xref: /titanic_50/usr/src/lib/libzonecfg/common/libzonecfg.c (revision 193974072f41a843678abf5f61979c748687e66b)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 
22 /*
23  * Copyright 2008 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 <libsysevent.h>
30 #include <pthread.h>
31 #include <stdlib.h>
32 #include <errno.h>
33 #include <fnmatch.h>
34 #include <strings.h>
35 #include <unistd.h>
36 #include <assert.h>
37 #include <libgen.h>
38 #include <libintl.h>
39 #include <alloca.h>
40 #include <ctype.h>
41 #include <sys/acl.h>
42 #include <sys/stat.h>
43 #include <sys/brand.h>
44 #include <sys/mntio.h>
45 #include <sys/mnttab.h>
46 #include <sys/nvpair.h>
47 #include <sys/types.h>
48 #include <sys/sockio.h>
49 #include <ftw.h>
50 #include <pool.h>
51 #include <libscf.h>
52 #include <libproc.h>
53 #include <sys/priocntl.h>
54 #include <libuutil.h>
55 #include <wait.h>
56 #include <bsm/adt.h>
57 
58 #include <arpa/inet.h>
59 #include <netdb.h>
60 
61 #include <libxml/xmlmemory.h>
62 #include <libxml/parser.h>
63 
64 #include <libdevinfo.h>
65 #include <uuid/uuid.h>
66 #include <dirent.h>
67 #include <libbrand.h>
68 
69 #include <libzonecfg.h>
70 #include "zonecfg_impl.h"
71 
72 #define	_PATH_TMPFILE	"/zonecfg.XXXXXX"
73 #define	ZONE_CB_RETRY_COUNT		10
74 #define	ZONE_EVENT_PING_SUBCLASS	"ping"
75 #define	ZONE_EVENT_PING_PUBLISHER	"solaris"
76 
77 /* Hard-code the DTD element/attribute/entity names just once, here. */
78 #define	DTD_ELEM_ATTR		(const xmlChar *) "attr"
79 #define	DTD_ELEM_COMMENT	(const xmlChar *) "comment"
80 #define	DTD_ELEM_DEVICE		(const xmlChar *) "device"
81 #define	DTD_ELEM_FS		(const xmlChar *) "filesystem"
82 #define	DTD_ELEM_FSOPTION	(const xmlChar *) "fsoption"
83 #define	DTD_ELEM_IPD		(const xmlChar *) "inherited-pkg-dir"
84 #define	DTD_ELEM_NET		(const xmlChar *) "network"
85 #define	DTD_ELEM_RCTL		(const xmlChar *) "rctl"
86 #define	DTD_ELEM_RCTLVALUE	(const xmlChar *) "rctl-value"
87 #define	DTD_ELEM_ZONE		(const xmlChar *) "zone"
88 #define	DTD_ELEM_DATASET	(const xmlChar *) "dataset"
89 #define	DTD_ELEM_TMPPOOL	(const xmlChar *) "tmp_pool"
90 #define	DTD_ELEM_PSET		(const xmlChar *) "pset"
91 #define	DTD_ELEM_MCAP		(const xmlChar *) "mcap"
92 #define	DTD_ELEM_PACKAGE	(const xmlChar *) "package"
93 #define	DTD_ELEM_PATCH		(const xmlChar *) "patch"
94 #define	DTD_ELEM_OBSOLETES	(const xmlChar *) "obsoletes"
95 #define	DTD_ELEM_DEV_PERM	(const xmlChar *) "dev-perm"
96 
97 #define	DTD_ATTR_ACTION		(const xmlChar *) "action"
98 #define	DTD_ATTR_ADDRESS	(const xmlChar *) "address"
99 #define	DTD_ATTR_AUTOBOOT	(const xmlChar *) "autoboot"
100 #define	DTD_ATTR_IPTYPE		(const xmlChar *) "ip-type"
101 #define	DTD_ATTR_DEFROUTER	(const xmlChar *) "defrouter"
102 #define	DTD_ATTR_DIR		(const xmlChar *) "directory"
103 #define	DTD_ATTR_LIMIT		(const xmlChar *) "limit"
104 #define	DTD_ATTR_LIMITPRIV	(const xmlChar *) "limitpriv"
105 #define	DTD_ATTR_BOOTARGS	(const xmlChar *) "bootargs"
106 #define	DTD_ATTR_SCHED		(const xmlChar *) "scheduling-class"
107 #define	DTD_ATTR_MATCH		(const xmlChar *) "match"
108 #define	DTD_ATTR_NAME		(const xmlChar *) "name"
109 #define	DTD_ATTR_PHYSICAL	(const xmlChar *) "physical"
110 #define	DTD_ATTR_POOL		(const xmlChar *) "pool"
111 #define	DTD_ATTR_PRIV		(const xmlChar *) "priv"
112 #define	DTD_ATTR_RAW		(const xmlChar *) "raw"
113 #define	DTD_ATTR_SPECIAL	(const xmlChar *) "special"
114 #define	DTD_ATTR_TYPE		(const xmlChar *) "type"
115 #define	DTD_ATTR_VALUE		(const xmlChar *) "value"
116 #define	DTD_ATTR_ZONEPATH	(const xmlChar *) "zonepath"
117 #define	DTD_ATTR_NCPU_MIN	(const xmlChar *) "ncpu_min"
118 #define	DTD_ATTR_NCPU_MAX	(const xmlChar *) "ncpu_max"
119 #define	DTD_ATTR_IMPORTANCE	(const xmlChar *) "importance"
120 #define	DTD_ATTR_PHYSCAP	(const xmlChar *) "physcap"
121 #define	DTD_ATTR_VERSION	(const xmlChar *) "version"
122 #define	DTD_ATTR_ID		(const xmlChar *) "id"
123 #define	DTD_ATTR_UID		(const xmlChar *) "uid"
124 #define	DTD_ATTR_GID		(const xmlChar *) "gid"
125 #define	DTD_ATTR_MODE		(const xmlChar *) "mode"
126 #define	DTD_ATTR_ACL		(const xmlChar *) "acl"
127 #define	DTD_ATTR_BRAND		(const xmlChar *) "brand"
128 
129 #define	DTD_ENTITY_BOOLEAN	"boolean"
130 #define	DTD_ENTITY_DEVPATH	"devpath"
131 #define	DTD_ENTITY_DRIVER	"driver"
132 #define	DTD_ENTITY_DRVMIN	"drv_min"
133 #define	DTD_ENTITY_FALSE	"false"
134 #define	DTD_ENTITY_INT		"int"
135 #define	DTD_ENTITY_STRING	"string"
136 #define	DTD_ENTITY_TRUE		"true"
137 #define	DTD_ENTITY_UINT		"uint"
138 
139 #define	DTD_ENTITY_BOOL_LEN	6	/* "false" */
140 
141 #define	ATTACH_FORCED	"SUNWattached.xml"
142 
143 #define	TMP_POOL_NAME	"SUNWtmp_%s"
144 #define	MAX_TMP_POOL_NAME	(ZONENAME_MAX + 9)
145 #define	RCAP_SERVICE	"system/rcap:default"
146 #define	POOLD_SERVICE	"system/pools/dynamic:default"
147 
148 /*
149  * rctl alias definitions
150  *
151  * This holds the alias, the full rctl name, the default priv value, action
152  * and lower limit.  The functions that handle rctl aliases step through
153  * this table, matching on the alias, and using the full values for setting
154  * the rctl entry as well the limit for validation.
155  */
156 static struct alias {
157 	char *shortname;
158 	char *realname;
159 	char *priv;
160 	char *action;
161 	uint64_t low_limit;
162 } aliases[] = {
163 	{ALIAS_MAXLWPS, "zone.max-lwps", "privileged", "deny", 100},
164 	{ALIAS_MAXSHMMEM, "zone.max-shm-memory", "privileged", "deny", 0},
165 	{ALIAS_MAXSHMIDS, "zone.max-shm-ids", "privileged", "deny", 0},
166 	{ALIAS_MAXMSGIDS, "zone.max-msg-ids", "privileged", "deny", 0},
167 	{ALIAS_MAXSEMIDS, "zone.max-sem-ids", "privileged", "deny", 0},
168 	{ALIAS_MAXLOCKEDMEM, "zone.max-locked-memory", "privileged", "deny", 0},
169 	{ALIAS_MAXSWAP, "zone.max-swap", "privileged", "deny", 0},
170 	{ALIAS_SHARES, "zone.cpu-shares", "privileged", "none", 0},
171 	{ALIAS_CPUCAP, "zone.cpu-cap", "privileged", "deny", 0},
172 	{NULL, NULL, NULL, NULL, 0}
173 };
174 
175 /*
176  * Structure for applying rctls to a running zone.  It allows important
177  * process values to be passed together easily.
