  * will not be called if the set is MN diskset.  The check has
2327  * been done early in meta_getnextside_devinfo.  However this
2328  * routine will be called when the devid support for MN set is
2329  * enabled and check is removed.
2330  */
2331 bool_t
2332 mdrpc_devinfo_by_devid_2_svc(
2333 	mdrpc_devidstr_args	*args,
2334 	mdrpc_devinfo_2_res	*res,
2335 	struct svc_req	  *rqstp		  /* RPC stuff */
2336 )
2337 {
2338 
2339 	char		*devidstr = args->enc_devid;
2340 	md_error_t	*ep = &res->status;
2341 	ddi_devid_t	devid;
2342 	char		*minor_name = NULL;
2343 	int		ret = 0;
2344 	int		err;
2345 	devid_nmlist_t	*disklist = NULL;
2346 	int		op_mode = R_OK;
2347 	mdname_t	*np;
2348 	mdsetname_t	*sp = args->sp;
2349 
2350 	/* setup, check permissions */
2351 	(void) memset(res, 0, sizeof (*res));
2352 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2353 		return (FALSE);
2354 	else if (err != 0)
2355 		return (TRUE);
2356 
2357 	if (check_set_lock(op_mode, NULL, ep))
2358 		return (TRUE);
2359 
2360 	if (devid_str_decode(devidstr, &devid, &minor_name) != 0)
2361 		return (TRUE);
2362 
2363 	/*
2364 	 * if we do not have a minor name then look for a character device.
2365 	 * This is because the caller (checkdrive_onnode) expects a character
2366 	 * device to be returned. The other client of this interface is
2367 	 * meta_getnextside_devinfo and this supplies a minor name.
2368 	 */
2369 	if (minor_name == NULL) {
2370 		ret = meta_deviceid_to_nmlist("/dev", devid,
2371 		    DEVID_MINOR_NAME_ALL_CHR, &disklist);
2372 	} else {
2373 		ret = meta_deviceid_to_nmlist("/dev", devid, minor_name,
2374 		    &disklist);
2375 		devid_str_free(minor_name);
2376 	}
2377 
2378 	devid_free(devid);
2379 	if (ret != 0) {
2380 		res->dev = NODEV64;
2381 		devid_free_nmlist(disklist);
2382 		return (TRUE);
2383 	}
2384 
2385 	np = metaname(&sp, disklist[0].devname, ep);
2386 	if (np != NULL) {
2387 		mdcinfo_t	*cinfo;
2388 		if ((cinfo = metagetcinfo(np, ep)) != NULL) {
2389 			res->drivername = Strdup(cinfo->dname);
2390 		}
2391 	}
2392 
2393 	res->dev = meta_expldev(disklist[0].dev);
2394 	res->devname = strdup(disklist[0].devname);
2395 
2396 	devid_free_nmlist(disklist);
2397 
2398 	err = svc_fini(ep);
2399 
2400 	return (TRUE);
2401 }
2402 
2403 /*
2404  * This routine should not be called for a multi-node diskset.
2405  *
2406  * The devid support is disabled for MN diskset so this routine
2407  * will not be called if the set is MN diskset.  The check has
2408  * been done early in meta_getnextside_devinfo.  However this
2409  * routine will be called when the devid support for MN set is
2410  * enabled and check is removed.
2411  *
2412  * This function will return the device info attempting to use
2413  * both the passed in devid and device name.  This is to deal
2414  * with systems that use multi-path disks but not running mpxio.
2415  * In this situation meta_deviceid_to_nmlist will return multiple
2416  * devices.  The orig_devname is used to disambiguate.
2417  *
2418  */
2419 bool_t
2420 mdrpc_devinfo_by_devid_name_2_svc(
2421 	mdrpc_devid_name_2_args	*args,
2422 	mdrpc_devinfo_2_res	*res,
2423 	struct svc_req	  *rqstp		  /* RPC stuff */
2424 )
2425 {
2426 
2427 	char		*devidstr;
2428 	char		*orig_devname;
2429 	md_error_t	*ep = &res->status;
2430 	ddi_devid_t	devid;
2431 	char		*minor_name = NULL;
2432 	int		ret = 0;
2433 	int		err;
2434 	int		i;
2435 	devid_nmlist_t	*disklist = NULL;
2436 	int		op_mode = R_OK;
2437 	mdname_t	*np;
2438 	mdsetname_t	*sp;
2439 
2440 	switch (args->rev) {
2441 	    case MD_METAD_ARGS_REV_1:
2442 		sp = (&(args->mdrpc_devid_name_2_args_u.rev1))->sp;
2443 		devidstr = (&(args->mdrpc_devid_name_2_args_u.rev1))->enc_devid;
2444 		orig_devname =
2445 			(&(args->mdrpc_devid_name_2_args_u.rev1))->orig_devname;
2446 		break;
2447 	    default:
2448 		return (FALSE);
2449 	}
2450 
2451 	/* setup, check permissions */
2452 	(void) memset(res, 0, sizeof (*res));
2453 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2454 		return (FALSE);
2455 	else if (err != 0)
2456 		return (TRUE);
2457 
2458 	if (check_set_lock(op_mode, NULL, ep))
2459 		return (TRUE);
2460 
2461 	if (devid_str_decode(devidstr, &devid, &minor_name) != 0)
2462 		return (TRUE);
2463 
2464 	/*
2465 	 * if we do not have a minor name then look for a character device.
2466 	 * This is because the caller (checkdrive_onnode) expects a character
2467 	 * device to be returned. The other client of this interface is
2468 	 * meta_getnextside_devinfo and this supplies a minor name.
2469 	 */
2470 	if (minor_name == NULL) {
2471 		ret = meta_deviceid_to_nmlist("/dev", devid,
2472 		    DEVID_MINOR_NAME_ALL_CHR, &disklist);
2473 	} else {
2474 		ret = meta_deviceid_to_nmlist("/dev", devid, minor_name,
2475 		    &disklist);
2476 		devid_str_free(minor_name);
2477 	}
2478 
2479 	devid_free(devid);
2480 	if (ret != 0) {
2481 		res->dev = NODEV64;
2482 		devid_free_nmlist(disklist);
2483 		return (TRUE);
2484 	}
2485 
2486 	/* attempt to match to the device name on the originating node */
2487 	for (i = 0; disklist[i].dev != NODEV; i++) {
2488 		if (strncmp(orig_devname, disklist[i].devname,
2489 					strlen(disklist[i].devname)) == 0)
2490 			break;
2491 	}
2492 
2493 	/* if it's not found then use the first disk in the list */
2494 	if (disklist[i].dev == NODEV)
2495 		i = 0;
2496 
2497 	np = metaname(&sp, disklist[i].devname, ep);
2498 	if (np != NULL) {
2499 		mdcinfo_t	*cinfo;
2500 		if ((cinfo = metagetcinfo(np, ep)) != NULL) {
2501 			res->drivername = Strdup(cinfo->dname);
2502 		}
2503 	}
2504 
2505 	res->dev = meta_expldev(disklist[i].dev);
2506 	res->devname = strdup(disklist[i].devname);
2507 
2508 	devid_free_nmlist(disklist);
2509 
2510 	err = svc_fini(ep);
2511 
2512 	return (TRUE);
2513 }
2514 
2515 static void
2516 drvused(mdsetname_t *sp, mddrivename_t *dnp, md_error_t *ep)
2517 {
2518 	if (meta_check_drivemounted(sp, dnp, ep))
2519 		return;
2520 
2521 	if (meta_check_driveswapped(sp, dnp, ep))
2522 		return;
2523 
2524 	if (meta_check_drive_inuse(metasetname(MD_LOCAL_NAME, ep), dnp,
2525 	    TRUE, ep))
2526 		return;
2527 
2528 	(void) meta_check_driveinset(sp, dnp, ep);
2529 }
2530 
2531 /*
2532  * determine if a device is in use.
2533  */
2534 bool_t
2535 mdrpc_drvused_common(
2536 	mdrpc_drvused_2_args_r1	*args,
2537 	mdrpc_generic_res	*res,
2538 	struct svc_req		*rqstp		/* RPC stuff */
2539 )
2540 {
2541 	md_error_t		*ep = &res->status;
2542 	int			slice;
2543 	mdname_t		*np;
2544 	mddrivename_t		*dnp = args->drivenamep;
2545 	int			err;
2546 	int			op_mode = R_OK;
2547 
2548 	/* setup, check permissions */
2549 	(void) memset(res, 0, sizeof (*res));
2550 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2551 		return (FALSE);
2552 	else if (err != 0)
2553 		return (TRUE);
2554 
2555 	if (check_set_lock(op_mode, NULL, ep))
2556 		return (TRUE);
2557 
2558 	if (dnp == NULL) {
2559 		/* no drive pointer specified */
2560 		return (TRUE);
2561 	}
2562 	/*
2563 	 * fix all the drivenamep's in the mdname_t's to
2564 	 * point to the right place.
2565 	 */
2566 	for (slice = 0; (slice < dnp->parts.parts_len); ++slice) {
2567 		if ((np = metaslicename(dnp, slice, ep)) == NULL)
2568 			return (TRUE);
2569 		np->drivenamep = dnp;
2570 	}
2571 
2572 	/* doit */
2573 	drvused(args->sp, dnp, ep);
2574 
2575 	err = svc_fini(ep);
2576 
2577 	return (TRUE);
2578 }
2579 
2580 /*
2581  * version 1 of the remote procedure. This procedure is called if the
2582  * client is running in version 1. We first convert version 1 arguments
2583  * into version 2 arguments and then call the common remote procedure.
2584  */
2585 bool_t
2586 mdrpc_drvused_1_svc(
2587 	mdrpc_drvused_args	*args,
2588 	mdrpc_generic_res	*res,
2589 	struct svc_req		*rqstp		/* RPC stuff */
2590 )
2591 {
2592 	bool_t			retval;
2593 	mdrpc_drvused_2_args_r1	v2_args;
2594 
2595 	/* allocate memory */
2596 	v2_args.drivenamep = Zalloc(sizeof (mddrivename_t));
2597 	v2_args.drivenamep->parts.parts_val =
2598 	    Zalloc(sizeof (mdname_t) * args->drivenamep->parts.parts_len);
2599 
2600 	/* build args */
2601 	v2_args.sp = args->sp;
2602 	v2_args.cl_sk = args->cl_sk;
2603 
2604 	/* convert v1 args to v2 (revision 1) args */
2605 	meta_conv_drvname_old2new(args->drivenamep, v2_args.drivenamep);
2606 	retval = mdrpc_drvused_common(&v2_args, res, rqstp);
2607 
2608 	free(v2_args.drivenamep);
2609 	free(v2_args.drivenamep->parts.parts_val);
2610 
2611 	return (retval);
2612 }
2613 
2614 bool_t
2615 mdrpc_drvused_2_svc(
2616 	mdrpc_drvused_2_args	*args,
2617 	mdrpc_generic_res	*res,
2618 	struct svc_req		*rqstp		/* RPC stuff */
2619 )
2620 {
2621 	switch (args->rev) {
2622 	    case MD_METAD_ARGS_REV_1:
2623 		return (mdrpc_drvused_common(
2624 		    &args->mdrpc_drvused_2_args_u.rev1, res, rqstp));
2625 	    default:
2626 		return (FALSE);
2627 	}
2628 }
2629 
2630 /*
2631  * return a set records selected by name or number.
2632  */
2633 bool_t
2634 mdrpc_getset_common(
2635 	mdrpc_getset_args	*args,
2636 	mdrpc_getset_res 	*res,
2637 	struct svc_req		*rqstp		/* RPC stuff */
2638 )
2639 {
2640 	md_error_t		*ep = &res->status;
2641 	int			err;
2642 	int			op_mode = R_OK;
2643 
2644 	/* setup, check permissions */
2645 	(void) memset(res, 0, sizeof (*res));
2646 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2647 		return (FALSE);
2648 	else if (err != 0)
2649 		return (TRUE);
2650 
2651 	/* Don't have a setno, so we don't check the lock */
2652 	if (check_set_lock(op_mode, NULL, ep))
2653 		return (TRUE);
2654 
2655 	/* doit */
2656 	if (args->setname && *args->setname)
2657 		res->sr = setdup(getsetbyname(args->setname, ep));
2658 	else if (args->setno > 0)
2659 		res->sr = setdup(getsetbynum(args->setno, ep));
2660 	else
2661 		res->sr = NULL;
2662 
2663 	err = svc_fini(ep);
2664 
2665 	return (TRUE);
2666 }
2667 
2668 bool_t
2669 mdrpc_getset_1_svc(
2670 	mdrpc_getset_args	*args,
2671 	mdrpc_getset_res 	*res,
2672 	struct svc_req		*rqstp		/* RPC stuff */
2673 )
2674 {
2675 	return (mdrpc_getset_common(args, res, rqstp));
2676 }
2677 
2678 bool_t
2679 mdrpc_getset_2_svc(
2680 	mdrpc_getset_2_args	*args,
2681 	mdrpc_getset_res 	*res,
2682 	struct svc_req		*rqstp		/* RPC stuff */
2683 )
2684 {
2685 	switch (args->rev) {
2686 	    case MD_METAD_ARGS_REV_1:
2687 		return (mdrpc_getset_common(
2688 		    &args->mdrpc_getset_2_args_u.rev1, res, rqstp));
2689 	    default:
2690 		return (FALSE);
2691 	}
2692 }
2693 
2694 /*
2695  * return a MN set record selected by name or number.