178  */
179 typedef struct pr_info_handle {
180 	struct ps_prochandle *pr;
181 	pid_t pid;
182 } pr_info_handle_t;
183 
184 struct zone_dochandle {
185 	char		*zone_dh_rootdir;
186 	xmlDocPtr	zone_dh_doc;
187 	xmlNodePtr	zone_dh_cur;
188 	xmlNodePtr	zone_dh_top;
189 	boolean_t	zone_dh_newzone;
190 	boolean_t	zone_dh_snapshot;
191 	boolean_t	zone_dh_sw_inv;
192 	char		zone_dh_delete_name[ZONENAME_MAX];
193 };
194 
195 struct znotify {
196 	void * zn_private;
197 	evchan_t *zn_eventchan;
198 	int (*zn_callback)(const  char *zonename, zoneid_t zid,
199 	    const char *newstate, const char *oldstate, hrtime_t when, void *p);
200 	pthread_mutex_t zn_mutex;
201 	pthread_cond_t zn_cond;
202 	pthread_mutex_t zn_bigmutex;
203 	volatile enum {ZN_UNLOCKED, ZN_LOCKED, ZN_PING_INFLIGHT,
204 	    ZN_PING_RECEIVED} zn_state;
205 	char zn_subscriber_id[MAX_SUBID_LEN];
206 	volatile boolean_t zn_failed;
207 	int zn_failure_count;
208 };
209 
210 /* used to track nested zone-lock operations */
211 static int zone_lock_cnt = 0;
212 
213 /* used to communicate lock status to children */
214 #define	LOCK_ENV_VAR	"_ZONEADM_LOCK_HELD"
215 static char zoneadm_lock_held[] = LOCK_ENV_VAR"=1";
216 static char zoneadm_lock_not_held[] = LOCK_ENV_VAR"=0";
217 
218 char *zonecfg_root = "";
219 
220 /*
221  * For functions which return int, which is most of the functions herein,
222  * the return values should be from the Z_foo set defined in <libzonecfg.h>.
223  * In some instances, we take pains mapping some libc errno values to Z_foo
224  * values from this set.
225  */
226 
227 /*
228  * Set the root (/) path for all zonecfg configuration files.  This is a
229  * private interface used by Live Upgrade extensions to access zone
230  * configuration inside mounted alternate boot environments.
231  */
232 void
233 zonecfg_set_root(const char *rootpath)
234 {
235 	if (*zonecfg_root != '\0')
236 		free(zonecfg_root);
237 	if (rootpath == NULL || rootpath[0] == '\0' || rootpath[1] == '\0' ||
238 	    (zonecfg_root = strdup(rootpath)) == NULL)
239 		zonecfg_root = "";
240 }
241 
242 const char *
243 zonecfg_get_root(void)
244 {
245 	return (zonecfg_root);
246 }
247 
248 boolean_t
249 zonecfg_in_alt_root(void)
250 {
251 	return (*zonecfg_root != '\0');
252 }
253 
254 /*
255  * Callers of the _file_path() functions are expected to have the second
256  * parameter be a (char foo[MAXPATHLEN]).
257  */
258 
259 static boolean_t
260 config_file_path(const char *zonename, char *answer)
261 {
262 	return (snprintf(answer, MAXPATHLEN, "%s%s/%s.xml", zonecfg_root,
263 	    ZONE_CONFIG_ROOT, zonename) < MAXPATHLEN);
264 }
265 
266 static boolean_t
267 snap_file_path(const char *zonename, char *answer)
268 {
269 	return (snprintf(answer, MAXPATHLEN, "%s%s/%s.snapshot.xml",
270 	    zonecfg_root, ZONE_SNAPSHOT_ROOT, zonename) < MAXPATHLEN);
271 }
272 
273 /*ARGSUSED*/
274 static void
275 zonecfg_error_func(void *ctx, const char *msg, ...)
276 {
277 	/*
278 	 * This function does nothing by design.  Its purpose is to prevent
279 	 * libxml from dumping unwanted messages to stdout/stderr.
280 	 */
281 }
282 
283 zone_dochandle_t
284 zonecfg_init_handle(void)
285 {
286 	zone_dochandle_t handle = calloc(1, sizeof (struct zone_dochandle));
287 	if (handle == NULL) {
288 		errno = Z_NOMEM;
289 		return (NULL);
290 	}
291 
292 	/* generic libxml initialization */
293 	xmlLineNumbersDefault(1);
294 	xmlLoadExtDtdDefaultValue |= XML_DETECT_IDS;
295 	xmlDoValidityCheckingDefaultValue = 1;
296 	(void) xmlKeepBlanksDefault(0);
297 	xmlGetWarningsDefaultValue = 0;
298 	xmlSetGenericErrorFunc(NULL, zonecfg_error_func);
299 
300 	return (handle);
301 }
302 
303 int
304 zonecfg_check_handle(zone_dochandle_t handle)
305 {
306 	if (handle == NULL || handle->zone_dh_doc == NULL)
307 		return (Z_BAD_HANDLE);
308 	return (Z_OK);
309 }
310 
311 void
312 zonecfg_fini_handle(zone_dochandle_t handle)
313 {
314 	if (zonecfg_check_handle(handle) == Z_OK)
315 		xmlFreeDoc(handle->zone_dh_doc);
316 	if (handle != NULL)
317 		free(handle);
318 }
319 
320 static int
321 zonecfg_destroy_impl(char *filename)
322 {
323 	if (unlink(filename) == -1) {
324 		if (errno == EACCES)
325 			return (Z_ACCES);
326 		if (errno == ENOENT)
327 			return (Z_NO_ZONE);
328 		return (Z_MISC_FS);
329 	}
330 	return (Z_OK);
331 }
332 
333 int
334 zonecfg_destroy(const char *zonename, boolean_t force)
335 {
336 	char path[MAXPATHLEN];
337 	struct zoneent ze;
338 	int err, state_err;
339 	zone_state_t state;
340 
341 	if (!config_file_path(zonename, path))
342 		return (Z_MISC_FS);
343 
344 	state_err = zone_get_state((char *)zonename, &state);
345 	err = access(path, W_OK);
346 
347 	/*
348 	 * If there is no file, and no index entry, reliably indicate that no
349 	 * such zone exists.
350 	 */
351 	if ((state_err == Z_NO_ZONE) && (err == -1) && (errno == ENOENT))
352 		return (Z_NO_ZONE);
353 
354 	/*
355 	 * Handle any other filesystem related errors (except if the XML
356 	 * file is missing, which we treat silently), unless we're forcing,
357 	 * in which case we plow on.
358 	 */
359 	if (err == -1 && errno != ENOENT) {
360 		if (errno == EACCES)
361 			return (Z_ACCES);
362 		else if (!force)
363 			return (Z_MISC_FS);
364 	}
365 
366 	if (state > ZONE_STATE_INSTALLED)
367 		return (Z_BAD_ZONE_STATE);
368 
369 	if (!force && state > ZONE_STATE_CONFIGURED)
370 		return (Z_BAD_ZONE_STATE);
371 
372 	/*
373 	 * Index deletion succeeds even if the entry doesn't exist.  So this
374 	 * will fail only if we've had some more severe problem.
375 	 */
376 	bzero(&ze, sizeof (ze));
377 	(void) strlcpy(ze.zone_name, zonename, sizeof (ze.zone_name));
378 	if ((err = putzoneent(&ze, PZE_REMOVE)) != Z_OK)
379 		if (!force)
380 			return (err);
381 
382 	err = zonecfg_destroy_impl(path);
383 
384 	/*
385 	 * Treat failure to find the XML file silently, since, well, it's
386 	 * gone, and with the index file cleaned up, we're done.
387 	 */
388 	if (err == Z_OK || err == Z_NO_ZONE)
389 		return (Z_OK);
390 	return (err);
391 }
392 
393 int
394 zonecfg_destroy_snapshot(const char *zonename)
395 {
396 	char path[MAXPATHLEN];
397 
398 	if (!snap_file_path(zonename, path))
399 		return (Z_MISC_FS);
400 	return (zonecfg_destroy_impl(path));
401 }
402 
403 static int
404 getroot(zone_dochandle_t handle, xmlNodePtr *root)
405 {
406 	if (zonecfg_check_handle(handle) == Z_BAD_HANDLE)
407 		return (Z_BAD_HANDLE);
408 
409 	*root = xmlDocGetRootElement(handle->zone_dh_doc);
410 
411 	if (*root == NULL)
412 		return (Z_EMPTY_DOCUMENT);
413 
414 	if (xmlStrcmp((*root)->name, DTD_ELEM_ZONE))
415 		return (Z_WRONG_DOC_TYPE);
416 
417 	return (Z_OK);
418 }
419 
420 static int
421 operation_prep(zone_dochandle_t handle)
422 {
423 	xmlNodePtr root;
424 	int err;
425 
426 	if ((err = getroot(handle, &root)) != 0)
427 		return (err);
428 
429 	handle->zone_dh_cur = root;
430 	handle->zone_dh_top = root;
431 	return (Z_OK);
432 }
433 
434 static int
435 fetchprop(xmlNodePtr cur, const xmlChar *propname, char *dst, size_t dstsize)
436 {
437 	xmlChar *property;
438 	size_t srcsize;
439 
440 	if ((property = xmlGetProp(cur, propname)) == NULL)
441 		return (Z_BAD_PROPERTY);
442 	srcsize = strlcpy(dst, (char *)property, dstsize);
443 	xmlFree(property);
444 	if (srcsize >= dstsize)
445 		return (Z_TOO_BIG);
446 	return (Z_OK);
447 }
448 
449 static int
450 fetch_alloc_prop(xmlNodePtr cur, const xmlChar *propname, char **dst)
451 {
452 	xmlChar *property;
453 
454 	if ((property = xmlGetProp(cur, propname)) == NULL)
455 		return (Z_BAD_PROPERTY);
456 	if ((*dst = strdup((char *)property)) == NULL) {
457 		xmlFree(property);
458 		return (Z_NOMEM);
459 	}
460 	xmlFree(property);
461 	return (Z_OK);
462 }
463 
464 static int
465 getrootattr(zone_dochandle_t handle, const xmlChar *propname,
466     char *propval, size_t propsize)
467 {
468 	xmlNodePtr root;
469 	int err;
470 
471 	if ((err = getroot(handle, &root)) != 0)
472 		return (err);
473 
474 	return (fetchprop(root, propname, propval, propsize));
475 }
476 
477 static int
478 get_alloc_rootattr(zone_dochandle_t handle, const xmlChar *propname,
479     char **propval)
480 {
481 	xmlNodePtr root;
482 	int err;
483 
484 	if ((err = getroot(handle, &root)) != 0)
485 		return (err);
486 
487 	return (fetch_alloc_prop(root, propname, propval));
488 }
489 
490 static int
491 setrootattr(zone_dochandle_t handle, const xmlChar *propname,
492     const char *propval)
493 {
494 	int err;
495 	xmlNodePtr root;
496 
497 	if ((err = getroot(handle, &root)) != Z_OK)
498 		return (err);
499 
500 	/*
501 	 * If we get a null propval remove the property (ignore return since it
502 	 * may not be set to begin with).
503 	 */
504 	if (propval == NULL) {
505 		(void) xmlUnsetProp(root, propname);
506 	} else {
507 		if (xmlSetProp(root, propname, (const xmlChar *) propval)
508 		    == NULL)
509 			return (Z_INVAL);
510 	}
511 	return (Z_OK);
512 }
513 
514 static void
515 addcomment(zone_dochandle_t handle, const char *comment)
516 {
517 	xmlNodePtr node;
518 	node = xmlNewComment((xmlChar *) comment);
519 
520 	if (node != NULL)
521 		(void) xmlAddPrevSibling(handle->zone_dh_top, node);
522 }
523 
524 static void
525 stripcomments(zone_dochandle_t handle)
526 {
527 	xmlDocPtr top;
528 	xmlNodePtr child, next;
529 
530 	top = handle->zone_dh_doc;
531 	for (child = top->xmlChildrenNode; child != NULL; child = next) {
532 		next = child->next;
533 		if (child->name == NULL)
534 			continue;
535 		if (xmlStrcmp(child->name, DTD_ELEM_COMMENT) == 0) {
536 			next = child->next;
537 			xmlUnlinkNode(child);
538 			xmlFreeNode(child);
539 		}
540 	}
541 }
542 
543 static void
544 strip_sw_inv(zone_dochandle_t handle)
545 {
546 	xmlNodePtr root, child, next;
547 
548 	root = xmlDocGetRootElement(handle->zone_dh_doc);
549 	for (child = root->xmlChildrenNode; child != NULL; child = next) {
550 		next = child->next;
551 		if (child->name == NULL)
552 			continue;
553 		if (xmlStrcmp(child->name, DTD_ELEM_PACKAGE) == 0 ||
554 		    xmlStrcmp(child->name, DTD_ELEM_PATCH) == 0) {
555 			next = child->next;
556 			xmlUnlinkNode(child);
557 			xmlFreeNode(child);
558 		}
559 	}
560 }
561 
562 static int
563 zonecfg_get_handle_impl(const char *zonename, const char *filename,
564     zone_dochandle_t handle)
565 {
566 	xmlValidCtxtPtr cvp;
567 	struct stat statbuf;
568 	int valid;
569 
570 	if (zonename == NULL)
571 		return (Z_NO_ZONE);
572 
573 	if ((handle->zone_dh_doc = xmlParseFile(filename)) == NULL) {
574 		/* distinguish file not found vs. found but not parsed */
575 		if (stat(filename, &statbuf) == 0)
576 			return (Z_INVALID_DOCUMENT);
577 		return (Z_NO_ZONE);
578 	}
579 	if ((cvp = xmlNewValidCtxt()) == NULL)
580 		return (Z_NOMEM);
581 	cvp->error = zonecfg_error_func;
582 	cvp->warning = zonecfg_error_func;
583 	valid = xmlValidateDocument(cvp, handle->zone_dh_doc);
584 	xmlFreeValidCtxt(cvp);
585 	if (valid == 0)
586 		return (Z_INVALID_DOCUMENT);
587 
588 	/* delete any comments such as inherited Sun copyright / ident str */
589 	stripcomments(handle);
590 	return (Z_OK);
591 }
592 
593 int
594 zonecfg_get_handle(const char *zonename, zone_dochandle_t handle)
595 {
596 	char path[MAXPATHLEN];
597 
598 	if (!config_file_path(zonename, path))
599 		return (Z_MISC_FS);
600 	handle->zone_dh_newzone = B_FALSE;
601 
602 	return (zonecfg_get_handle_impl(zonename, path, handle));
603 }
604 
605 int
606 zonecfg_get_attach_handle(const char *path, const char *fname,
607     const char *zonename, boolean_t preserve_sw, zone_dochandle_t handle)
608 {
609 	char		migpath[MAXPATHLEN];
610 	int		err;
611 	struct stat	buf;
612 
613 	if (snprintf(migpath, sizeof (migpath), "%s/root", path) >=
614 	    sizeof (migpath))
615 		return (Z_NOMEM);
616 
617 	if (stat(migpath, &buf) == -1 || !S_ISDIR(buf.st_mode))
618 		return (Z_NO_ZONE);
619 
620 	if (snprintf(migpath, sizeof (migpath), "%s/%s", path, fname) >=
621 	    sizeof (migpath))
622 		return (Z_NOMEM);
623 
624 	if ((err = zonecfg_get_handle_impl(zonename, migpath, handle)) != Z_OK)
625 		return (err);
626 
627 	if (!preserve_sw)
628 		strip_sw_inv(handle);
629 
630 	handle->zone_dh_newzone = B_TRUE;
631 	if ((err = setrootattr(handle, DTD_ATTR_ZONEPATH, path)) != Z_OK)
632 		return (err);
633 
634 	return (setrootattr(handle, DTD_ATTR_NAME, zonename));
635 }
636 
637 int
638 zonecfg_get_snapshot_handle(const char *zonename, zone_dochandle_t handle)
639 {
640 	char path[MAXPATHLEN];
641 
642 	if (!snap_file_path(zonename, path))
643 		return (Z_MISC_FS);
644 	handle->zone_dh_newzone = B_FALSE;
645 	return (zonecfg_get_handle_impl(zonename, path, handle));
646 }
647 
648 int
649 zonecfg_get_template_handle(const char *template, const char *zonename,
650     zone_dochandle_t handle)
651 {
652 	char path[MAXPATHLEN];
653 	int err;
654 
655 	if (!config_file_path(template, path))
656 		return (Z_MISC_FS);
657 
658 	if ((err = zonecfg_get_handle_impl(template, path, handle)) != Z_OK)
659 		return (err);
660 	handle->zone_dh_newzone = B_TRUE;
661 	return (setrootattr(handle, DTD_ATTR_NAME, zonename));
662 }
663 
664 int
665 zonecfg_get_xml_handle(const char *path, zone_dochandle_t handle)
666 {
667 	struct stat buf;
668 	int err;
669 
670 	if (stat(path, &buf) == -1)
671 		return (Z_MISC_FS);
672 
673 	if ((err = zonecfg_get_handle_impl("xml", path, handle)) != Z_OK)
674 		return (err);
675 	handle->zone_dh_newzone = B_TRUE;
676 	return (Z_OK);
677 }
678 
679 /*
680  * Initialize two handles from the manifest read on fd.  The rem_handle
681  * is initialized from the input file, including the sw inventory.  The
682  * local_handle is initialized with the same zone configuration but with
683  * no sw inventory.