2696  */
2697 bool_t
2698 mdrpc_mngetset_common(
2699 	mdrpc_getset_args	*args,
2700 	mdrpc_mngetset_res 	*res,
2701 	struct svc_req		*rqstp		/* RPC stuff */
2702 )
2703 {
2704 	md_error_t		*ep = &res->status;
2705 	int			err;
2706 	int			op_mode = R_OK;
2707 	md_set_record		*sr = NULL;
2708 	md_mnset_record		*mnsr = NULL;
2709 
2710 	/* setup, check permissions */
2711 	(void) memset(res, 0, sizeof (*res));
2712 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2713 		return (FALSE);
2714 	else if (err != 0)
2715 		return (TRUE);
2716 
2717 	/* Don't have a setno, so we don't check the lock */
2718 	if (check_set_lock(op_mode, NULL, ep))
2719 		return (TRUE);
2720 
2721 	/* doit */
2722 	res->mnsr = NULL;
2723 	if (args->setname && *args->setname)
2724 		sr = getsetbyname(args->setname, ep);
2725 	else if (args->setno > 0)
2726 		sr = getsetbynum(args->setno, ep);
2727 
2728 	if ((sr) && (MD_MNSET_REC(sr))) {
2729 		mnsr = (struct md_mnset_record *)sr;
2730 		res->mnsr = mnsetdup(mnsr);
2731 	}
2732 
2733 	err = svc_fini(ep);
2734 
2735 	return (TRUE);
2736 }
2737 
2738 bool_t
2739 mdrpc_mngetset_2_svc(
2740 	mdrpc_getset_2_args	*args,
2741 	mdrpc_mngetset_res	*res,
2742 	struct svc_req		*rqstp		/* RPC stuff */
2743 )
2744 {
2745 	switch (args->rev) {
2746 	    case MD_METAD_ARGS_REV_1:
2747 		return (mdrpc_mngetset_common(
2748 		    &args->mdrpc_getset_2_args_u.rev1, res, rqstp));
2749 	    default:
2750 		return (FALSE);
2751 	}
2752 }
2753 
2754 static void
2755 upd_setmaster(
2756 	mdsetname_t	*sp,
2757 	md_node_nm_t	master_nodenm,
2758 	int		master_nodeid,
2759 	md_error_t	*ep
2760 )
2761 {
2762 	mdsetname_t	*local_sp;
2763 	md_set_record	*sr;
2764 	md_mnset_record	*mnsr;
2765 	mddb_setmaster_config_t	sm;
2766 
2767 	if ((local_sp = metasetname(sp->setname, ep)) == NULL)
2768 		return;
2769 
2770 	metaflushsetname(local_sp);
2771 
2772 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
2773 		return;
2774 
2775 	if (MD_MNSET_REC(sr)) {
2776 		mnsr = (struct md_mnset_record *)sr;
2777 		strlcpy(mnsr->sr_master_nodenm, master_nodenm,
2778 			MD_MAX_NODENAME);
2779 		mnsr->sr_master_nodeid = master_nodeid;
2780 		if (master_nodeid != 0) {
2781 			(void) memset(&sm, 0, sizeof (sm));
2782 			sm.c_setno = sp->setno;
2783 			/* Use magic to help protect ioctl against attack. */
2784 			sm.c_magic = MDDB_SETMASTER_MAGIC;
2785 			if (strcmp(master_nodenm, mynode()) == 0) {
2786 				sm.c_current_host_master = 1;
2787 			} else {
2788 				sm.c_current_host_master = 0;
2789 			}
2790 			(void) metaioctl(MD_SETMASTER, &sm, &sm.c_mde, NULL);
2791 			mdclrerror(&sm.c_mde);
2792 		}
2793 	}
2794 
2795 out:
2796 	commitset(sr, FALSE, ep);
2797 	free_sr(sr);
2798 }
2799 
2800 /*
2801  * set the master and nodeid in node record
2802  */
2803 bool_t
2804 mdrpc_mnsetmaster_common(
2805 	mdrpc_mnsetmaster_args	*args,
2806 	mdrpc_generic_res 	*res,
2807 	struct svc_req		*rqstp		/* RPC stuff */
2808 )
2809 {
2810 	md_error_t		*ep = &res->status;
2811 	int			err;
2812 	int			op_mode = W_OK;
2813 
2814 	/* setup, check permissions */
2815 	(void) memset(res, 0, sizeof (*res));
2816 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2817 		return (FALSE);
2818 	else if (err != 0)
2819 		return (TRUE);
2820 
2821 	if (check_set_lock(op_mode, args->cl_sk, ep))
2822 		return (TRUE);
2823 
2824 	/* doit */
2825 	upd_setmaster(args->sp, args->master_nodenm, args->master_nodeid, ep);
2826 
2827 	err = svc_fini(ep);
2828 
2829 	return (TRUE);
2830 }
2831 
2832 bool_t
2833 mdrpc_mnsetmaster_2_svc(
2834 	mdrpc_mnsetmaster_2_args	*args,
2835 	mdrpc_generic_res	*res,
2836 	struct svc_req		*rqstp		/* RPC stuff */
2837 )
2838 {
2839 	switch (args->rev) {
2840 	    case MD_METAD_ARGS_REV_1:
2841 		return (mdrpc_mnsetmaster_common(
2842 		    &args->mdrpc_mnsetmaster_2_args_u.rev1, res, rqstp));
2843 	    default:
2844 		return (FALSE);
2845 	}
2846 }
2847 
2848 /*
2849  * Join this node to the diskset.
2850  * Pass stale_flag information to snarf_set so that snarf code
2851  * can choose a STALE or non-STALE state when starting the set.
2852  * If master is STALE, any joining node will join a stale set regardless
2853  * of the number of accessible mddbs.  Also, if master is at 50%
2854  * accessible replicas and is in the TOOFEW state, don't mark newly
2855  * joining node as STALE; mark it TOOFEW instead.
2856  */
2857 static void
2858 joinset(
2859 	mdsetname_t	*sp,
2860 	int		flags,
2861 	md_error_t	*ep
2862 )
2863 {
2864 	mdsetname_t		*local_sp;
2865 	md_drive_desc		*mydd;
2866 	bool_t			stale_bool;
2867 	mddb_block_parm_t	mbp;
2868 	md_error_t		xep = mdnullerror;
2869 
2870 	if ((local_sp = metasetname(sp->setname, ep)) == NULL)
2871 		return;
2872 
2873 	/*
2874 	 * Start mddoors daemon here.
2875 	 * mddoors itself takes care there will be
2876 	 * only one instance running, so starting it twice won't hurt
2877 	 */
2878 	pclose(popen(MDDOORS, "w"));
2879 
2880 	/*
2881 	 * Get latest copy of data.  If a drive was just added causing
2882 	 * nodes to get joined - this drive won't be in the local
2883 	 * name caches drive list yet.
2884 	 */
2885 	metaflushsetname(local_sp);
2886 
2887 	mydd = metaget_drivedesc(local_sp, (MD_BASICNAME_OK | PRINT_FAST), ep);
2888 	if (mydd) {
2889 		/* Causes mddbs to be loaded in kernel */
2890 		if (setup_db_bydd(local_sp, mydd, 0, ep) == -1) {
2891 			/* If ep isn't set for some reason, set it */
2892 			if (! mdisok(ep)) {
2893 				(void) mdmddberror(ep, MDE_DB_NOTNOW,
2894 				    (minor_t)NODEV64, sp->setno, 0, NULL);
2895 			}
2896 			return;
2897 		}
2898 
2899 		if (flags & MNSET_IS_STALE)
2900 			stale_bool = TRUE;
2901 		else
2902 			stale_bool = FALSE;
2903 
2904 		/*
2905 		 * Snarf the set.  No failure has occurred if STALE or
2906 		 * ACCOK error was set.  Otherwise, fail the call setting
2907 		 * a generic error if no error was already set.
2908 		 *
2909 		 * STALE means that set has < 50% mddbs.
2910 		 * ACCOK means that the mediator provided an extra vote.
2911 		 */
2912 		if (snarf_set(local_sp, stale_bool, ep) != 0) {
2913 			if (!(mdismddberror(ep, MDE_DB_STALE)) &&
2914 			    !(mdismddberror(ep, MDE_DB_ACCOK))) {
2915 				return;
2916 			} else if (mdisok(ep)) {
2917 				/* If snarf failed, but no error set - set it */
2918 				(void) mdmddberror(ep, MDE_DB_NOTNOW,
2919 				    (minor_t)NODEV64, sp->setno, 0, NULL);
2920 				return;
2921 			}
2922 		}
2923 
2924 		/*
2925 		 * If node is joining during reconfig cycle, then
2926 		 * set mddb_parse to be in blocked state so that
2927 		 * mddb reparse messages are not generated until
2928 		 * the commd has been resumed later in the reconfig
2929 		 * cycle.
2930 		 */
2931 		if (flags & MNSET_IN_RECONFIG) {
2932 			(void) memset(&mbp, 0, sizeof (mbp));
2933 			if (s_ownset(sp->setno, &xep) == MD_SETOWNER_YES) {
2934 				(void) memset(&mbp, 0, sizeof (mbp));
2935 				mbp.c_setno = local_sp->setno;
2936 				mbp.c_blk_flags = MDDB_BLOCK_PARSE;
2937 				if (metaioctl(MD_MN_MDDB_BLOCK, &mbp,
2938 				    &mbp.c_mde, NULL)) {
2939 					mdstealerror(&xep, &mbp.c_mde);
2940 					mde_perror(ep, gettext(
2941 					    "Could not block set %s"),
2942 					    sp->setname);
2943 					return;
2944 				}
2945 			}
2946 			/*
2947 			 * If s_ownset fails and snarf_set succeeded,
2948 			 * then can steal the ownset failure information
2949 			 * and store it into ep. If snarf_set failed,
2950 			 * don't overwrite critical ep information even
2951 			 * if s_ownset failed.
2952 			 */
2953 			if (!mdisok(&xep)) {
2954 				/*
2955 				 * If snarf_set succeeded or snarf_set failed
2956 				 * with MDE_DB_ACCOK (which is set if the
2957 				 * mediator provided the extra vote) then
2958 				 * steal the xep failure information and put
2959 				 * into ep.
2960 				 */
2961 				if (mdisok(ep) ||
2962 				    mdismddberror(ep, MDE_DB_ACCOK)) {
2963 					mdstealerror(ep, &xep);
2964 				}
2965 			}
2966 		}
2967 	}
2968 }
2969 
2970 /*
2971  * Have this node join the set.
2972  * This is called when a node has been
2973  * added to a MN diskset that has drives.
2974  * Also, called when a node is an alive
2975  * member of a MN diskset and the first
2976  * drive has been added.
2977  */
2978 bool_t
2979 mdrpc_joinset_common(
2980 	mdrpc_sp_flags_args	*args,
2981 	mdrpc_generic_res 	*res,
2982 	struct svc_req		*rqstp		/* RPC stuff */
2983 )
2984 {
2985 	md_error_t		*ep = &res->status;
2986 	int			err;
2987 	int			op_mode = W_OK;
2988 
2989 	/* setup, check permissions */
2990 	(void) memset(res, 0, sizeof (*res));
2991 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
2992 		return (FALSE);
2993 	else if (err != 0)
2994 		return (TRUE);
2995 
2996 	/*
2997 	 * During reconfig, joinset can happen without
2998 	 * locking first.  Turn off reconfig flag before calling
2999 	 * joinset.
3000 	 */
3001 	if (!(args->flags & MNSET_IN_RECONFIG)) {
3002 		if (check_set_lock(op_mode, args->cl_sk, ep))
3003 			return (TRUE);
3004 	}
3005 
3006 	/* doit */
3007 	joinset(args->sp, args->flags, ep);
3008 
3009 	err = svc_fini(ep);
3010 
3011 	return (TRUE);
3012 }
3013 
3014 bool_t
3015 mdrpc_joinset_2_svc(
3016 	mdrpc_sp_flags_2_args	*args,
3017 	mdrpc_generic_res	*res,
3018 	struct svc_req		*rqstp		/* RPC stuff */
3019 )
3020 {
3021 	switch (args->rev) {
3022 	    case MD_METAD_ARGS_REV_1:
3023 		return (mdrpc_joinset_common(
3024 		    &args->mdrpc_sp_flags_2_args_u.rev1, res, rqstp));
3025 	    default:
3026 		return (FALSE);
3027 	}
3028 }
3029 
3030 static void
3031 withdrawset(
3032 	mdsetname_t	*sp,
3033 	md_error_t	*ep
3034 )
3035 {
3036 	mdsetname_t	*my_sp;
3037 
3038 	if ((my_sp = metasetname(sp->setname, ep)) == NULL)
3039 		return;
3040 
3041 	(void) halt_set(my_sp, ep);
3042 }
3043 
3044 /*
3045  * Have this node withdraw from set.
3046  * In response to a failure that occurred
3047  * on the client after a joinset.