684  */
685 int
686 zonecfg_attach_manifest(int fd, zone_dochandle_t local_handle,
687     zone_dochandle_t rem_handle)
688 {
689 	xmlValidCtxtPtr cvp;
690 	int valid;
691 
692 	/* load the manifest into the handle for the remote system */
693 	if ((rem_handle->zone_dh_doc = xmlReadFd(fd, NULL, NULL, 0)) == NULL) {
694 		return (Z_INVALID_DOCUMENT);
695 	}
696 	if ((cvp = xmlNewValidCtxt()) == NULL)
697 		return (Z_NOMEM);
698 	cvp->error = zonecfg_error_func;
699 	cvp->warning = zonecfg_error_func;
700 	valid = xmlValidateDocument(cvp, rem_handle->zone_dh_doc);
701 	xmlFreeValidCtxt(cvp);
702 	if (valid == 0)
703 		return (Z_INVALID_DOCUMENT);
704 
705 	/* delete any comments such as inherited Sun copyright / ident str */
706 	stripcomments(rem_handle);
707 
708 	rem_handle->zone_dh_newzone = B_TRUE;
709 	rem_handle->zone_dh_sw_inv = B_TRUE;
710 
711 	/*
712 	 * Now use the remote system handle to generate a local system handle
713 	 * with an identical zones configuration but no sw inventory.
714 	 */
715 	if ((local_handle->zone_dh_doc = xmlCopyDoc(rem_handle->zone_dh_doc,
716 	    1)) == NULL) {
717 		return (Z_INVALID_DOCUMENT);
718 	}
719 
720 	/*
721 	 * We need to re-run xmlValidateDocument on local_handle to properly
722 	 * update the in-core representation of the configuration.
723 	 */
724 	if ((cvp = xmlNewValidCtxt()) == NULL)
725 		return (Z_NOMEM);
726 	cvp->error = zonecfg_error_func;
727 	cvp->warning = zonecfg_error_func;
728 	valid = xmlValidateDocument(cvp, local_handle->zone_dh_doc);
729 	xmlFreeValidCtxt(cvp);
730 	if (valid == 0)
731 		return (Z_INVALID_DOCUMENT);
732 
733 	strip_sw_inv(local_handle);
734 
735 	local_handle->zone_dh_newzone = B_TRUE;
736 	local_handle->zone_dh_sw_inv = B_FALSE;
737 
738 	return (Z_OK);
739 }
740 
741 static boolean_t
742 is_renaming(zone_dochandle_t handle)
743 {
744 	if (handle->zone_dh_newzone)
745 		return (B_FALSE);
746 	if (strlen(handle->zone_dh_delete_name) > 0)
747 		return (B_TRUE);
748 	return (B_FALSE);
749 }
750 
751 static boolean_t
752 is_new(zone_dochandle_t handle)
753 {
754 	return (handle->zone_dh_newzone || handle->zone_dh_snapshot);
755 }
756 
757 static boolean_t
758 is_snapshot(zone_dochandle_t handle)
759 {
760 	return (handle->zone_dh_snapshot);
761 }
762 
763 /*
764  * It would be great to be able to use libc's ctype(3c) macros, but we
765  * can't, as they are locale sensitive, and it would break our limited thread
766  * safety if this routine had to change the app locale on the fly.
767  */
768 int
769 zonecfg_validate_zonename(const char *zone)
770 {
771 	int i;
772 
773 	if (strcmp(zone, GLOBAL_ZONENAME) == 0)
774 		return (Z_BOGUS_ZONE_NAME);
775 
776 	if (strlen(zone) >= ZONENAME_MAX)
777 		return (Z_BOGUS_ZONE_NAME);
778 
779 	if (!((zone[0] >= 'a' && zone[0] <= 'z') ||
780 	    (zone[0] >= 'A' && zone[0] <= 'Z') ||
781 	    (zone[0] >= '0' && zone[0] <= '9')))
782 		return (Z_BOGUS_ZONE_NAME);
783 
784 	for (i = 1; zone[i] != '\0'; i++) {
785 		if (!((zone[i] >= 'a' && zone[i] <= 'z') ||
786 		    (zone[i] >= 'A' && zone[i] <= 'Z') ||
787 		    (zone[i] >= '0' && zone[i] <= '9') ||
788 		    (zone[i] == '-') || (zone[i] == '_') || (zone[i] == '.')))
789 			return (Z_BOGUS_ZONE_NAME);
790 	}
791 
792 	return (Z_OK);
793 }
794 
795 /*
796  * Changing the zone name requires us to track both the old and new
797  * name of the zone until commit time.
798  */
799 int
800 zonecfg_get_name(zone_dochandle_t handle, char *name, size_t namesize)
801 {
802 	return (getrootattr(handle, DTD_ATTR_NAME, name, namesize));
803 }
804 
805 int
806 zonecfg_set_name(zone_dochandle_t handle, char *name)
807 {
808 	zone_state_t state;
809 	char curname[ZONENAME_MAX], old_delname[ZONENAME_MAX];
810 	int err;
811 
812 	if ((err = getrootattr(handle, DTD_ATTR_NAME, curname,
813 	    sizeof (curname))) != Z_OK)
814 		return (err);
815 
816 	if (strcmp(name, curname) == 0)
817 		return (Z_OK);
818 
819 	/*
820 	 * Switching zone names to one beginning with SUNW is not permitted.
821 	 */
822 	if (strncmp(name, "SUNW", 4) == 0)
823 		return (Z_BOGUS_ZONE_NAME);
824 
825 	if ((err = zonecfg_validate_zonename(name)) != Z_OK)
826 		return (err);
827 
828 	/*
829 	 * Setting the name back to the original name (effectively a revert of
830 	 * the name) is fine.  But if we carry on, we'll falsely identify the
831 	 * name as "in use," so special case here.
832 	 */
833 	if (strcmp(name, handle->zone_dh_delete_name) == 0) {
834 		err = setrootattr(handle, DTD_ATTR_NAME, name);
835 		handle->zone_dh_delete_name[0] = '\0';
836 		return (err);
837 	}
838 
839 	/* Check to see if new name chosen is already in use */
840 	if (zone_get_state(name, &state) != Z_NO_ZONE)
841 		return (Z_NAME_IN_USE);
842 
843 	/*
844 	 * If this isn't already "new" or in a renaming transition, then
845 	 * we're initiating a rename here; so stash the "delete name"
846 	 * (i.e. the name of the zone we'll be removing) for the rename.
847 	 */
848 	(void) strlcpy(old_delname, handle->zone_dh_delete_name,
849 	    sizeof (old_delname));
850 	if (!is_new(handle) && !is_renaming(handle)) {
851 		/*
852 		 * Name change is allowed only when the zone we're altering
853 		 * is not ready or running.
854 		 */
855 		err = zone_get_state(curname, &state);
856 		if (err == Z_OK) {
857 			if (state > ZONE_STATE_INSTALLED)
858 				return (Z_BAD_ZONE_STATE);
859 		} else if (err != Z_NO_ZONE) {
860 			return (err);
861 		}
862 
863 		(void) strlcpy(handle->zone_dh_delete_name, curname,
864 		    sizeof (handle->zone_dh_delete_name));
865 		assert(is_renaming(handle));
866 	} else if (is_renaming(handle)) {
867 		err = zone_get_state(handle->zone_dh_delete_name, &state);
868 		if (err == Z_OK) {
869 			if (state > ZONE_STATE_INSTALLED)
870 				return (Z_BAD_ZONE_STATE);
871 		} else if (err != Z_NO_ZONE) {
872 			return (err);
873 		}
874 	}
875 
876 	if ((err = setrootattr(handle, DTD_ATTR_NAME, name)) != Z_OK) {
877 		/*
878 		 * Restore the deletename to whatever it was at the
879 		 * top of the routine, since we've had a failure.
880 		 */
881 		(void) strlcpy(handle->zone_dh_delete_name, old_delname,
882 		    sizeof (handle->zone_dh_delete_name));
883 		return (err);
884 	}
885 
886 	return (Z_OK);
887 }
888 
889 int
890 zonecfg_get_zonepath(zone_dochandle_t handle, char *path, size_t pathsize)
891 {
892 	size_t len;
893 
894 	if ((len = strlcpy(path, zonecfg_root, pathsize)) >= pathsize)
895 		return (Z_TOO_BIG);
896 	return (getrootattr(handle, DTD_ATTR_ZONEPATH, path + len,
897 	    pathsize - len));
898 }
899 
900 int
901 zonecfg_set_zonepath(zone_dochandle_t handle, char *zonepath)
902 {
903 	size_t len;
904 
905 	/*
906 	 * The user deals in absolute paths in the running global zone, but the
907 	 * internal configuration files deal with boot environment relative
908 	 * paths.  Strip out the alternate root when specified.