3048  */
3049 bool_t
3050 mdrpc_withdrawset_common(
3051 	mdrpc_sp_args		*args,
3052 	mdrpc_generic_res 	*res,
3053 	struct svc_req		*rqstp		/* RPC stuff */
3054 )
3055 {
3056 	md_error_t		*ep = &res->status;
3057 	int			err;
3058 	int			op_mode = W_OK;
3059 
3060 	/* setup, check permissions */
3061 	(void) memset(res, 0, sizeof (*res));
3062 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3063 		return (FALSE);
3064 	else if (err != 0)
3065 		return (TRUE);
3066 
3067 	if (check_set_lock(op_mode, args->cl_sk, ep))
3068 		return (TRUE);
3069 
3070 	/* doit */
3071 	withdrawset(args->sp, ep);
3072 
3073 	err = svc_fini(ep);
3074 
3075 	return (TRUE);
3076 }
3077 
3078 bool_t
3079 mdrpc_withdrawset_2_svc(
3080 	mdrpc_sp_2_args		*args,
3081 	mdrpc_generic_res	*res,
3082 	struct svc_req		*rqstp		/* RPC stuff */
3083 )
3084 {
3085 	switch (args->rev) {
3086 	    case MD_METAD_ARGS_REV_1:
3087 		return (mdrpc_withdrawset_common(
3088 		    &args->mdrpc_sp_2_args_u.rev1, res, rqstp));
3089 	    default:
3090 		return (FALSE);
3091 	}
3092 }
3093 
3094 static mhd_mhiargs_t *
3095 gtimeout(mdsetname_t *sp, md_error_t *ep)
3096 {
3097 	md_set_record		*sr;
3098 	mhd_mhiargs_t		*mhiargs;
3099 
3100 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
3101 		return (NULL);
3102 
3103 	mhiargs = Zalloc(sizeof (*mhiargs));
3104 	*mhiargs = sr->sr_mhiargs;
3105 
3106 	free_sr(sr);
3107 	return (mhiargs);
3108 }
3109 
3110 /*
3111  * Get the MH timeout values for this set.
3112  */
3113 bool_t
3114 mdrpc_gtimeout_common(
3115 	mdrpc_sp_args		*args,
3116 	mdrpc_gtimeout_res 	*res,
3117 	struct svc_req		*rqstp		/* RPC stuff */
3118 )
3119 {
3120 	md_error_t		*ep = &res->status;
3121 	int			err;
3122 	int			op_mode = R_OK;
3123 
3124 	/* setup, check permissions */
3125 	(void) memset(res, 0, sizeof (*res));
3126 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3127 		return (FALSE);
3128 	else if (err != 0)
3129 		return (TRUE);
3130 
3131 	if (check_set_lock(op_mode, NULL, ep))
3132 		return (TRUE);
3133 
3134 	/* doit */
3135 	res->mhiargsp = gtimeout(args->sp, ep);
3136 
3137 	err = svc_fini(ep);
3138 
3139 	return (TRUE);
3140 }
3141 
3142 bool_t
3143 mdrpc_gtimeout_1_svc(
3144 	mdrpc_sp_args		*args,
3145 	mdrpc_gtimeout_res 	*res,
3146 	struct svc_req		*rqstp		/* RPC stuff */
3147 )
3148 {
3149 	return (mdrpc_gtimeout_common(args, res, rqstp));
3150 }
3151 
3152 bool_t
3153 mdrpc_gtimeout_2_svc(
3154 	mdrpc_sp_2_args		*args,
3155 	mdrpc_gtimeout_res 	*res,
3156 	struct svc_req		*rqstp		/* RPC stuff */
3157 )
3158 {
3159 	switch (args->rev) {
3160 	    case MD_METAD_ARGS_REV_1:
3161 		return (mdrpc_gtimeout_common(
3162 		    &args->mdrpc_sp_2_args_u.rev1, res, rqstp));
3163 	    default:
3164 		return (FALSE);
3165 	}
3166 }
3167 
3168 /*
3169  * return the official host name for the callee
3170  */
3171 /*ARGSUSED*/
3172 bool_t
3173 mdrpc_hostname_common(
3174 	mdrpc_null_args		*args,
3175 	mdrpc_hostname_res 	*res,
3176 	struct svc_req		*rqstp		/* RPC stuff */
3177 )
3178 {
3179 	md_error_t		*ep = &res->status;
3180 	int			err;
3181 	int			op_mode = R_OK;
3182 
3183 	/* setup, check permissions */
3184 	(void) memset(res, 0, sizeof (*res));
3185 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3186 		return (FALSE);
3187 	else if (err != 0)
3188 		return (TRUE);
3189 
3190 	if (check_set_lock(op_mode, NULL, ep))
3191 		return (TRUE);
3192 
3193 	/* doit */
3194 	res->hostname = Strdup(mynode());
3195 
3196 	err = svc_fini(ep);
3197 
3198 	return (TRUE);
3199 }
3200 
3201 bool_t
3202 mdrpc_hostname_1_svc(
3203 	mdrpc_null_args		*args,
3204 	mdrpc_hostname_res 	*res,
3205 	struct svc_req		*rqstp		/* RPC stuff */
3206 )
3207 {
3208 	return (mdrpc_hostname_common(args, res, rqstp));
3209 }
3210 
3211 bool_t
3212 mdrpc_hostname_2_svc(
3213 	mdrpc_null_args		*args,
3214 	mdrpc_hostname_res 	*res,
3215 	struct svc_req		*rqstp		/* RPC stuff */
3216 )
3217 {
3218 	return (mdrpc_hostname_common(args, res, rqstp));
3219 }
3220 
3221 /*
3222  * return a response
3223  */
3224 /*ARGSUSED*/
3225 bool_t
3226 mdrpc_nullproc_common(
3227 	void		*args,
3228 	md_error_t	*ep,
3229 	struct svc_req	*rqstp		/* RPC stuff */
3230 )
3231 {
3232 	*ep = mdnullerror;
3233 	/* do nothing */
3234 	return (TRUE);
3235 }
3236 
3237 bool_t
3238 mdrpc_nullproc_1_svc(
3239 	void		*args,
3240 	md_error_t	*ep,
3241 	struct svc_req	*rqstp		/* RPC stuff */
3242 )
3243 {
3244 	return (mdrpc_nullproc_common(args, ep, rqstp));
3245 }
3246 
3247 bool_t
3248 mdrpc_nullproc_2_svc(
3249 	void		*args,
3250 	md_error_t	*ep,
3251 	struct svc_req	*rqstp		/* RPC stuff */
3252 )
3253 {
3254 	return (mdrpc_nullproc_common(args, ep, rqstp));
3255 }
3256 
3257 /*
3258  * determine if the caller owns the set.
3259  */
3260 bool_t
3261 mdrpc_ownset_common(
3262 	mdrpc_sp_args		*args,
3263 	mdrpc_bool_res		*res,
3264 	struct svc_req		*rqstp		/* RPC stuff */
3265 )
3266 {
3267 	md_error_t		*ep = &res->status;
3268 	int			err;
3269 	int			op_mode = R_OK;
3270 
3271 	/* setup, check permissions */
3272 	(void) memset(res, 0, sizeof (*res));
3273 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3274 		return (FALSE);
3275 	else if (err != 0)
3276 		return (TRUE);
3277 
3278 	if (check_set_lock(op_mode, NULL, ep))
3279 		return (TRUE);
3280 
3281 	/* doit */
3282 	if (s_ownset(args->sp->setno, ep))
3283 		res->value = TRUE;
3284 	else
3285 		res->value = FALSE;
3286 
3287 	err = svc_fini(ep);
3288 
3289 	return (TRUE);
3290 }
3291 
3292 bool_t
3293 mdrpc_ownset_1_svc(
3294 	mdrpc_sp_args		*args,
3295 	mdrpc_bool_res		*res,
3296 	struct svc_req		*rqstp		/* RPC stuff */
3297 )
3298 {
3299 	return (mdrpc_ownset_common(args, res, rqstp));
3300 }
3301 
3302 bool_t
3303 mdrpc_ownset_2_svc(
3304 	mdrpc_sp_2_args		*args,
3305 	mdrpc_bool_res		*res,
3306 	struct svc_req		*rqstp		/* RPC stuff */
3307 )
3308 {
3309 	switch (args->rev) {
3310 	    case MD_METAD_ARGS_REV_1:
3311 		return (mdrpc_ownset_common(
3312 		    &args->mdrpc_sp_2_args_u.rev1, res, rqstp));
3313 	    default:
3314 		return (FALSE);
3315 	}
3316 }
3317 
3318 static int
3319 setnameok(char *setname, md_error_t *ep)
3320 {
3321 	int			rval = 0;
3322 	struct	stat		statb;
3323 	md_set_record		*sr = NULL;
3324 	char			*setlink = NULL;
3325 
3326 	setlink = Strdup("/dev/md/");
3327 	setlink = Realloc(setlink, strlen(setlink) + strlen(setname) + 1);
3328 	(void) strcat(setlink, setname);
3329 
3330 	if (lstat(setlink, &statb) == -1) {
3331 		/*
3332 		 * If lstat() fails with ENOENT, setname is OK, if it
3333 		 * fails for other than that, we fail the RPC
3334 		 */
3335 		if (errno == ENOENT) {
3336 			rval = 1;
3337 			goto out;
3338 		}
3339 
3340 		(void) mdsyserror(ep, errno, setlink);
3341 		goto out;
3342 	}
3343 
3344 	/*
3345 	 * If the lstat() succeeded, then we see what type of object
3346 	 * we are dealing with, if it is a symlink, we do some further
3347 	 * checking, if it is not a symlink, then we return an
3348 	 * indication that the set name is NOT acceptable.
3349 	 */
3350 	if (! S_ISLNK(statb.st_mode))
3351 		goto out;
3352 
3353 	/*
3354 	 * We look up the setname to see if there is a set
3355 	 * with that name, if there is, then we return
3356 	 * an indication that the set name is NOT acceptable.
3357 	 */
3358 	if ((sr = getsetbyname(setname, ep)) != NULL)
3359 		goto out;
3360 
3361 	if (! mdiserror(ep, MDE_NO_SET))
3362 		goto out;
3363 
3364 	mdclrerror(ep);
3365 
3366 	rval = 1;
3367 out:
3368 	if (sr != NULL)
3369 		free_sr(sr);
3370 	Free(setlink);
3371 	return (rval);
3372 }
3373 
3374 /*
3375  * Make sure the name of the set is OK.
3376  */
3377 bool_t
3378 mdrpc_setnameok_common(
3379 	mdrpc_sp_args		*args,	/* device name */
3380 	mdrpc_bool_res		*res,
3381 	struct svc_req		*rqstp	/* RPC stuff */
3382 )
3383 {
3384 	md_error_t		*ep = &res->status;
3385 	int			err;
3386 	int			op_mode = R_OK;
3387 
3388 	/* setup, check permissions */
3389 	(void) memset(res, 0, sizeof (*res));
3390 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3391 		return (FALSE);
3392 	else if (err != 0)
3393 		return (TRUE);
3394 
3395 	if (check_set_lock(op_mode, NULL, ep))
3396 		return (TRUE);
3397 
3398 	/* doit */
3399 	res->value = setnameok(args->sp->setname, ep);
3400 
3401 	err = svc_fini(ep);
3402 
3403 	return (TRUE);
3404 }
3405 
3406 bool_t
3407 mdrpc_setnameok_1_svc(
3408 	mdrpc_sp_args		*args,	/* device name */
3409 	mdrpc_bool_res		*res,
3410 	struct svc_req		*rqstp	/* RPC stuff */
3411 )
3412 {
3413 	return (mdrpc_setnameok_common(args, res, rqstp));
3414 }
3415 
3416 bool_t
3417 mdrpc_setnameok_2_svc(
3418 	mdrpc_sp_2_args		*args,	/* device name */
3419 	mdrpc_bool_res		*res,
3420 	struct svc_req		*rqstp	/* RPC stuff */
3421 )
3422 {
3423 	switch (args->rev) {
3424 	    case MD_METAD_ARGS_REV_1:
3425 		return (mdrpc_setnameok_common(
3426 		    &args->mdrpc_sp_2_args_u.rev1, res, rqstp));
3427 	    default:
3428 		return (FALSE);
3429 	}
3430 }
3431 
3432 /*
3433  * determine if the setnumber we want to share is in use.
3434  */
3435 bool_t
3436 mdrpc_setnumbusy_common(
3437 	mdrpc_setno_args	*args,
3438 	mdrpc_bool_res		*res,
3439 	struct svc_req		*rqstp		/* RPC stuff */
3440 )
3441 {
3442 	md_error_t		*ep = &res->status;
3443 	md_set_record		*sr = NULL;
3444 	int			err;
3445 	int			op_mode = R_OK;
3446 
3447 	/* setup, check permissions */
3448 	(void) memset(res, 0, sizeof (*res));
3449 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3450 		return (FALSE);
3451 	else if (err != 0)
3452 		return (TRUE);
3453 
3454 	if (check_set_lock(op_mode, NULL, ep))
3455 		return (TRUE);
3456 
3457 	/* doit */
3458 	if ((sr = getsetbynum(args->setno, ep)) != NULL) {
3459 		res->value = TRUE;
3460 		free_sr(sr);
3461 		return (TRUE);
3462 	}
3463 	res->value = FALSE;
3464 	if (mdiserror(ep, MDE_NO_SET))
3465 		mdclrerror(ep);
3466 
3467 	err = svc_fini(ep);
3468 
3469 	return (TRUE);
3470 }
3471 
3472 bool_t
3473 mdrpc_setnumbusy_1_svc(
3474 	mdrpc_setno_args	*args,
3475 	mdrpc_bool_res		*res,
3476 	struct svc_req		*rqstp		/* RPC stuff */
3477 )
3478 {
3479 	return (mdrpc_setnumbusy_common(args, res, rqstp));
3480 }
3481 
3482 bool_t
3483 mdrpc_setnumbusy_2_svc(
3484 	mdrpc_setno_2_args	*args,
3485 	mdrpc_bool_res		*res,
3486 	struct svc_req		*rqstp		/* RPC stuff */
3487 )
3488 {
3489 	switch (args->rev) {
3490 	    case MD_METAD_ARGS_REV_1:
3491 		return (mdrpc_setnumbusy_common(
3492 		    &args->mdrpc_setno_2_args_u.rev1, res, rqstp));
3493 	    default:
3494 		return (FALSE);
3495 	}
3496 }
3497 
3498 static void
3499 stimeout(
3500 	mdsetname_t	*sp,
3501 	mhd_mhiargs_t	*mhiargsp,
3502 	int		version,	/* RPC version of calling routine */
3503 	md_error_t	*ep
3504 )
3505 {
3506 	mddb_userreq_t		req;
3507 	md_set_record		*sr;
3508 
3509 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
3510 		return;
3511 
3512 	sr->sr_mhiargs = *mhiargsp;
3513 
3514 	(void) memset(&req, '\0', sizeof (req));
3515 
3516 	METAD_SETUP_SR(MD_DB_SETDATA, sr->sr_selfid)
3517 	/* Do MN operation if rpc version supports it and if a MN set */
3518 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
3519 		req.ur_size = sizeof (struct md_mnset_record);
3520 	} else {
3521 		req.ur_size = sizeof (*sr);
3522 	}
3523 	req.ur_data = (uintptr_t)sr;
3524 
3525 	/*
3526 	 * Cluster nodename support
3527 	 * Convert nodename -> nodeid
3528 	 * Don't do this for MN disksets since we've already stored
3529 	 * both the nodeid and name.
3530 	 */
3531 	if ((version == METAD_VERSION) ||
3532 	    ((version == METAD_VERSION_DEVID) && (!(MD_MNSET_REC(sr)))))
3533 		sdssc_cm_sr_nm2nid(sr);
3534 
3535 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
3536 		(void) mdstealerror(ep, &req.ur_mde);
3537 		return;
3538 	}
3539 
3540 	(void) memset(&req, '\0', sizeof (req));
3541 	METAD_SETUP_SR(MD_DB_COMMIT_ONE, sr->sr_selfid)
3542 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0)
3543 		(void) mdstealerror(ep, &req.ur_mde);
3544 
3545 	/*
3546 	 * Cluster nodename support
3547 	 * Convert nodeid -> nodename
3548 	 * Don't do this for MN disksets since we've already stored
3549 	 * both the nodeid and name.
3550 	 */
3551 	if ((version == METAD_VERSION) ||
3552 	    ((version == METAD_VERSION_DEVID) && (!(MD_MNSET_REC(sr)))))
3553 		sdssc_cm_sr_nid2nm(sr);
3554 
3555 	free_sr(sr);
3556 }
3557 
3558 /*
3559  * Set MH ioctl timeout values.
3560  */
3561 bool_t
3562 mdrpc_stimeout_common(
3563 	mdrpc_stimeout_args	*args,
3564 	mdrpc_generic_res	*res,
3565 	struct svc_req		*rqstp,		/* RPC stuff */
3566 	int			version		/* RPC version */
3567 )
3568 {
3569 	md_error_t		*ep = &res->status;
3570 	int			err;
3571 	int			op_mode = W_OK;
3572 
3573 	/* setup, check permissions */
3574 	(void) memset(res, 0, sizeof (*res));
3575 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3576 		return (FALSE);
3577 	else if (err != 0)
3578 		return (TRUE);
3579 
3580 	if (check_set_lock(op_mode, NULL, ep))
3581 		return (TRUE);
3582 
3583 	/* doit */
3584 	stimeout(args->sp, args->mhiargsp, version, ep);
3585 
3586 	err = svc_fini(ep);
3587 
3588 	return (TRUE);
3589 }
3590 
3591 bool_t
3592 mdrpc_stimeout_1_svc(
3593 	mdrpc_stimeout_args	*args,
3594 	mdrpc_generic_res	*res,
3595 	struct svc_req		*rqstp		/* RPC stuff */
3596 )
3597 {
3598 	/* Pass RPC version (METAD_VERSION) to common routine */
3599 	return (mdrpc_stimeout_common(args, res, rqstp, METAD_VERSION));
3600 }
3601 
3602 bool_t
3603 mdrpc_stimeout_2_svc(
3604 	mdrpc_stimeout_2_args	*args,
3605 	mdrpc_generic_res	*res,
3606 	struct svc_req		*rqstp		/* RPC stuff */
3607 )
3608 {
3609 	switch (args->rev) {
3610 	    case MD_METAD_ARGS_REV_1:
3611 		/* Pass RPC version (METAD_VERSION_DEVID) to common routine */
3612 		return (mdrpc_stimeout_common(
3613 		    &args->mdrpc_stimeout_2_args_u.rev1, res,
3614 		    rqstp, METAD_VERSION_DEVID));
3615 	    default:
3616 		return (FALSE);
3617 	}
3618 }
3619 
3620 static void
3621 upd_dr_dbinfo(
3622 	mdsetname_t	*sp,
3623 	md_drive_desc	*dd,
3624 	md_error_t	*ep
3625 )
3626 {
3627 	mdsetname_t	*local_sp;
3628 	md_set_record	*sr;
3629 	md_drive_record	*dr;
3630 	md_drive_desc	*p;
3631 	mddrivename_t	*dn, *dn1;
3632 	ddi_devid_t	devid_remote = NULL;
3633 	ddi_devid_t	devid_local = NULL;
3634 	int		devid_same = -1;
3635 	side_t		sideno;
3636 	int		using_devid = 0;
3637 
3638 	if ((local_sp = metasetname(sp->setname, ep)) == NULL)
3639 		return;
3640 
3641 	metaflushsetname(local_sp);
3642 
3643 	if ((sideno = getmyside(local_sp, ep)) == MD_SIDEWILD)
3644 		return;
3645 
3646 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
3647 		return;
3648 
3649 	if (dd->dd_dnp == NULL)
3650 		return;
3651 
3652 	/*
3653 	 * The system is either all devid or all
3654 	 * non-devid so we determine this by looking
3655 	 * at the first item in the list.
3656 	 *
3657 	 * For did disks, the dd_dnp->devid is a valid pointer which
3658 	 * points to a '' string of devid.  We need to check this
3659 	 * before set the using_devid.
3660 	 */
3661 	if ((dd->dd_dnp->devid != NULL) && (dd->dd_dnp->devid[0] != '\0') &&
3662 	    (!(MD_MNSET_REC(sr))))
3663 		using_devid = 1;
3664 
3665 	for (p = dd; p != NULL; p = p->dd_next) {
3666 		dn = p->dd_dnp;
3667 		devid_remote = NULL;
3668 
3669 		if (dn->devid != NULL && (strlen(dn->devid) != 0) &&
3670 		    using_devid) {
3671 			/*
3672 			 * We have a devid so use it.
3673 			 */
3674 			(void) devid_str_decode(dn->devid, &devid_remote, NULL);
3675 		}
3676 
3677 		/* check to make sure using_devid agrees with reality... */
3678 		if ((using_devid == 1) && (devid_remote == NULL)) {
3679 			/* something went really wrong. Can't process */
3680 			(void) mddserror(ep, MDE_DS_INVALIDDEVID, sp->setno,
3681 			    mynode(), dn->cname, sp->setname);
3682 			return;
3683 		}
3684 
3685 		for (dr = sr->sr_drivechain; dr; dr = dr->dr_next) {
3686 			devid_same = -1;
3687 
3688 			dn1 = metadrivename_withdrkey(local_sp, sideno,
3689 			    dr->dr_key, MD_BASICNAME_OK, ep);
3690 
3691 			if (dn1 == NULL) {
3692 				if (devid_remote)
3693 					devid_free(devid_remote);
3694 				goto out;
3695 			}
3696 
3697 			if (dn1->devid != NULL && using_devid) {
3698 				if (devid_str_decode(dn1->devid, &devid_local,
3699 				    NULL) == 0) {
3700 					devid_same = devid_compare(devid_remote,
3701 					    devid_local);
3702 					devid_free(devid_local);
3703 				}
3704 			}
3705 
3706 			if (using_devid && devid_same == 0)
3707 				break;
3708 
3709 			if (!using_devid &&
3710 			    strcmp(dn->cname, dn1->cname) == 0)
3711 				break;
3712 		}
3713 
3714 		if (dr) {
3715 			/* Adjust the fields in the copy */
3716 			dr->dr_dbcnt = p->dd_dbcnt;
3717 			dr->dr_dbsize = p->dd_dbsize;
3718 		}
3719 		if (devid_remote)
3720 			devid_free(devid_remote);
3721 	}
3722 
3723 
3724 out:
3725 	commitset(sr, FALSE, ep);
3726 	free_sr(sr);
3727 }
3728 
3729 /*
3730  * update the database count and size field of drive records.
3731  */
3732 bool_t
3733 mdrpc_upd_dr_dbinfo_common(
3734 	mdrpc_drives_2_args_r1	*args,
3735 	mdrpc_generic_res	*res,
3736 	struct svc_req		*rqstp		/* RPC stuff */
3737 )
3738 {
3739 	md_error_t		*ep = &res->status;
3740 	int			err;
3741 	int			op_mode = W_OK;
3742 
3743 	/* setup, check permissions */
3744 	(void) memset(res, 0, sizeof (*res));
3745 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3746 		return (FALSE);
3747 	else if (err != 0)
3748 		return (TRUE);
3749 
3750 	if (check_set_lock(op_mode, args->cl_sk, ep))
3751 		return (TRUE);
3752 
3753 	/* doit */
3754 	upd_dr_dbinfo(args->sp, args->drivedescs, ep);
3755 
3756 	err = svc_fini(ep);
3757 
3758 	return (TRUE);
3759 }
3760 
3761 /*
3762  * version 1 of the remote procedure. This procedure is called if the
3763  * client is running in version 1. We first convert version 1 arguments
3764  * into version 2 arguments and then call the common remote procedure.
3765  */
3766 bool_t
3767 mdrpc_upd_dr_dbinfo_1_svc(
3768 	mdrpc_drives_args	*args,
3769 	mdrpc_generic_res	*res,
3770 	struct svc_req		*rqstp		/* RPC stuff */
3771 )
3772 {
3773 	bool_t			retval;
3774 	mdrpc_drives_2_args_r1	v2_args;
3775 
3776 	/* allocate memory */
3777 	alloc_newdrvdesc(args->drivedescs, &v2_args.drivedescs);
3778 
3779 	/* build args */
3780 	v2_args.cl_sk = args->cl_sk;
3781 	v2_args.sp = args->sp;
3782 	/* convert v1 args to v2 (revision 1) args */
3783 	meta_conv_drvdesc_old2new(args->drivedescs, v2_args.drivedescs);
3784 	v2_args.timestamp = args->timestamp;
3785 	v2_args.genid = args->genid;
3786 
3787 	retval = mdrpc_upd_dr_dbinfo_common(&v2_args, res, rqstp);
3788 
3789 	free_newdrvdesc(v2_args.drivedescs);
3790 
3791 	return (retval);
3792 }
3793 
3794 bool_t
3795 mdrpc_upd_dr_dbinfo_2_svc(
3796 	mdrpc_drives_2_args	*args,
3797 	mdrpc_generic_res	*res,
3798 	struct svc_req		*rqstp		/* RPC stuff */
3799 )
3800 {
3801 	switch (args->rev) {
3802 	    case MD_METAD_ARGS_REV_1:
3803 		return (mdrpc_upd_dr_dbinfo_common(
3804 		    &args->mdrpc_drives_2_args_u.rev1, res, rqstp));
3805 	    default:
3806 		return (FALSE);
3807 	}
3808 }
3809 
3810 static void
3811 upd_dr_flags(
3812 	mdsetname_t	*sp,
3813 	md_drive_desc	*dd,
3814 	uint_t		new_flags,
3815 	md_error_t	*ep
3816 )
3817 {
3818 	mdsetname_t	*local_sp;
3819 	md_set_record	*sr;
3820 	md_drive_record	*dr;
3821 	md_drive_desc	*p;
3822 	mddrivename_t	*dn, *dn1;
3823 	ddi_devid_t	devid_remote = NULL;
3824 	ddi_devid_t	devid_local = NULL;
3825 	int		devid_same = -1;
3826 	side_t		sideno;
3827 	int		using_devid = 0;
3828 
3829 	if ((local_sp = metasetname(sp->setname, ep)) == NULL)
3830 		return;
3831 
3832 	metaflushsetname(local_sp);
3833 
3834 	if ((sideno = getmyside(local_sp, ep)) == MD_SIDEWILD)
3835 		return;
3836 
3837 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
3838 		return;
3839 
3840 	if (dd->dd_dnp == NULL)
3841 		return;
3842 
3843 	/*
3844 	 * The system is either all devid or all
3845 	 * non-devid so we determine this by looking
3846 	 * at the first item in the list.
3847 	 *
3848 	 * For did disks, the dd_dnp->devid is a valid pointer which
3849 	 * points to a '' string of devid.  We need to check this
3850 	 * before set the using_devid.
3851 	 */
3852 	if ((dd->dd_dnp->devid != NULL) && (dd->dd_dnp->devid[0] != '\0') &&
3853 	    (!(MD_MNSET_REC(sr))))
3854 		using_devid = 1;
3855 
3856 	for (p = dd; p != NULL; p = p->dd_next) {
3857 		dn = p->dd_dnp;
3858 		devid_remote = NULL;
3859 
3860 		if (dn->devid != NULL && (strlen(dn->devid) != 0) &&
3861 		    using_devid) {
3862 			/*
3863 			 * We have a devid so use it.