909 	 */
910 	len = strlen(zonecfg_root);
911 	if (strncmp(zonepath, zonecfg_root, len) != 0 || zonepath[len] != '/')
912 		return (Z_BAD_PROPERTY);
913 	zonepath += len;
914 	return (setrootattr(handle, DTD_ATTR_ZONEPATH, zonepath));
915 }
916 
917 int
918 zonecfg_get_brand(zone_dochandle_t handle, char *brand, size_t brandsize)
919 {
920 	int ret, sz;
921 
922 	ret = getrootattr(handle, DTD_ATTR_BRAND, brand, brandsize);
923 
924 	/* If the zone has no brand, it is native. */
925 	if (ret == Z_OK && brand[0] == '\0') {
926 		sz = strlcpy(brand, NATIVE_BRAND_NAME, brandsize);
927 		if (sz >= brandsize)
928 			ret = Z_TOO_BIG;
929 		else
930 			ret = Z_OK;
931 	}
932 
933 	return (ret);
934 }
935 
936 int
937 zonecfg_set_brand(zone_dochandle_t handle, char *brand)
938 {
939 	return (setrootattr(handle, DTD_ATTR_BRAND, brand));
940 }
941 
942 int
943 zonecfg_get_autoboot(zone_dochandle_t handle, boolean_t *autoboot)
944 {
945 	char autobootstr[DTD_ENTITY_BOOL_LEN];
946 	int ret;
947 
948 	if ((ret = getrootattr(handle, DTD_ATTR_AUTOBOOT, autobootstr,
949 	    sizeof (autobootstr))) != Z_OK)
950 		return (ret);
951 
952 	if (strcmp(autobootstr, DTD_ENTITY_TRUE) == 0)
953 		*autoboot = B_TRUE;
954 	else if (strcmp(autobootstr, DTD_ENTITY_FALSE) == 0)
955 		*autoboot = B_FALSE;
956 	else
957 		ret = Z_BAD_PROPERTY;
958 	return (ret);
959 }
960 
961 int
962 zonecfg_set_autoboot(zone_dochandle_t handle, boolean_t autoboot)
963 {
964 	return (setrootattr(handle, DTD_ATTR_AUTOBOOT,
965 	    autoboot ? DTD_ENTITY_TRUE : DTD_ENTITY_FALSE));
966 }
967 
968 int
969 zonecfg_get_pool(zone_dochandle_t handle, char *pool, size_t poolsize)
970 {
971 	return (getrootattr(handle, DTD_ATTR_POOL, pool, poolsize));
972 }
973 
974 int
975 zonecfg_set_pool(zone_dochandle_t handle, char *pool)
976 {
977 	return (setrootattr(handle, DTD_ATTR_POOL, pool));
978 }
979 
980 int
981 zonecfg_get_limitpriv(zone_dochandle_t handle, char **limitpriv)
982 {
983 	return (get_alloc_rootattr(handle, DTD_ATTR_LIMITPRIV, limitpriv));
984 }
985 
986 int
987 zonecfg_set_limitpriv(zone_dochandle_t handle, char *limitpriv)
988 {
989 	return (setrootattr(handle, DTD_ATTR_LIMITPRIV, limitpriv));
990 }
991 
992 int
993 zonecfg_get_bootargs(zone_dochandle_t handle, char *bargs, size_t bargssize)
994 {
995 	return (getrootattr(handle, DTD_ATTR_BOOTARGS, bargs, bargssize));
996 }
997 
998 int
999 zonecfg_set_bootargs(zone_dochandle_t handle, char *bargs)
1000 {
1001 	return (setrootattr(handle, DTD_ATTR_BOOTARGS, bargs));
1002 }
1003 
1004 int
1005 zonecfg_get_sched_class(zone_dochandle_t handle, char *sched, size_t schedsize)
1006 {
1007 	return (getrootattr(handle, DTD_ATTR_SCHED, sched, schedsize));
1008 }
1009 
1010 int
1011 zonecfg_set_sched(zone_dochandle_t handle, char *sched)
1012 {
1013 	return (setrootattr(handle, DTD_ATTR_SCHED, sched));
1014 }
1015 
1016 /*
1017  * /etc/zones/index caches a vital piece of information which is also
1018  * in the <zonename>.xml file: the path to the zone.  This is for performance,
1019  * since we need to walk all zonepath's in order to be able to detect conflicts
1020  * (see crosscheck_zonepaths() in the zoneadm command).
1021  *
1022  * An additional complexity is that when doing a rename, we'd like the entire
1023  * index update operation (rename, and potential state changes) to be atomic.
1024  * In general, the operation of this function should succeed or fail as
1025  * a unit.
1026  */
1027 int
1028 zonecfg_refresh_index_file(zone_dochandle_t handle)
1029 {
1030 	char name[ZONENAME_MAX], zonepath[MAXPATHLEN];
1031 	struct zoneent ze;
1032 	int err;
1033 	int opcode;
1034 	char *zn;
1035 
1036 	bzero(&ze, sizeof (ze));
1037 	ze.zone_state = -1;	/* Preserve existing state in index */
1038 
1039 	if ((err = zonecfg_get_name(handle, name, sizeof (name))) != Z_OK)
1040 		return (err);
1041 	(void) strlcpy(ze.zone_name, name, sizeof (ze.zone_name));
1042 
1043 	if ((err = zonecfg_get_zonepath(handle, zonepath,
1044 	    sizeof (zonepath))) != Z_OK)
1045 		return (err);
1046 	(void) strlcpy(ze.zone_path, zonepath + strlen(zonecfg_root),
1047 	    sizeof (ze.zone_path));
1048 
1049 	if (is_renaming(handle)) {
1050 		opcode = PZE_MODIFY;
1051 		(void) strlcpy(ze.zone_name, handle->zone_dh_delete_name,
1052 		    sizeof (ze.zone_name));
1053 		(void) strlcpy(ze.zone_newname, name, sizeof (ze.zone_newname));
1054 	} else if (is_new(handle)) {
1055 		FILE *cookie;
1056 		/*
1057 		 * Be tolerant of the zone already existing in the index file,
1058 		 * since we might be forcibly overwriting an existing
1059 		 * configuration with a new one (for example 'create -F'
1060 		 * in zonecfg).
1061 		 */
1062 		opcode = PZE_ADD;
1063 		cookie = setzoneent();
1064 		while ((zn = getzoneent(cookie)) != NULL) {
1065 			if (strcmp(zn, name) == 0) {
1066 				opcode = PZE_MODIFY;
1067 				free(zn);
1068 				break;
1069 			}
1070 			free(zn);
1071 		}
1072 		endzoneent(cookie);
1073 		ze.zone_state = ZONE_STATE_CONFIGURED;
1074 	} else {
1075 		opcode = PZE_MODIFY;
1076 	}
1077 
1078 	if ((err = putzoneent(&ze, opcode)) != Z_OK)
1079 		return (err);
1080 
1081 	return (Z_OK);
1082 }
1083 
1084 /*
1085  * The goal of this routine is to cause the index file update and the
1086  * document save to happen as an atomic operation.  We do the document
1087  * first, saving a backup copy using a hard link; if that succeeds, we go
1088  * on to the index.  If that fails, we roll the document back into place.
1089  *
1090  * Strategy:
1091  *
1092  * New zone 'foo' configuration:
1093  * 	Create tmpfile (zonecfg.xxxxxx)
1094  * 	Write XML to tmpfile
1095  * 	Rename tmpfile to xmlfile (zonecfg.xxxxxx -> foo.xml)
1096  * 	Add entry to index file
1097  * 	If it fails, delete foo.xml, leaving nothing behind.
1098  *
1099  * Save existing zone 'foo':
1100  * 	Make backup of foo.xml -> .backup
1101  * 	Create tmpfile (zonecfg.xxxxxx)
1102  * 	Write XML to tmpfile
1103  * 	Rename tmpfile to xmlfile (zonecfg.xxxxxx -> foo.xml)
1104  * 	Modify index file as needed
1105  * 	If it fails, recover from .backup -> foo.xml
1106  *
1107  * Rename 'foo' to 'bar':
1108  * 	Create tmpfile (zonecfg.xxxxxx)
1109  * 	Write XML to tmpfile
1110  * 	Rename tmpfile to xmlfile (zonecfg.xxxxxx -> bar.xml)
1111  * 	Add entry for 'bar' to index file, Remove entry for 'foo' (refresh)
1112  * 	If it fails, delete bar.xml; foo.xml is left behind.