3864 			 */
3865 			(void) devid_str_decode(dn->devid, &devid_remote, NULL);
3866 		}
3867 
3868 		/* check to make sure using_devid agrees with reality... */
3869 		if ((using_devid == 1) && (devid_remote == NULL)) {
3870 			/* something went really wrong. Can't process */
3871 			(void) mddserror(ep, MDE_DS_INVALIDDEVID, sp->setno,
3872 			    mynode(), dn->cname, sp->setname);
3873 			return;
3874 		}
3875 
3876 		for (dr = sr->sr_drivechain; dr; dr = dr->dr_next) {
3877 			devid_same = -1;
3878 
3879 			dn1 = metadrivename_withdrkey(local_sp, sideno,
3880 			    dr->dr_key, MD_BASICNAME_OK, ep);
3881 
3882 			if (dn1 == NULL) {
3883 				if (devid_remote)
3884 					devid_free(devid_remote);
3885 				goto out;
3886 			}
3887 
3888 			if (dn1->devid != NULL && using_devid) {
3889 				if (devid_str_decode(dn1->devid,
3890 				    &devid_local, NULL) == 0) {
3891 					devid_same = devid_compare(devid_remote,
3892 					    devid_local);
3893 					devid_free(devid_local);
3894 				}
3895 			}
3896 
3897 			if (using_devid && devid_same == 0)
3898 				break;
3899 
3900 			if (!using_devid &&
3901 			    strcmp(dn->cname, dn1->cname) == 0)
3902 				break;
3903 		}
3904 
3905 		if (dr)
3906 			dr->dr_flags = new_flags;
3907 		if (devid_remote)
3908 			devid_free(devid_remote);
3909 	}
3910 out:
3911 	commitset(sr, TRUE, ep);
3912 	free_sr(sr);
3913 }
3914 
3915 /*
3916  * update the database count and size field of drive records.
3917  */
3918 bool_t
3919 mdrpc_upd_dr_flags_common(
3920 	mdrpc_upd_dr_flags_2_args_r1	*args,
3921 	mdrpc_generic_res		*res,
3922 	struct svc_req			*rqstp		/* RPC stuff */
3923 )
3924 {
3925 	md_error_t		*ep = &res->status;
3926 	int			err;
3927 	int			op_mode = W_OK;
3928 
3929 	/* setup, check permissions */
3930 	(void) memset(res, 0, sizeof (*res));
3931 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
3932 		return (FALSE);
3933 	else if (err != 0)
3934 		return (TRUE);
3935 
3936 	if (check_set_lock(op_mode, args->cl_sk, ep))
3937 		return (TRUE);
3938 
3939 	/* doit */
3940 	upd_dr_flags(args->sp, args->drivedescs, args->new_flags, ep);
3941 
3942 	err = svc_fini(ep);
3943 
3944 	return (TRUE);
3945 }
3946 
3947 /*
3948  * version 1 of the remote procedure. This procedure is called if the
3949  * client is running in version 1. We first convert version 1 arguments
3950  * into version 2 arguments and then call the common remote procedure.
3951  */
3952 bool_t
3953 mdrpc_upd_dr_flags_1_svc(
3954 	mdrpc_upd_dr_flags_args	*args,
3955 	mdrpc_generic_res	*res,
3956 	struct svc_req		*rqstp		/* RPC stuff */
3957 )
3958 {
3959 	bool_t				retval;
3960 	mdrpc_upd_dr_flags_2_args_r1	v2_args;
3961 
3962 	/* allocate memory */
3963 	alloc_newdrvdesc(args->drivedescs, &v2_args.drivedescs);
3964 
3965 	/* build args */
3966 	v2_args.cl_sk = args->cl_sk;
3967 	v2_args.sp = args->sp;
3968 	/* convert v1 args to v2 (revision 1) args */
3969 	meta_conv_drvdesc_old2new(args->drivedescs, v2_args.drivedescs);
3970 	v2_args.new_flags = args->new_flags;
3971 
3972 	retval = mdrpc_upd_dr_flags_common(&v2_args, res, rqstp);
3973 
3974 	free_newdrvdesc(v2_args.drivedescs);
3975 
3976 	return (retval);
3977 }
3978 
3979 bool_t
3980 mdrpc_upd_dr_flags_2_svc(
3981 	mdrpc_upd_dr_flags_2_args	*args,
3982 	mdrpc_generic_res		*res,
3983 	struct svc_req			*rqstp		/* RPC stuff */
3984 )
3985 {
3986 	switch (args->rev) {
3987 	    case MD_METAD_ARGS_REV_1:
3988 		return (mdrpc_upd_dr_flags_common(
3989 		    &args->mdrpc_upd_dr_flags_2_args_u.rev1, res, rqstp));
3990 	    default:
3991 		return (FALSE);
3992 	}
3993 }
3994 
3995 static void
3996 upd_sr_flags(
3997 	mdsetname_t	*sp,
3998 	uint_t		new_flags,
3999 	md_error_t	*ep
4000 )
4001 {
4002 	md_set_record	*sr;
4003 
4004 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
4005 		return;
4006 
4007 	sr->sr_flags = new_flags;
4008 	commitset(sr, TRUE, ep);
4009 	free_sr(sr);
4010 }
4011 
4012 /*
4013  * update the set record flags
4014  */
4015 bool_t
4016 mdrpc_upd_sr_flags_common(
4017 	mdrpc_upd_sr_flags_args	*args,
4018 	mdrpc_generic_res	*res,
4019 	struct svc_req		*rqstp		/* RPC stuff */
4020 )
4021 {
4022 	md_error_t		*ep = &res->status;
4023 	int			err;
4024 	int			op_mode = W_OK;
4025 
4026 	/* setup, check permissions */
4027 	(void) memset(res, 0, sizeof (*res));
4028 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4029 		return (FALSE);
4030 	else if (err != 0)
4031 		return (TRUE);
4032 
4033 	if (check_set_lock(op_mode, args->cl_sk, ep))
4034 		return (TRUE);
4035 
4036 	/* doit */
4037 	upd_sr_flags(args->sp, args->new_flags, ep);
4038 
4039 	err = svc_fini(ep);
4040 
4041 	return (TRUE);
4042 }
4043 
4044 bool_t
4045 mdrpc_upd_sr_flags_1_svc(
4046 	mdrpc_upd_sr_flags_args	*args,
4047 	mdrpc_generic_res	*res,
4048 	struct svc_req		*rqstp		/* RPC stuff */
4049 )
4050 {
4051 	return (mdrpc_upd_sr_flags_common(args, res, rqstp));
4052 }
4053 
4054 bool_t
4055 mdrpc_upd_sr_flags_2_svc(
4056 	mdrpc_upd_sr_flags_2_args	*args,
4057 	mdrpc_generic_res		*res,
4058 	struct svc_req			*rqstp		/* RPC stuff */
4059 )
4060 {
4061 	switch (args->rev) {
4062 	    case MD_METAD_ARGS_REV_1:
4063 		return (mdrpc_upd_sr_flags_common(
4064 		    &args->mdrpc_upd_sr_flags_2_args_u.rev1, res, rqstp));
4065 	    default:
4066 		return (FALSE);
4067 	}
4068 }
4069 
4070 /*
4071  * upd_nr_flags updates the node records stored in this node's local mddb
4072  * given a node desciptor list and an action.  upd_nr_flags then commits
4073  * the node records to the local mddb.
4074  *
4075  * nd - A linked list of node descriptors that describes the node records
4076  *	in this diskset on which the action applies.
4077  * flag_action: action to be taken on node records that match the nd list.
4078  *	flag_action can be:
4079  *		MD_NR_JOIN: set OWN flag in node records
4080  *		MD_NR_WITHDRAW: reset OWN flag in node records
4081  *		MD_NR_OK: reset ADD flags and set OK flag in node records
4082  *		MD_NR_SET: set node record flags based on flags stored in nd
4083  *
4084  * Typically, the JOIN, WITHDRAW and OK flag_actions are used when setting
4085  * all nodes in a diskset to JOIN (add first disk to set), WITHDRAW
4086  * (remove last disk from set) or OK (after addition of host to set).
4087  *
4088  * The SET flag_action is typically used when nodelist contains all nodes
4089  * in the diskset, but specific nodes have had flag changes.  An example of
4090  * this would be the join/withdraw of a specific node to/from the set.
4091  *
4092  * Ignore the MD_MN_NODE_RB_JOIN flag if set in node record flag.  This
4093  * flag is used by the client to recover in case of failure and should not
4094  * be set in the node record flags.
4095  */
4096 static void
4097 upd_nr_flags(
4098 	mdsetname_t	*sp,
4099 	md_mnnode_desc	*nd,
4100 	uint_t		flag_action,
4101 	md_error_t	*ep
4102 )
4103 {
4104 	mdsetname_t		*local_sp;
4105 	md_set_record		*sr;
4106 	md_mnset_record		*mnsr;
4107 	md_mnnode_desc		*ndp;
4108 	md_mnnode_record	*nrp;
4109 
4110 	if ((local_sp = metasetname(sp->setname, ep)) == NULL)
4111 		return;
4112 
4113 	metaflushsetname(local_sp);
4114 
4115 	if ((sr = getsetbyname(sp->setname, ep)) == NULL)
4116 		return;
4117 
4118 	if (!(MD_MNSET_REC(sr))) {
4119 		return;
4120 	}
4121 	mnsr = (struct md_mnset_record *)sr;
4122 
4123 	switch (flag_action) {
4124 	case MD_NR_JOIN:
4125 	case MD_NR_WITHDRAW:
4126 	case MD_NR_SET:
4127 	case MD_NR_OK:
4128 	case MD_NR_DEL:
4129 		break;
4130 	default:
4131 		return;
4132 	}
4133 
4134 	for (ndp = nd; ndp != NULL; ndp = ndp->nd_next) {
4135 		/* Find matching node record for given node descriptor */
4136 		for (nrp = mnsr->sr_nodechain; nrp != NULL;
4137 		    nrp = nrp->nr_next) {
4138 			if (ndp->nd_nodeid == nrp->nr_nodeid) {
4139 				switch (flag_action) {
4140 				case MD_NR_JOIN:
4141 					nrp->nr_flags |= MD_MN_NODE_OWN;
4142 					break;
4143 				case MD_NR_WITHDRAW:
4144 					nrp->nr_flags &= ~MD_MN_NODE_OWN;
4145 					break;
4146 				case MD_NR_OK:
4147 					nrp->nr_flags &=
4148 					    ~(MD_MN_NODE_ADD | MD_MN_NODE_DEL);
4149 					nrp->nr_flags |= MD_MN_NODE_OK;
4150 					break;
4151 				case MD_NR_DEL:
4152 					nrp->nr_flags &=
4153 					    ~(MD_MN_NODE_OK | MD_MN_NODE_ADD);
4154 					nrp->nr_flags |= MD_MN_NODE_DEL;
4155 					break;
4156 				case MD_NR_SET:
4157 					/* Do not set RB_JOIN flag */
4158 					nrp->nr_flags =
4159 					    ndp->nd_flags & ~MD_MN_NODE_RB_JOIN;
4160 					break;
4161 				}
4162 				break;
4163 			}
4164 		}
4165 	}
4166 out:
4167 	/* Don't increment set genid for node record flag update */
4168 	commitset(sr, FALSE, ep);
4169 	free_sr(sr);
4170 }
4171 
4172 /*
4173  * init/fini wrapper around upd_nr_flags
4174  */
4175 bool_t
4176 mdrpc_upd_nr_flags_common(
4177 	mdrpc_upd_nr_flags_args	*args,
4178 	mdrpc_generic_res	*res,
4179 	struct svc_req		*rqstp		/* RPC stuff */
4180 )
4181 {
4182 	md_error_t		*ep = &res->status;
4183 	int			err;
4184 	int			op_mode = W_OK;
4185 
4186 	/* setup, check permissions */
4187 	(void) memset(res, 0, sizeof (*res));
4188 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4189 		return (FALSE);
4190 	else if (err != 0)
4191 		return (TRUE);
4192 
4193 	/*
4194 	 * During reconfig, node record flags can be updated without
4195 	 * locking first.
4196 	 */
4197 	if (!(args->flags & MNSET_IN_RECONFIG)) {
4198 		if (check_set_lock(op_mode, args->cl_sk, ep))
4199 			return (TRUE);
4200 	}
4201 
4202 	/* doit */
4203 	upd_nr_flags(args->sp, args->nodedescs, args->flag_action, ep);
4204 
4205 	err = svc_fini(ep);
4206 
4207 	return (TRUE);
4208 }
4209 
4210 /*
4211  * update the node records using given flag action.