1113  */
1114 static int
1115 zonecfg_save_impl(zone_dochandle_t handle, char *filename)
1116 {
1117 	char tmpfile[MAXPATHLEN];
1118 	char bakdir[MAXPATHLEN], bakbase[MAXPATHLEN], bakfile[MAXPATHLEN];
1119 	int tmpfd, err, valid;
1120 	xmlValidCtxt cvp = { NULL };
1121 	boolean_t backup;
1122 
1123 	(void) strlcpy(tmpfile, filename, sizeof (tmpfile));
1124 	(void) dirname(tmpfile);
1125 	(void) strlcat(tmpfile, _PATH_TMPFILE, sizeof (tmpfile));
1126 
1127 	tmpfd = mkstemp(tmpfile);
1128 	if (tmpfd == -1) {
1129 		(void) unlink(tmpfile);
1130 		return (Z_TEMP_FILE);
1131 	}
1132 	(void) close(tmpfd);
1133 
1134 	cvp.error = zonecfg_error_func;
1135 	cvp.warning = zonecfg_error_func;
1136 
1137 	/*
1138 	 * We do a final validation of the document.  Since the library has
1139 	 * malfunctioned if it fails to validate, we follow-up with an
1140 	 * assert() that the doc is valid.
1141 	 */
1142 	valid = xmlValidateDocument(&cvp, handle->zone_dh_doc);
1143 	assert(valid != 0);
1144 
1145 	if (xmlSaveFormatFile(tmpfile, handle->zone_dh_doc, 1) <= 0)
1146 		goto err;
1147 
1148 	(void) chmod(tmpfile, 0644);
1149 
1150 	/*
1151 	 * In the event we are doing a standard save, hard link a copy of the
1152 	 * original file in .backup.<pid>.filename so we can restore it if
1153 	 * something goes wrong.
1154 	 */
1155 	if (!is_new(handle) && !is_renaming(handle)) {
1156 		backup = B_TRUE;
1157 
1158 		(void) strlcpy(bakdir, filename, sizeof (bakdir));
1159 		(void) strlcpy(bakbase, filename, sizeof (bakbase));
1160 		(void) snprintf(bakfile, sizeof (bakfile), "%s/.backup.%d.%s",
1161 		    dirname(bakdir), getpid(), basename(bakbase));
1162 
1163 		if (link(filename, bakfile) == -1) {
1164 			err = errno;
1165 			(void) unlink(tmpfile);
1166 			if (errno == EACCES)
1167 				return (Z_ACCES);
1168 			return (Z_MISC_FS);
1169 		}
1170 	}
1171 
1172 	/*
1173 	 * Move the new document over top of the old.
1174 	 * i.e.:   zonecfg.XXXXXX  ->  myzone.xml
1175 	 */
1176 	if (rename(tmpfile, filename) == -1) {
1177 		err = errno;
1178 		(void) unlink(tmpfile);
1179 		if (backup)
1180 			(void) unlink(bakfile);
1181 		if (err == EACCES)
1182 			return (Z_ACCES);
1183 		return (Z_MISC_FS);
1184 	}
1185 
1186 	/*
1187 	 * If this is a snapshot, we're done-- don't add an index entry.
1188 	 */
1189 	if (is_snapshot(handle))
1190 		return (Z_OK);
1191 
1192 	/* now update the index file to reflect whatever we just did */
1193 	if ((err = zonecfg_refresh_index_file(handle)) != Z_OK) {
1194 		if (backup) {
1195 			/*
1196 			 * Try to restore from our backup.
1197 			 */
1198 			(void) unlink(filename);
1199 			(void) rename(bakfile, filename);
1200 		} else {
1201 			/*
1202 			 * Either the zone is new, in which case we can delete
1203 			 * new.xml, or we're doing a rename, so ditto.
1204 			 */
1205 			assert(is_new(handle) || is_renaming(handle));
1206 			(void) unlink(filename);
1207 		}
1208 		return (Z_UPDATING_INDEX);
1209 	}
1210 
1211 	if (backup)
1212 		(void) unlink(bakfile);
1213 
1214 	return (Z_OK);
1215 
1216 err:
1217 	(void) unlink(tmpfile);
1218 	return (Z_SAVING_FILE);
1219 }
1220 
1221 int
1222 zonecfg_save(zone_dochandle_t handle)
1223 {
1224 	char zname[ZONENAME_MAX], path[MAXPATHLEN];
1225 	char delpath[MAXPATHLEN];
1226 	int err = Z_SAVING_FILE;
1227 
1228 	if (zonecfg_check_handle(handle) != Z_OK)
1229 		return (Z_BAD_HANDLE);
1230 
1231 	/*
1232 	 * We don't support saving snapshots or a tree containing a sw
1233 	 * inventory at this time.
1234 	 */
1235 	if (handle->zone_dh_snapshot || handle->zone_dh_sw_inv)
1236 		return (Z_INVAL);
1237 
1238 	if ((err = zonecfg_get_name(handle, zname, sizeof (zname))) != Z_OK)
1239 		return (err);
1240 
1241 	if (!config_file_path(zname, path))
1242 		return (Z_MISC_FS);
1243 
1244 	addcomment(handle, "\n    DO NOT EDIT THIS "
1245 	    "FILE.  Use zonecfg(1M) instead.\n");
1246 
1247 	err = zonecfg_save_impl(handle, path);
1248 
1249 	stripcomments(handle);
1250 
1251 	if (err != Z_OK)
1252 		return (err);
1253 
1254 	handle->zone_dh_newzone = B_FALSE;
1255 
1256 	if (is_renaming(handle)) {
1257 		if (config_file_path(handle->zone_dh_delete_name, delpath))
1258 			(void) unlink(delpath);
1259 		handle->zone_dh_delete_name[0] = '\0';
1260 	}
1261 
1262 	return (Z_OK);
1263 }
1264 
1265 int
1266 zonecfg_verify_save(zone_dochandle_t handle, char *filename)
1267 {
1268 	int valid;
1269 
1270 	xmlValidCtxt cvp = { NULL };
1271 
1272 	if (zonecfg_check_handle(handle) != Z_OK)
1273 		return (Z_BAD_HANDLE);
1274 
1275 	cvp.error = zonecfg_error_func;
1276 	cvp.warning = zonecfg_error_func;
1277 
1278 	/*
1279 	 * We do a final validation of the document.  Since the library has
1280 	 * malfunctioned if it fails to validate, we follow-up with an
1281 	 * assert() that the doc is valid.
1282 	 */
1283 	valid = xmlValidateDocument(&cvp, handle->zone_dh_doc);
1284 	assert(valid != 0);
1285 
1286 	if (xmlSaveFormatFile(filename, handle->zone_dh_doc, 1) <= 0)
1287 		return (Z_SAVING_FILE);
1288 
1289 	return (Z_OK);
1290 }
1291 
1292 int
1293 zonecfg_detach_save(zone_dochandle_t handle, uint_t flags)
1294 {
1295 	char zname[ZONENAME_MAX];
1296 	char path[MAXPATHLEN];
1297 	char migpath[MAXPATHLEN];
1298 	xmlValidCtxt cvp = { NULL };
1299 	int err = Z_SAVING_FILE;
1300 	int valid;
1301 
1302 	if (zonecfg_check_handle(handle) != Z_OK)
1303 		return (Z_BAD_HANDLE);
1304 
1305 	if (flags & ZONE_DRY_RUN) {
1306 		(void) strlcpy(migpath, "-", sizeof (migpath));
1307 	} else {
1308 		if ((err = zonecfg_get_name(handle, zname, sizeof (zname)))
1309 		    != Z_OK)
1310 			return (err);
1311 
1312 		if ((err = zone_get_zonepath(zname, path, sizeof (path)))
1313 		    != Z_OK)
1314 			return (err);
1315 
1316 		if (snprintf(migpath, sizeof (migpath), "%s/%s", path,
1317 		    ZONE_DETACHED) >= sizeof (migpath))
1318 			return (Z_NOMEM);
1319 	}
1320 
1321 	if ((err = operation_prep(handle)) != Z_OK)
1322 		return (err);
1323 
1324 	addcomment(handle, "\n    DO NOT EDIT THIS FILE.  "
1325 	    "Use zonecfg(1M) and zoneadm(1M) attach.\n");
1326 
1327 	cvp.error = zonecfg_error_func;
1328 	cvp.warning = zonecfg_error_func;
1329 
1330 	/*
1331 	 * We do a final validation of the document.  Since the library has
1332 	 * malfunctioned if it fails to validate, we follow-up with an
1333 	 * assert() that the doc is valid.
1334 	 */
1335 	valid = xmlValidateDocument(&cvp, handle->zone_dh_doc);
1336 	assert(valid != 0);
1337 
1338 	if (xmlSaveFormatFile(migpath, handle->zone_dh_doc, 1) <= 0)
1339 		return (Z_SAVING_FILE);
1340 
1341 	if (!(flags & ZONE_DRY_RUN))
1342 		(void) chmod(migpath, 0644);
1343 
1344 	stripcomments(handle);
1345 
1346 	handle->zone_dh_newzone = B_FALSE;
1347 
1348 	return (Z_OK);
1349 }
1350 
1351 boolean_t
1352 zonecfg_detached(const char *path)
1353 {
1354 	char		migpath[MAXPATHLEN];
1355 	struct stat	buf;
1356 
1357 	if (snprintf(migpath, sizeof (migpath), "%s/%s", path, ZONE_DETACHED) >=
1358 	    sizeof (migpath))
1359 		return (B_FALSE);
1360 
1361 	if (stat(migpath, &buf) != -1)
1362 		return (B_TRUE);
1363 
1364 	return (B_FALSE);
1365 }
1366 
1367 void
1368 zonecfg_rm_detached(zone_dochandle_t handle, boolean_t forced)
1369 {
1370 	char zname[ZONENAME_MAX];
1371 	char path[MAXPATHLEN];
1372 	char detached[MAXPATHLEN];
1373 	char attached[MAXPATHLEN];
1374 
1375 	if (zonecfg_check_handle(handle) != Z_OK)
1376 		return;
1377 
1378 	if (zonecfg_get_name(handle, zname, sizeof (zname)) != Z_OK)
1379 		return;
1380 
1381 	if (zone_get_zonepath(zname, path, sizeof (path)) != Z_OK)
1382 		return;
1383 
1384 	(void) snprintf(detached, sizeof (detached), "%s/%s", path,
1385 	    ZONE_DETACHED);
1386 	(void) snprintf(attached, sizeof (attached), "%s/%s", path,
1387 	    ATTACH_FORCED);
1388 
1389 	if (forced) {
1390 		(void) rename(detached, attached);
1391 	} else {
1392 		(void) unlink(attached);
1393 		(void) unlink(detached);
1394 	}
1395 }
1396 
1397 /*
1398  * Special case: if access(2) fails with ENOENT, then try again using
1399  * ZONE_CONFIG_ROOT instead of config_file_path(zonename).  This is how we
1400  * work around the case of a config file which has not been created yet:
1401  * the user will need access to the directory so use that as a heuristic.
1402  */
1403 
1404 int
1405 zonecfg_access(const char *zonename, int amode)
1406 {
1407 	char path[MAXPATHLEN];
1408 
1409 	if (!config_file_path(zonename, path))
1410 		return (Z_INVAL);
1411 	if (access(path, amode) == 0)
1412 		return (Z_OK);
1413 	if (errno == ENOENT) {
1414 		if (snprintf(path, sizeof (path), "%s%s", zonecfg_root,
1415 		    ZONE_CONFIG_ROOT) >= sizeof (path))
1416 			return (Z_INVAL);
1417 		if (access(path, amode) == 0)
1418 			return (Z_OK);
1419 	}
1420 	if (errno == EACCES)
1421 		return (Z_ACCES);
1422 	if (errno == EINVAL)
1423 		return (Z_INVAL);
1424 	return (Z_MISC_FS);
1425 }
1426 
1427 int
1428 zonecfg_create_snapshot(const char *zonename)
1429 {
1430 	zone_dochandle_t handle;
1431 	char path[MAXPATHLEN], zonepath[MAXPATHLEN], rpath[MAXPATHLEN];
1432 	int error = Z_OK, res;
1433 
1434 	if ((handle = zonecfg_init_handle()) == NULL) {
1435 		return (Z_NOMEM);
1436 	}
1437 
1438 	handle->zone_dh_newzone = B_TRUE;
1439 	handle->zone_dh_snapshot = B_TRUE;
1440 
1441 	if ((error = zonecfg_get_handle(zonename, handle)) != Z_OK)
1442 		goto out;
1443 	if ((error = operation_prep(handle)) != Z_OK)
1444 		goto out;
1445 	error = zonecfg_get_zonepath(handle, zonepath, sizeof (zonepath));
1446 	if (error != Z_OK)
1447 		goto out;
1448 	if ((res = resolvepath(zonepath, rpath, sizeof (rpath))) == -1) {
1449 		error = Z_RESOLVED_PATH;
1450 		goto out;
1451 	}
1452 	/*
1453 	 * If the resolved path is not the same as the original path, then
1454 	 * save the resolved path in the snapshot, thus preventing any
1455 	 * potential problems down the line when zoneadmd goes to unmount
1456 	 * file systems and depends on initial string matches with resolved
1457 	 * paths.
1458 	 */
1459 	rpath[res] = '\0';
1460 	if (strcmp(zonepath, rpath) != 0) {
1461 		if ((error = zonecfg_set_zonepath(handle, rpath)) != Z_OK)
1462 			goto out;
1463 	}
1464 	if (snprintf(path, sizeof (path), "%s%s", zonecfg_root,
1465 	    ZONE_SNAPSHOT_ROOT) >= sizeof (path)) {
1466 		error = Z_MISC_FS;
1467 		goto out;
1468 	}
1469 	if ((mkdir(path, S_IRWXU) == -1) && (errno != EEXIST)) {
1470 		error = Z_MISC_FS;
1471 		goto out;
1472 	}
1473 
1474 	if (!snap_file_path(zonename, path)) {
1475 		error = Z_MISC_FS;
1476 		goto out;
1477 	}
1478 
1479 	addcomment(handle, "\n    DO NOT EDIT THIS FILE.  "
1480 	    "It is a snapshot of running zone state.\n");
1481 
1482 	error = zonecfg_save_impl(handle, path);
1483 
1484 	stripcomments(handle);
1485 
1486 out:
1487 	zonecfg_fini_handle(handle);
1488 	return (error);
1489 }
1490 
1491 int
1492 zonecfg_get_iptype(zone_dochandle_t handle, zone_iptype_t *iptypep)
1493 {
1494 	char property[10]; /* 10 is big enough for "shared"/"exclusive" */
1495 	int err;
1496 
1497 	err = getrootattr(handle, DTD_ATTR_IPTYPE, property, sizeof (property));
1498 	if (err == Z_BAD_PROPERTY) {
1499 		/* Return default value */
1500 		*iptypep = ZS_SHARED;
1501 		return (Z_OK);
1502 	} else if (err != Z_OK) {
1503 		return (err);
1504 	}
1505 
1506 	if (strlen(property) == 0 ||
1507 	    strcmp(property, "shared") == 0)
1508 		*iptypep = ZS_SHARED;
1509 	else if (strcmp(property, "exclusive") == 0)
1510 		*iptypep = ZS_EXCLUSIVE;
1511 	else
1512 		return (Z_INVAL);
1513 
1514 	return (Z_OK);
1515 }
1516 
1517 int
1518 zonecfg_set_iptype(zone_dochandle_t handle, zone_iptype_t iptype)
1519 {
1520 	xmlNodePtr cur;
1521 
1522 	if (handle == NULL)
1523 		return (Z_INVAL);
1524 
1525 	cur = xmlDocGetRootElement(handle->zone_dh_doc);
1526 	if (cur == NULL) {
1527 		return (Z_EMPTY_DOCUMENT);
1528 	}
1529 
1530 	if (xmlStrcmp(cur->name, DTD_ELEM_ZONE) != 0) {
1531 		return (Z_WRONG_DOC_TYPE);
1532 	}
1533 	switch (iptype) {
1534 	case ZS_SHARED:
1535 		/*
1536 		 * Since "shared" is the default, we don't write it to the
1537 		 * configuration file, so that it's easier to migrate those
1538 		 * zones elsewhere, eg., to systems which are not IP-Instances
1539 		 * aware.
1540 		 * xmlUnsetProp only fails when the attribute doesn't exist,
1541 		 * which we don't care.