4212  */
4213 bool_t
4214 mdrpc_upd_nr_flags_2_svc(
4215 	mdrpc_upd_nr_flags_2_args	*args,
4216 	mdrpc_generic_res		*res,
4217 	struct svc_req			*rqstp		/* RPC stuff */
4218 )
4219 {
4220 	switch (args->rev) {
4221 	    case MD_METAD_ARGS_REV_1:
4222 		return (mdrpc_upd_nr_flags_common(
4223 		    &args->mdrpc_upd_nr_flags_2_args_u.rev1, res, rqstp));
4224 	    default:
4225 		return (FALSE);
4226 	}
4227 }
4228 
4229 void
4230 free_sk(md_setkey_t *skp)
4231 {
4232 	Free(skp->sk_setname);
4233 	Free(skp->sk_host);
4234 	Free(skp);
4235 }
4236 
4237 void
4238 del_sk(set_t setno)
4239 {
4240 	md_setkey_t	*skp;
4241 	md_setkey_t	*tskp;
4242 
4243 	for (skp = tskp = my_svc_sk; skp; tskp = skp, skp = skp->sk_next) {
4244 		if (setno == skp->sk_setno) {
4245 			if (skp == my_svc_sk)
4246 				my_svc_sk = skp->sk_next;
4247 			else
4248 				tskp->sk_next = skp->sk_next;
4249 
4250 			Free(skp->sk_setname);
4251 			Free(skp->sk_host);
4252 			Free(skp);
4253 			break;
4254 		}
4255 	}
4256 }
4257 
4258 md_setkey_t *
4259 dupsk(md_setkey_t *skp)
4260 {
4261 	md_setkey_t	*tskp;
4262 
4263 	tskp = Zalloc(sizeof (md_setkey_t));
4264 
4265 	*tskp = *skp;
4266 	tskp->sk_host = Strdup(skp->sk_host);
4267 	tskp->sk_setname = Strdup(skp->sk_setname);
4268 
4269 	return (tskp);
4270 }
4271 
4272 md_setkey_t *
4273 svc_get_setkey(set_t setno)
4274 {
4275 	md_setkey_t	*skp;
4276 
4277 	for (skp = my_svc_sk; skp != NULL; skp = skp->sk_next)
4278 		if (setno == skp->sk_setno)
4279 			return (dupsk(skp));
4280 	return (NULL);
4281 }
4282 
4283 void
4284 svc_set_setkey(md_setkey_t *svc_sk)
4285 {
4286 	md_setkey_t	*skp;
4287 
4288 	if (my_svc_sk == NULL) {
4289 		my_svc_sk = dupsk(svc_sk);
4290 		return;
4291 	}
4292 
4293 	for (skp = my_svc_sk; skp->sk_next != NULL; skp = skp->sk_next)
4294 		assert(svc_sk->sk_setno != skp->sk_setno);
4295 
4296 	skp->sk_next = dupsk(svc_sk);
4297 }
4298 
4299 /*
4300  * Unlock the set
4301  *
4302  * To unlock the set, the user must have the correct key, once this is verified
4303  * the set is unlocked and the cached information for the set is flushed.
4304  */
4305 bool_t
4306 mdrpc_unlock_set_common(
4307 	mdrpc_null_args		*args,
4308 	mdrpc_setlock_res	*res,
4309 	struct svc_req		*rqstp		/* RPC stuff */
4310 )
4311 {
4312 	md_error_t		*ep = &res->status;
4313 	int			err;
4314 	int			op_mode = W_OK;
4315 	md_setkey_t		*svc_skp;
4316 	md_set_desc		*sd;
4317 	mdsetname_t		*sp;
4318 	int			multi_node = 0;
4319 	md_error_t		xep = mdnullerror;
4320 
4321 	/* setup, check permissions */
4322 	(void) memset(res, 0, sizeof (*res));
4323 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4324 		return (FALSE);
4325 	else if (err != 0)
4326 		return (TRUE);
4327 
4328 	/*
4329 	 * Is diskset a MN diskset?
4330 	 * Don't set error from this check since unlock set can be
4331 	 * called after a set has been deleted.
4332 	 */
4333 	if (((sp = metasetnosetname(args->cl_sk->sk_setno, &xep)) != NULL) &&
4334 	    ((sd = metaget_setdesc(sp, &xep)) != NULL)) {
4335 		if ((MD_MNSET_DESC(sd))) {
4336 			multi_node = 1;
4337 		}
4338 	}
4339 
4340 	/* Get the set key, if any */
4341 	svc_skp = svc_get_setkey(args->cl_sk->sk_setno);
4342 
4343 	/* The set is locked */
4344 	if (svc_skp != NULL) {
4345 
4346 		/* Make sure the opener has the right key. */
4347 		if (args->cl_sk->sk_key.tv_sec != svc_skp->sk_key.tv_sec ||
4348 		    args->cl_sk->sk_key.tv_usec != svc_skp->sk_key.tv_usec) {
4349 			(void) mddserror(ep, MDE_DS_ULKSBADKEY,
4350 			    svc_skp->sk_setno, mynode(), svc_skp->sk_host,
4351 			    svc_skp->sk_setname);
4352 			free_sk(svc_skp);
4353 			return (TRUE);
4354 		}
4355 
4356 		/* Unlock the set */
4357 		del_sk(args->cl_sk->sk_setno);
4358 
4359 		/* Cleanup */
4360 		free_sk(svc_skp);
4361 
4362 		goto out;
4363 	}
4364 
4365 
4366 	/*
4367 	 * It is possible on a MN diskset to attempt to unlock a set that
4368 	 * is unlocked.  This could occur when the metaset or metadb  command
4369 	 * is failing due to another metaset or metadb command running.
4370 	 * So, print no warning for MN disksets.
4371 	 */
4372 	if (multi_node == 0) {
4373 		md_eprintf("Warning: set unlocked when unlock_set called!\n");
4374 	}
4375 
4376 out:
4377 	res->cl_sk = svc_get_setkey(args->cl_sk->sk_setno);
4378 
4379 	/* Flush the set cache */
4380 	sr_cache_flush_setno(args->cl_sk->sk_setno);
4381 
4382 	return (TRUE);
4383 }
4384 
4385 bool_t
4386 mdrpc_unlock_set_1_svc(
4387 	mdrpc_null_args		*args,
4388 	mdrpc_setlock_res	*res,
4389 	struct svc_req		*rqstp		/* RPC stuff */
4390 )
4391 {
4392 	return (mdrpc_unlock_set_common(args, res, rqstp));
4393 }
4394 
4395 bool_t
4396 mdrpc_unlock_set_2_svc(
4397 	mdrpc_null_args		*args,
4398 	mdrpc_setlock_res	*res,
4399 	struct svc_req		*rqstp		/* RPC stuff */
4400 )
4401 {
4402 	return (mdrpc_unlock_set_common(args, res, rqstp));
4403 }
4404 
4405 /*
4406  * Lock the set
4407  *
4408  * If the user does not hand us a key, then we generate a new key and lock the
4409  * set using this new key that was generated, if the user hands us a key then
4410  * we use the key to lock the set.
4411  */
4412 bool_t
4413 mdrpc_lock_set_common(
4414 	mdrpc_null_args		*args,
4415 	mdrpc_setlock_res	*res,
4416 	struct svc_req		*rqstp		/* RPC stuff */
4417 )
4418 {
4419 	md_error_t		*ep = &res->status;
4420 	int			err;
4421 	md_error_t		xep = mdnullerror;
4422 	int			op_mode = W_OK;
4423 	md_setkey_t		*svc_skp;
4424 	md_setkey_t		new_sk;
4425 	md_set_desc		*sd;
4426 	mdsetname_t		*sp;
4427 
4428 	/* setup, check permissions */
4429 	(void) memset(res, 0, sizeof (*res));
4430 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4431 		return (FALSE);
4432 	else if (err != 0)
4433 		return (TRUE);
4434 
4435 	svc_skp = svc_get_setkey(args->cl_sk->sk_setno);
4436 
4437 	/* The set is unlocked */
4438 	if (svc_skp == NULL) {
4439 		/* If we have been given a key, use it. */
4440 		if (args->cl_sk->sk_key.tv_sec || args->cl_sk->sk_key.tv_usec) {
4441 			svc_set_setkey(args->cl_sk);
4442 			res->cl_sk = svc_get_setkey(args->cl_sk->sk_setno);
4443 			goto out;
4444 		}
4445 
4446 		/* We need to lock it, with a new key */
4447 		new_sk = *args->cl_sk;
4448 		if (meta_gettimeofday(&new_sk.sk_key) == -1) {
4449 			(void) mdsyserror(ep, errno, "meta_gettimeofday()");
4450 			mde_perror(&xep, "");
4451 			md_exit(NULL, 1);
4452 		}
4453 		svc_set_setkey(&new_sk);
4454 
4455 		res->cl_sk = svc_get_setkey(args->cl_sk->sk_setno);
4456 		goto out;
4457 	}
4458 
4459 	/*
4460 	 * If a MN diskset, the lock_set routine is used as a locking
4461 	 * mechanism to keep multiple metaset and/or metadb commads
4462 	 * from interfering with each other.  If two metaset/metadb
4463 	 * commands are issued at the same time - one will complete
4464 	 * and the other command will fail with MDE_DS_NOTNOW_CMD.
4465 	 */
4466 	if (((sp = metasetnosetname(args->cl_sk->sk_setno, ep)) != NULL) &&
4467 	    ((sd = metaget_setdesc(sp, ep)) != NULL) &&
4468 	    (MD_MNSET_DESC(sd))) {
4469 		(void) mddserror(ep, MDE_DS_NOTNOW_CMD,
4470 		    svc_skp->sk_setno, mynode(),
4471 		    svc_skp->sk_host, svc_skp->sk_setname);
4472 		goto out;
4473 	}
4474 
4475 	md_eprintf("Warning: set locked when lock_set called!\n");
4476 
4477 	md_eprintf("Lock info:\n");
4478 
4479 	md_eprintf("\tLock(svc):\n");
4480 	md_eprintf("\t\tSetname: %s\n", svc_skp->sk_setname);
4481 	md_eprintf("\t\tSetno:   %d\n", svc_skp->sk_setno);
4482 	md_eprintf("\t\tHost:    %s\n", svc_skp->sk_host);
4483 	md_eprintf("\t\tKey:     %d/%d %s",
4484 	    svc_skp->sk_key.tv_sec, svc_skp->sk_key.tv_usec,
4485 	    ctime((const time_t *)&svc_skp->sk_key.tv_sec));
4486 
4487 	md_eprintf("\tLock(cl):\n");
4488 	md_eprintf("\t\tSetname: %s\n", args->cl_sk->sk_setname);
4489 	md_eprintf("\t\tSetno:   %d\n", args->cl_sk->sk_setno);
4490 	md_eprintf("\t\tHost:    %s\n", args->cl_sk->sk_host);
4491 	md_eprintf("\t\tKey:     %d/%d %s",
4492 	    args->cl_sk->sk_key.tv_sec, args->cl_sk->sk_key.tv_usec,
4493 	    ctime((const time_t *)&args->cl_sk->sk_key.tv_sec));
4494 
4495 	/* The set is locked, do we have the key? */
4496 	if (args->cl_sk->sk_key.tv_sec == svc_skp->sk_key.tv_sec &&
4497 	    args->cl_sk->sk_key.tv_usec == svc_skp->sk_key.tv_usec) {
4498 		res->cl_sk = svc_get_setkey(args->cl_sk->sk_setno);
4499 		goto out;
4500 	}
4501 
4502 	/*
4503 	 * The set is locked and we do not have the key, so we set up an error.
4504 	 */
4505 	(void) mddserror(ep, MDE_DS_LKSBADKEY, svc_skp->sk_setno, mynode(),
4506 	    svc_skp->sk_host, args->cl_sk->sk_setname);
4507 
4508 out:
4509 	if (svc_skp != NULL)
4510 		free_sk(svc_skp);
4511 
4512 	/* Flush the set cache */
4513 	sr_cache_flush_setno(args->cl_sk->sk_setno);
4514 
4515 	return (TRUE);
4516 }
4517 
4518 bool_t
4519 mdrpc_lock_set_1_svc(
4520 	mdrpc_null_args		*args,
4521 	mdrpc_setlock_res	*res,
4522 	struct svc_req		*rqstp		/* RPC stuff */
4523 )
4524 {
4525 	return (mdrpc_lock_set_common(args, res, rqstp));
4526 }
4527 
4528 bool_t
4529 mdrpc_lock_set_2_svc(
4530 	mdrpc_null_args		*args,
4531 	mdrpc_setlock_res	*res,
4532 	struct svc_req		*rqstp		/* RPC stuff */
4533 )
4534 {
4535 	return (mdrpc_lock_set_common(args, res, rqstp));
4536 }
4537 
4538 static void
4539 updmeds(
4540 	char		*setname,
4541 	md_h_arr_t	*medp,
4542 	int		version,	/* RPC version of calling routine */
4543 	md_error_t	*ep
4544 )
4545 {
4546 	mddb_userreq_t		req;
4547 	md_set_record		*sr;
4548 	mddb_med_parm_t		mp;
4549 
4550 	if ((sr = getsetbyname(setname, ep)) == NULL)
4551 		return;
4552 
4553 	sr->sr_med = *medp;			/* structure assignment */
4554 
4555 	(void) memset(&req, '\0', sizeof (req));
4556 
4557 	METAD_SETUP_SR(MD_DB_SETDATA, sr->sr_selfid)
4558 	/* Do MN operation if rpc version supports it and if a MN set */
4559 	if ((version != METAD_VERSION) && (MD_MNSET_REC(sr))) {
4560 		req.ur_size = sizeof (struct md_mnset_record);
4561 	} else {
4562 		req.ur_size = sizeof (*sr);
4563 	}
4564 	req.ur_data = (uintptr_t)sr;
4565 	if (metaioctl(MD_DB_USERREQ, &req, &req.ur_mde, NULL) != 0) {
4566 		(void) mdstealerror(ep, &req.ur_mde);
4567 		free_sr(sr);
4568 		return;
4569 	}
4570 
4571 	commitset(sr, TRUE, ep);
4572 
4573 	/*
4574 	 * If a MN disket, send the mediator list to the kernel.
4575 	 */
4576 	if (MD_MNSET_REC(sr)) {
4577 		(void) memset(&mp, '\0', sizeof (mddb_med_parm_t));
4578 		mp.med_setno = sr->sr_setno;
4579 		if (meta_h2hi(medp, &mp.med, ep)) {
4580 			free_sr(sr);
4581 			return;
4582 		}
4583 
4584 		/* Resolve the IP addresses for the host list */
4585 		if (meta_med_hnm2ip(&mp.med, ep)) {
4586 			free_sr(sr);
4587 			return;
4588 		}
4589 
4590 		/* If node not yet joined to set, failure is ok. */
4591 		if (metaioctl(MD_MED_SET_LST, &mp, &mp.med_mde, NULL) != 0) {
4592 			if (!mdismddberror(&mp.med_mde, MDE_DB_NOTOWNER)) {
4593 				(void) mdstealerror(ep, &mp.med_mde);
4594 			}
4595 		}
4596 	}
4597 	free_sr(sr);
4598 }
4599 
4600 /*
4601  * Update the mediator data in the set record
4602  */
4603 bool_t
4604 mdrpc_updmeds_common(
4605 	mdrpc_updmeds_args	*args,
4606 	mdrpc_generic_res	*res,
4607 	struct svc_req		*rqstp,		/* RPC stuff */
4608 	int			version		/* RPC version */
4609 )
4610 {
4611 	md_error_t		*ep = &res->status;
4612 	int			err;
4613 	int			op_mode = W_OK;
4614 
4615 	/* setup, check permissions */
4616 	(void) memset(res, 0, sizeof (*res));
4617 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4618 		return (FALSE);
4619 	else if (err != 0)
4620 		return (TRUE);
4621 
4622 	if (check_set_lock(op_mode, args->cl_sk, ep))
4623 		return (TRUE);
4624 
4625 	/* doit */
4626 	updmeds(args->sp->setname, &args->meds, version, ep);
4627 
4628 	err = svc_fini(ep);
4629 
4630 	return (TRUE);
4631 }
4632 
4633 bool_t
4634 mdrpc_updmeds_1_svc(
4635 	mdrpc_updmeds_args	*args,
4636 	mdrpc_generic_res	*res,
4637 	struct svc_req		*rqstp		/* RPC stuff */
4638 )
4639 {
4640 	/* Pass RPC version (METAD_VERSION) to common routine */
4641 	return (mdrpc_updmeds_common(args, res, rqstp, METAD_VERSION));
4642 }
4643 
4644 bool_t
4645 mdrpc_updmeds_2_svc(
4646 	mdrpc_updmeds_2_args	*args,
4647 	mdrpc_generic_res	*res,
4648 	struct svc_req		*rqstp		/* RPC stuff */
4649 )
4650 {
4651 	switch (args->rev) {
4652 	    case MD_METAD_ARGS_REV_1:
4653 		/* Pass RPC version (METAD_VERSION_DEVID) to common routine */
4654 		return (mdrpc_updmeds_common(
4655 		    &args->mdrpc_updmeds_2_args_u.rev1, res,
4656 		    rqstp, METAD_VERSION_DEVID));
4657 	    default:
4658 		return (FALSE);
4659 	}
4660 }
4661 
4662 /*
4663  * Call routines to suspend, reinit and resume mdcommd.
4664  * Called during metaset and metadb command.
4665  * NOT called during reconfig cycle.
4666  */
4667 bool_t
4668 mdrpc_mdcommdctl_2_svc(
4669 	mdrpc_mdcommdctl_2_args	*args,
4670 	mdrpc_generic_res	*res,
4671 	struct svc_req		*rqstp		/* RPC stuff */
4672 )
4673 {
4674 	mdrpc_mdcommdctl_args	*args_cc;
4675 	md_error_t		*ep = &res->status;
4676 	int			err;
4677 	int			op_mode = R_OK;
4678 	int			suspend_ret;
4679 
4680 	switch (args->rev) {
4681 	    case MD_METAD_ARGS_REV_1:
4682 		/* setup, check permissions */
4683 		(void) memset(res, 0, sizeof (*res));
4684 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4685 			return (FALSE);
4686 		else if (err != 0)
4687 			return (TRUE);
4688 
4689 		args_cc = &(args->mdrpc_mdcommdctl_2_args_u.rev1);
4690 		switch (args_cc->flag_action) {
4691 			case COMMDCTL_SUSPEND:
4692 				suspend_ret = mdmn_suspend(args_cc->setno,
4693 				    args_cc->class);
4694 				if (suspend_ret != 0) {
4695 					(void) mddserror(ep, suspend_ret,
4696 					    args_cc->setno, mynode(),
4697 					    NULL, mynode());
4698 				}
4699 				break;
4700 			case COMMDCTL_RESUME:
4701 				if (mdmn_resume(args_cc->setno,
4702 				    args_cc->class, args_cc->flags)) {
4703 					(void) mddserror(ep,
4704 					    MDE_DS_COMMDCTL_RESUME_FAIL,
4705 					    args_cc->setno, mynode(),
4706 					    NULL, mynode());
4707 				}
4708 				break;
4709 			case COMMDCTL_REINIT:
4710 				if (mdmn_reinit_set(args_cc->setno)) {
4711 					(void) mddserror(ep,
4712 					    MDE_DS_COMMDCTL_REINIT_FAIL,
4713 					    args_cc->setno, mynode(),
4714 					    NULL, mynode());
4715 				}
4716 				break;
4717 		}
4718 		err = svc_fini(ep);
4719 		return (TRUE);
4720 
4721 	    default:
4722 		return (FALSE);
4723 	}
4724 }
4725 
4726 /*
4727  * Return TRUE if set is stale.
4728  */
4729 bool_t
4730 mdrpc_mn_is_stale_2_svc(
4731 	mdrpc_setno_2_args	*args,
4732 	mdrpc_bool_res		*res,
4733 	struct svc_req		*rqstp		/* RPC stuff */
4734 )
4735 {
4736 	md_error_t	*ep = &res->status;
4737 	mddb_config_t	c;
4738 	int		err;
4739 	int		op_mode = R_OK;
4740 
4741 	(void) memset(&c, 0, sizeof (c));
4742 	switch (args->rev) {
4743 	    case MD_METAD_ARGS_REV_1:
4744 		c.c_id = 0;
4745 		c.c_setno = args->mdrpc_setno_2_args_u.rev1.setno;
4746 
4747 		/* setup, check permissions */
4748 		(void) memset(res, 0, sizeof (*res));
4749 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4750 			return (FALSE);
4751 		else if (err != 0)
4752 			return (TRUE);
4753 
4754 		if (metaioctl(MD_DB_GETDEV, &c, &c.c_mde, NULL) != 0) {
4755 			mdstealerror(ep, &c.c_mde);
4756 			return (TRUE);
4757 		}
4758 
4759 		if (c.c_flags & MDDB_C_STALE) {
4760 			res->value = TRUE;
4761 		} else {
4762 			res->value = FALSE;
4763 		}
4764 
4765 		err = svc_fini(ep);
4766 		return (TRUE);
4767 
4768 	    default:
4769 		return (FALSE);
4770 	}
4771 }
4772 
4773 /*
4774  * Clear out all clnt_locks held by all MN disksets.
4775  * This is only used during a reconfig cycle.
4776  */
4777 /* ARGSUSED */
4778 int
4779 mdrpc_clr_mnsetlock_2_svc(
4780 	mdrpc_null_args		*args,
4781 	mdrpc_generic_res	*res,
4782 	struct svc_req		*rqstp		/* RPC stuff */
4783 )
4784 {
4785 	set_t			max_sets, setno;
4786 	md_error_t		*ep = &res->status;
4787 	int			err;
4788 	int			op_mode = W_OK;
4789 	mdsetname_t		*sp;
4790 
4791 	/* setup, check permissions */
4792 	(void) memset(res, 0, sizeof (*res));
4793 
4794 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4795 		return (FALSE);
4796 	else if (err != 0)
4797 		return (TRUE);
4798 
4799 	/*
4800 	 * Walk through all possible disksets.
4801 	 * For each MN set, delete all keys associated with that set.
4802 	 */
4803 	if ((max_sets = get_max_sets(ep)) == 0) {
4804 		return (TRUE);
4805 	}
4806 
4807 	/* start walking through all possible disksets */
4808 	for (setno = 1; setno < max_sets; setno++) {
4809 		if ((sp = metasetnosetname(setno, ep)) == NULL) {
4810 			if (mdiserror(ep, MDE_NO_SET)) {
4811 				/* No set for this setno - continue */
4812 				mdclrerror(ep);
4813 				continue;
4814 			} else {
4815 				mde_perror(ep, gettext(
4816 				    "Unable to get set %s information"),
4817 				    sp->setname);
4818 				mdclrerror(ep);
4819 				continue;
4820 			}
4821 		}
4822 
4823 		/* only check multi-node disksets */
4824 		if (!meta_is_mn_set(sp, ep)) {
4825 			mdclrerror(ep);
4826 			continue;
4827 		}
4828 
4829 		/* Delete keys associated with rpc.metad clnt_lock */
4830 		del_sk(setno);
4831 	}
4832 
4833 	*ep = mdnullerror;
4834 
4835 	err = svc_fini(ep);
4836 
4837 	return (TRUE);
4838 }
4839 
4840 /*
4841  * Get drive desc on this host for given setno.
4842  * This is only used during a reconfig cycle.
4843  * Returns a drive desc structure for the given mdsetname
4844  * from this host.
4845  *
4846  * Returned drive desc structure is partially filled in with
4847  * the drive name but is not filled in with any other strings
4848  * in the drivename structure.
4849  */
4850 bool_t
4851 mdrpc_getdrivedesc_2_svc(
4852 	mdrpc_sp_2_args		*args,
4853 	mdrpc_getdrivedesc_res 	*res,
4854 	struct svc_req		*rqstp		/* RPC stuff */
4855 )
4856 {
4857 	md_drive_desc		*dd;
4858 	md_error_t		*ep = &res->status;
4859 	int			err;
4860 	int			op_mode = R_OK;
4861 	mdsetname_t		*my_sp;
4862 	mdrpc_sp_args		*args_r1;
4863 
4864 	switch (args->rev) {
4865 	    case MD_METAD_ARGS_REV_1:
4866 		/* setup, check permissions */
4867 		(void) memset(res, 0, sizeof (*res));
4868 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4869 			return (FALSE);
4870 		else if (err != 0)
4871 			return (TRUE);
4872 
4873 		/* doit */
4874 		args_r1 = &args->mdrpc_sp_2_args_u.rev1;
4875 		if ((my_sp = metasetname(args_r1->sp->setname, ep)) == NULL)
4876 			return (TRUE);
4877 
4878 		dd = metaget_drivedesc(my_sp,
4879 			(MD_BASICNAME_OK | PRINT_FAST), ep);
4880 
4881 		res->dd = dd_list_dup(dd);
4882 
4883 		err = svc_fini(ep);
4884 
4885 		return (TRUE);
4886 	    default:
4887 		return (FALSE);
4888 	}
4889 }
4890 
4891 /*
4892  * Update drive records given list from master during reconfig.
4893  * Make this node's list match the master's list which may include
4894  * deleting a drive record that is known by this node and not known
4895  * by the master node.
4896  *
4897  * Sync up the set/node/drive record genids to match the genid
4898  * passed in the dd structure (all genids in this structure
4899  * are the same).
4900  */
4901 bool_t
4902 mdrpc_upd_dr_reconfig_common(
4903 	mdrpc_upd_dr_flags_2_args_r1	*args,
4904 	mdrpc_generic_res		*res,
4905 	struct svc_req			*rqstp		/* RPC stuff */
4906 )
4907 {
4908 	md_error_t		*ep = &res->status;
4909 	int			err;
4910 	mdsetname_t		*local_sp;
4911 	md_set_record		*sr;
4912 	md_mnset_record		*mnsr;
4913 	md_drive_record		*dr, *dr_placeholder = NULL;
4914 	md_drive_desc		*dd;
4915 	mddrivename_t		*dn, *dn1;
4916 	side_t			sideno;
4917 	md_mnnode_record	*nrp;
4918 	int			op_mode = W_OK;
4919 	int			change = 0;
4920 
4921 	/* setup, check permissions */
4922 	(void) memset(res, 0, sizeof (*res));
4923 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
4924 		return (FALSE);
4925 	else if (err != 0)
4926 		return (TRUE);
4927 
4928 	if ((local_sp = metasetname(args->sp->setname, ep)) == NULL)
4929 		return (TRUE);
4930 
4931 	metaflushsetname(local_sp);
4932 
4933 	if ((sideno = getmyside(local_sp, ep)) == MD_SIDEWILD)
4934 		return (TRUE);
4935 
4936 	if ((sr = getsetbyname(args->sp->setname, ep)) == NULL)
4937 		return (TRUE);
4938 
4939 	if (!(MD_MNSET_REC(sr))) {
4940 		free_sr(sr);
4941 		return (TRUE);
4942 	}
4943 
4944 	mnsr = (md_mnset_record *)sr;
4945 	/* Setup genid on set and node records */
4946 	if (args->drivedescs) {
4947 		if (mnsr->sr_genid != args->drivedescs->dd_genid) {
4948 			change = 1;
4949 			mnsr->sr_genid = args->drivedescs->dd_genid;
4950 		}
4951 		nrp = mnsr->sr_nodechain;
4952 		while (nrp) {
4953 			if (nrp->nr_genid != args->drivedescs->dd_genid) {
4954 				change = 1;
4955 				nrp->nr_genid = args->drivedescs->dd_genid;
4956 			}
4957 			nrp = nrp->nr_next;
4958 		}
4959 	}
4960 	for (dr = mnsr->sr_drivechain; dr; dr = dr->dr_next) {
4961 		dn1 = metadrivename_withdrkey(local_sp, sideno,
4962 		    dr->dr_key, (MD_BASICNAME_OK | PRINT_FAST), ep);
4963 		if (dn1 == NULL)
4964 			goto out;
4965 		for (dd = args->drivedescs; dd != NULL; dd = dd->dd_next) {
4966 			dn = dd->dd_dnp;
4967 			/* Found this node's drive rec to match dd */
4968 			if (strcmp(dn->cname, dn1->cname) == 0)
4969 				break;
4970 		}
4971 
4972 		/*
4973 		 * If drive found in master's list, make slave match master.
4974 		 * If drive not found in master's list, remove drive.
4975 		 */
4976 		if (dd) {
4977 			if ((dr->dr_flags != dd->dd_flags) ||
4978 			    (dr->dr_genid != dd->dd_genid)) {
4979 				change = 1;
4980 				dr->dr_flags = dd->dd_flags;
4981 				dr->dr_genid = dd->dd_genid;
4982 			}
4983 		} else {
4984 			/*
4985 			 * Delete entry from linked list.  Need to use
4986 			 * dr_placeholder so that dr->dr_next points to
4987 			 * the next drive record in the list.
4988 			 */
4989 			if (dr_placeholder == NULL) {
4990 				dr_placeholder =
4991 					Zalloc(sizeof (md_drive_record));
4992 			}
4993 			dr_placeholder->dr_next = dr->dr_next;
4994 			dr_placeholder->dr_key = dr->dr_key;
4995 			sr_del_drv(sr, dr->dr_selfid);
4996 			(void) del_sideno_sidenm(dr_placeholder->dr_key,
4997 				sideno, ep);
4998 			change = 1;
4999 			dr = dr_placeholder;
5000 		}
5001 	}
5002 out:
5003 	/* If incore records are correct, don't need to write to disk */
5004 	if (change) {
5005 		/* Don't increment the genid in commitset */
5006 		commitset(sr, FALSE, ep);
5007 	}
5008 	free_sr(sr);
5009 
5010 	err = svc_fini(ep);
5011 
5012 	if (dr_placeholder != NULL)
5013 		Free(dr_placeholder);
5014 
5015 	return (TRUE);
5016 }
5017 
5018 /*
5019  * Version 2 routine to update this node's drive records based on
5020  * list passed in from master node.
5021  */
5022 bool_t
5023 mdrpc_upd_dr_reconfig_2_svc(
5024 	mdrpc_upd_dr_flags_2_args	*args,
5025 	mdrpc_generic_res		*res,
5026 	struct svc_req			*rqstp		/* RPC stuff */
5027 )
5028 {
5029 	switch (args->rev) {
5030 	    case MD_METAD_ARGS_REV_1:
5031 		return (mdrpc_upd_dr_reconfig_common(
5032 		    &args->mdrpc_upd_dr_flags_2_args_u.rev1, res, rqstp));
5033 	    default:
5034 		return (FALSE);
5035 	}
5036 }
5037 
5038 /*
5039  * reset mirror owner for mirrors owned by deleted
5040  * or withdrawn host(s).  Hosts being deleted or
5041  * withdrawn are designated by nodeid since host is
5042  * already deleted or withdrawn from set and may not
5043  * be able to translate between a nodename and a nodeid.
5044  * If an error occurs, ep will be set to that error information.
5045  */
5046 static void
5047 reset_mirror_owner(
5048 	char		*setname,
5049 	int		node_c,
5050 	int		*node_id,	/* Array of node ids */
5051 	md_error_t	*ep
5052 )
5053 {
5054 	mdsetname_t		*local_sp;
5055 	int			i;
5056 	mdnamelist_t		*devnlp = NULL;
5057 	mdnamelist_t		*p;
5058 	mdname_t		*devnp = NULL;
5059 	md_set_mmown_params_t	ownpar_p;
5060 	md_set_mmown_params_t	*ownpar = &ownpar_p;
5061 	char			*miscname;
5062 
5063 	if ((local_sp = metasetname(setname, ep)) == NULL)
5064 		return;
5065 
5066 	/* get a list of all the mirrors for current set */
5067 	if (meta_get_mirror_names(local_sp, &devnlp, 0, ep) < 0)
5068 		return;
5069 
5070 	/* for each mirror */
5071 	for (p = devnlp; (p != NULL); p = p->next) {
5072 		devnp = p->namep;
5073 
5074 		/*
5075 		 * we can only do these for mirrors so make sure we
5076 		 * really have a mirror device and not a softpartition
5077 		 * imitating one. meta_get_mirror_names seems to think
5078 		 * softparts on top of a mirror are mirrors!
5079 		 */
5080 		if ((miscname = metagetmiscname(devnp, ep)) == NULL)
5081 			goto out;
5082 		if (strcmp(miscname, MD_MIRROR) != 0)
5083 			continue;
5084 
5085 		(void) memset(ownpar, 0, sizeof (*ownpar));
5086 		ownpar->d.mnum = meta_getminor(devnp->dev);
5087 		MD_SETDRIVERNAME(ownpar, MD_MIRROR, local_sp->setno);
5088 
5089 		/* get the current owner id */
5090 		if (metaioctl(MD_MN_GET_MM_OWNER, ownpar, ep,
5091 		    "MD_MN_GET_MM_OWNER") != 0) {
5092 			mde_perror(ep, gettext(
5093 			    "Unable to get mirror owner for %s/d%u"),
5094 			    local_sp->setname,
5095 			    (unsigned)MD_MIN2UNIT(ownpar->d.mnum));
5096 			goto out;
5097 		}
5098 
5099 		if (ownpar->d.owner == MD_MN_MIRROR_UNOWNED) {
5100 			mdclrerror(ep);
5101 			continue;
5102 		}
5103 		/*
5104 		 * reset owner only if the current owner is
5105 		 * in the list of nodes being deleted.
5106 		 */
5107 		for (i = 0; i < node_c; i++) {
5108 			if (ownpar->d.owner == node_id[i]) {
5109 				if (meta_mn_change_owner(&ownpar,
5110 				    local_sp->setno, ownpar->d.mnum,
5111 				    MD_MN_MIRROR_UNOWNED,
5112 				    MD_MN_MM_ALLOW_CHANGE) == -1) {
5113 					mde_perror(ep, gettext(
5114 					    "Unable to reset mirror owner for"
5115 					    " %s/d%u"), local_sp->setname,
5116 					    (unsigned)MD_MIN2UNIT(
5117 					    ownpar->d.mnum));
5118 					goto out;
5119 				}
5120 				break;
5121 			}
5122 		}
5123 	}
5124 
5125 out:
5126 	/* cleanup */
5127 	metafreenamelist(devnlp);
5128 }
5129 
5130 /*
5131  * Wrapper routine for reset_mirror_owner.
5132  * Called when hosts are deleted or withdrawn
5133  * in order to reset any mirror owners that are needed.
5134  */
5135 bool_t
5136 mdrpc_reset_mirror_owner_common(
5137 	mdrpc_nodeid_args	*args,
5138 	mdrpc_generic_res	*res,
5139 	struct svc_req		*rqstp		/* RPC stuff */
5140 )
5141 {
5142 	md_error_t		*ep = &res->status;
5143 	int			err;
5144 	int			op_mode = W_OK;
5145 
5146 	/* setup, check permissions */
5147 	(void) memset(res, 0, sizeof (*res));
5148 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
5149 		return (FALSE);
5150 	else if (err != 0)
5151 		return (TRUE);
5152 
5153 	if (check_set_lock(op_mode, args->cl_sk, ep))
5154 		return (TRUE);
5155 
5156 	/* doit */
5157 	reset_mirror_owner(args->sp->setname, args->nodeid.nodeid_len,
5158 	    args->nodeid.nodeid_val, ep);
5159 
5160 	err = svc_fini(ep);
5161 
5162 	return (TRUE);
5163 }
5164 
5165 /*
5166  * RPC service routine to reset the mirror owner for mirrors owned
5167  * by the given hosts.  Typically, the list of given hosts is a list
5168  * of nodes being deleted or withdrawn from a diskset.
5169  * The given hosts are designated by nodeid since host may
5170  * already be deleted or withdrawn from set and may not
5171  * be able to translate between a nodename and a nodeid.
5172  */
5173 bool_t
5174 mdrpc_reset_mirror_owner_2_svc(
5175 	mdrpc_nodeid_2_args	*args,
5176 	mdrpc_generic_res	*res,
5177 	struct svc_req		*rqstp		/* RPC stuff */
5178 )
5179 {
5180 	switch (args->rev) {
5181 	    case MD_METAD_ARGS_REV_1:
5182 		return (mdrpc_reset_mirror_owner_common(
5183 		    &args->mdrpc_nodeid_2_args_u.rev1, res,
5184 		    rqstp));
5185 	    default:
5186 		return (FALSE);
5187 	}
5188 }
5189 
5190 /*
5191  * Call routines to suspend and resume I/O for the given diskset(s).
5192  * Called during reconfig cycle.
5193  * Diskset of 0 represents all MN disksets.
5194  */
5195 bool_t
5196 mdrpc_mn_susp_res_io_2_svc(
5197 	mdrpc_mn_susp_res_io_2_args	*args,
5198 	mdrpc_generic_res		*res,
5199 	struct svc_req			*rqstp		/* RPC stuff */
5200 )
5201 {
5202 	mdrpc_mn_susp_res_io_args	*args_sr;
5203 	md_error_t			*ep = &res->status;
5204 	int				err;
5205 	int				op_mode = R_OK;
5206 
5207 	switch (args->rev) {
5208 	    case MD_METAD_ARGS_REV_1:
5209 		/* setup, check permissions */
5210 		(void) memset(res, 0, sizeof (*res));
5211 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
5212 			return (FALSE);
5213 		else if (err != 0)
5214 			return (TRUE);
5215 
5216 		args_sr = &(args->mdrpc_mn_susp_res_io_2_args_u.rev1);
5217 		switch (args_sr->susp_res_cmd) {
5218 		case MN_SUSP_IO:
5219 			(void) (metaioctl(MD_MN_SUSPEND_SET,
5220 				&args_sr->susp_res_setno, ep, NULL));
5221 			break;
5222 		case MN_RES_IO:
5223 			(void) (metaioctl(MD_MN_RESUME_SET,
5224 				&args_sr->susp_res_setno, ep, NULL));
5225 			break;
5226 		}
5227 		err = svc_fini(ep);
5228 		return (TRUE);
5229 
5230 	    default:
5231 		return (FALSE);
5232 	}
5233 }
5234 
5235 /*
5236  * Resnarf a set after it has been imported
5237  */
5238 bool_t
5239 mdrpc_resnarf_set_2_svc(
5240 	mdrpc_setno_2_args	*args,
5241 	mdrpc_generic_res	*res,
5242 	struct svc_req		*rqstp		/* RPC stuff */
5243 )
5244 {
5245 	mdrpc_setno_args	*setno_args;
5246 	md_error_t		*ep = &res->status;
5247 	int			err;
5248 	int			op_mode = R_OK;
5249 
5250 	switch (args->rev) {
5251 	    case MD_METAD_ARGS_REV_1:
5252 		setno_args = &args->mdrpc_setno_2_args_u.rev1;
5253 		break;
5254 	    default:
5255 		return (FALSE);
5256 	}
5257 
5258 	(void) memset(res, 0, sizeof (*res));
5259 	if ((err = svc_init(rqstp, op_mode, ep)) < 0)
5260 		return (FALSE);
5261 	else if (err != 0)
5262 		return (TRUE);
5263 
5264 	/* do it */
5265 	if (resnarf_set(setno_args->setno, ep) < 0)
5266 		return (FALSE);
5267 
5268 	err = svc_fini(ep);
5269 	return (TRUE);
5270 }
5271 
5272 /*
5273  * Creates a resync thread.
5274  * Always returns true.
5275  */
5276 bool_t
5277 mdrpc_mn_mirror_resync_all_2_svc(
5278 	mdrpc_setno_2_args	*args,
5279 	mdrpc_generic_res	*res,
5280 	struct svc_req		*rqstp		/* RPC stuff */
5281 )
5282 {
5283 	md_error_t		*ep = &res->status;
5284 	mdrpc_setno_args	*setno_args;
5285 	int			err;
5286 	int			op_mode = R_OK;
5287 
5288 	switch (args->rev) {
5289 	    case MD_METAD_ARGS_REV_1:
5290 		/* setup, check permissions */
5291 		(void) memset(res, 0, sizeof (*res));
5292 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
5293 			return (FALSE);
5294 		else if (err != 0)
5295 			return (TRUE);
5296 		setno_args = &args->mdrpc_setno_2_args_u.rev1;
5297 
5298 		/*
5299 		 * Need to invoke a metasync on a node newly added to a set.
5300 		 */
5301 		meta_mn_mirror_resync_all(&(setno_args->setno));
5302 
5303 		err = svc_fini(ep);
5304 		return (TRUE);
5305 
5306 	    default:
5307 		return (FALSE);
5308 	}
5309 }
5310 
5311 /*
5312  * Updates ABR state for all softpartitions. Calls meta_mn_sp_update_abr(),
5313  * which forks a daemon process to perform this action.
5314  * Always returns true.
5315  */
5316 bool_t
5317 mdrpc_mn_sp_update_abr_2_svc(
5318 	mdrpc_setno_2_args	*args,
5319 	mdrpc_generic_res	*res,
5320 	struct svc_req		*rqstp		/* RPC stuff */
5321 )
5322 {
5323 	md_error_t		*ep = &res->status;
5324 	mdrpc_setno_args	*setno_args;
5325 	int			err;
5326 	int			op_mode = R_OK;
5327 
5328 	switch (args->rev) {
5329 	    case MD_METAD_ARGS_REV_1:
5330 		/* setup, check permissions */
5331 		(void) memset(res, 0, sizeof (*res));
5332 		if ((err = svc_init(rqstp, op_mode, ep)) < 0)
5333 			return (FALSE);
5334 		else if (err != 0)
5335 			return (TRUE);
5336 		setno_args = &args->mdrpc_setno_2_args_u.rev1;
5337 
5338 		meta_mn_sp_update_abr(&(setno_args->setno));
5339 
5340 		err = svc_fini(ep);
5341 		return (TRUE);
5342 
5343 	    default:
5344 		return (FALSE);
5345 	}
5346 }
5347