1542 		 */
1543 		(void) xmlUnsetProp(cur, DTD_ATTR_IPTYPE);
1544 		break;
1545 	case ZS_EXCLUSIVE:
1546 		if (xmlSetProp(cur, DTD_ATTR_IPTYPE,
1547 		    (const xmlChar *) "exclusive") == NULL)
1548 			return (Z_INVAL);
1549 		break;
1550 	}
1551 	return (Z_OK);
1552 }
1553 
1554 static int
1555 newprop(xmlNodePtr node, const xmlChar *attrname, char *src)
1556 {
1557 	xmlAttrPtr newattr;
1558 
1559 	newattr = xmlNewProp(node, attrname, (xmlChar *)src);
1560 	if (newattr == NULL) {
1561 		xmlUnlinkNode(node);
1562 		xmlFreeNode(node);
1563 		return (Z_BAD_PROPERTY);
1564 	}
1565 	return (Z_OK);
1566 }
1567 
1568 static int
1569 zonecfg_add_filesystem_core(zone_dochandle_t handle, struct zone_fstab *tabptr)
1570 {
1571 	xmlNodePtr newnode, cur = handle->zone_dh_cur, options_node;
1572 	zone_fsopt_t *ptr;
1573 	int err;
1574 
1575 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_FS, NULL);
1576 	if ((err = newprop(newnode, DTD_ATTR_SPECIAL,
1577 	    tabptr->zone_fs_special)) != Z_OK)
1578 		return (err);
1579 	if (tabptr->zone_fs_raw[0] != '\0' &&
1580 	    (err = newprop(newnode, DTD_ATTR_RAW, tabptr->zone_fs_raw)) != Z_OK)
1581 		return (err);
1582 	if ((err = newprop(newnode, DTD_ATTR_DIR, tabptr->zone_fs_dir)) != Z_OK)
1583 		return (err);
1584 	if ((err = newprop(newnode, DTD_ATTR_TYPE,
1585 	    tabptr->zone_fs_type)) != Z_OK)
1586 		return (err);
1587 	if (tabptr->zone_fs_options != NULL) {
1588 		for (ptr = tabptr->zone_fs_options; ptr != NULL;
1589 		    ptr = ptr->zone_fsopt_next) {
1590 			options_node = xmlNewTextChild(newnode, NULL,
1591 			    DTD_ELEM_FSOPTION, NULL);
1592 			if ((err = newprop(options_node, DTD_ATTR_NAME,
1593 			    ptr->zone_fsopt_opt)) != Z_OK)
1594 				return (err);
1595 		}
1596 	}
1597 	return (Z_OK);
1598 }
1599 
1600 int
1601 zonecfg_add_filesystem(zone_dochandle_t handle, struct zone_fstab *tabptr)
1602 {
1603 	int err;
1604 
1605 	if (tabptr == NULL)
1606 		return (Z_INVAL);
1607 
1608 	if ((err = operation_prep(handle)) != Z_OK)
1609 		return (err);
1610 
1611 	if ((err = zonecfg_add_filesystem_core(handle, tabptr)) != Z_OK)
1612 		return (err);
1613 
1614 	return (Z_OK);
1615 }
1616 
1617 static int
1618 zonecfg_add_ipd_core(zone_dochandle_t handle, struct zone_fstab *tabptr)
1619 {
1620 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
1621 	int err;
1622 
1623 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_IPD, NULL);
1624 	if ((err = newprop(newnode, DTD_ATTR_DIR, tabptr->zone_fs_dir)) != Z_OK)
1625 		return (err);
1626 	return (Z_OK);
1627 }
1628 
1629 int
1630 zonecfg_add_ipd(zone_dochandle_t handle, struct zone_fstab *tabptr)
1631 {
1632 	int err;
1633 
1634 	if (tabptr == NULL)
1635 		return (Z_INVAL);
1636 
1637 	if ((err = operation_prep(handle)) != Z_OK)
1638 		return (err);
1639 
1640 	if ((err = zonecfg_add_ipd_core(handle, tabptr)) != Z_OK)
1641 		return (err);
1642 
1643 	return (Z_OK);
1644 }
1645 
1646 int
1647 zonecfg_add_fs_option(struct zone_fstab *tabptr, char *option)
1648 {
1649 	zone_fsopt_t *last, *old, *new;
1650 
1651 	last = tabptr->zone_fs_options;
1652 	for (old = last; old != NULL; old = old->zone_fsopt_next)
1653 		last = old;	/* walk to the end of the list */
1654 	new = (zone_fsopt_t *)malloc(sizeof (zone_fsopt_t));
1655 	if (new == NULL)
1656 		return (Z_NOMEM);
1657 	(void) strlcpy(new->zone_fsopt_opt, option,
1658 	    sizeof (new->zone_fsopt_opt));
1659 	new->zone_fsopt_next = NULL;
1660 	if (last == NULL)
1661 		tabptr->zone_fs_options = new;
1662 	else
1663 		last->zone_fsopt_next = new;
1664 	return (Z_OK);
1665 }
1666 
1667 int
1668 zonecfg_remove_fs_option(struct zone_fstab *tabptr, char *option)
1669 {
1670 	zone_fsopt_t *last, *this, *next;
1671 
1672 	last = tabptr->zone_fs_options;
1673 	for (this = last; this != NULL; this = this->zone_fsopt_next) {
1674 		if (strcmp(this->zone_fsopt_opt, option) == 0) {
1675 			next = this->zone_fsopt_next;
1676 			if (this == tabptr->zone_fs_options)
1677 				tabptr->zone_fs_options = next;
1678 			else
1679 				last->zone_fsopt_next = next;
1680 			free(this);
1681 			return (Z_OK);
1682 		} else
1683 			last = this;
1684 	}
1685 	return (Z_NO_PROPERTY_ID);
1686 }
1687 
1688 void
1689 zonecfg_free_fs_option_list(zone_fsopt_t *list)
1690 {
1691 	zone_fsopt_t *this, *next;
1692 
1693 	for (this = list; this != NULL; this = next) {
1694 		next = this->zone_fsopt_next;
1695 		free(this);
1696 	}
1697 }
1698 
1699 void
1700 zonecfg_free_rctl_value_list(struct zone_rctlvaltab *valtab)
1701 {
1702 	if (valtab == NULL)
1703 		return;
1704 	zonecfg_free_rctl_value_list(valtab->zone_rctlval_next);
1705 	free(valtab);
1706 }
1707 
1708 static boolean_t
1709 match_prop(xmlNodePtr cur, const xmlChar *attr, char *user_prop)
1710 {
1711 	xmlChar *gotten_prop;
1712 	int prop_result;
1713 
1714 	gotten_prop = xmlGetProp(cur, attr);
1715 	if (gotten_prop == NULL)	/* shouldn't happen */
1716 		return (B_FALSE);
1717 	prop_result = xmlStrcmp(gotten_prop, (const xmlChar *) user_prop);
1718 	xmlFree(gotten_prop);
1719 	return ((prop_result == 0));
1720 }
1721 
1722 static int
1723 zonecfg_delete_filesystem_core(zone_dochandle_t handle,
1724     struct zone_fstab *tabptr)
1725 {
1726 	xmlNodePtr cur = handle->zone_dh_cur;
1727 	boolean_t dir_match, spec_match, raw_match, type_match;
1728 
1729 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
1730 		if (xmlStrcmp(cur->name, DTD_ELEM_FS))
1731 			continue;
1732 		dir_match = match_prop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir);
1733 		spec_match = match_prop(cur, DTD_ATTR_SPECIAL,
1734 		    tabptr->zone_fs_special);
1735 		raw_match = match_prop(cur, DTD_ATTR_RAW,
1736 		    tabptr->zone_fs_raw);
1737 		type_match = match_prop(cur, DTD_ATTR_TYPE,
1738 		    tabptr->zone_fs_type);
1739 		if (dir_match && spec_match && raw_match && type_match) {
1740 			xmlUnlinkNode(cur);
1741 			xmlFreeNode(cur);
1742 			return (Z_OK);
1743 		}
1744 	}
1745 	return (Z_NO_RESOURCE_ID);
1746 }
1747 
1748 int
1749 zonecfg_delete_filesystem(zone_dochandle_t handle, struct zone_fstab *tabptr)
1750 {
1751 	int err;
1752 
1753 	if (tabptr == NULL)
1754 		return (Z_INVAL);
1755 
1756 	if ((err = operation_prep(handle)) != Z_OK)
1757 		return (err);
1758 
1759 	if ((err = zonecfg_delete_filesystem_core(handle, tabptr)) != Z_OK)
1760 		return (err);
1761 
1762 	return (Z_OK);
1763 }
1764 
1765 int
1766 zonecfg_modify_filesystem(
1767 	zone_dochandle_t handle,
1768 	struct zone_fstab *oldtabptr,
1769 	struct zone_fstab *newtabptr)
1770 {
1771 	int err;
1772 
1773 	if (oldtabptr == NULL || newtabptr == NULL)
1774 		return (Z_INVAL);
1775 
1776 	if ((err = operation_prep(handle)) != Z_OK)
1777 		return (err);
1778 
1779 	if ((err = zonecfg_delete_filesystem_core(handle, oldtabptr)) != Z_OK)
1780 		return (err);
1781 
1782 	if ((err = zonecfg_add_filesystem_core(handle, newtabptr)) != Z_OK)
1783 		return (err);
1784 
1785 	return (Z_OK);
1786 }
1787 
1788 static int
1789 zonecfg_delete_ipd_core(zone_dochandle_t handle, struct zone_fstab *tabptr)
1790 {
1791 	xmlNodePtr cur = handle->zone_dh_cur;
1792 
1793 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
1794 		if (xmlStrcmp(cur->name, DTD_ELEM_IPD))
1795 			continue;
1796 		if (match_prop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir)) {
1797 			xmlUnlinkNode(cur);
1798 			xmlFreeNode(cur);
1799 			return (Z_OK);
1800 		}
1801 	}
1802 	return (Z_NO_RESOURCE_ID);
1803 }
1804 
1805 int
1806 zonecfg_delete_ipd(zone_dochandle_t handle, struct zone_fstab *tabptr)
1807 {
1808 	int err;
1809 
1810 	if (tabptr == NULL)
1811 		return (Z_INVAL);
1812 
1813 	if ((err = operation_prep(handle)) != Z_OK)
1814 		return (err);
1815 
1816 	if ((err = zonecfg_delete_ipd_core(handle, tabptr)) != Z_OK)
1817 		return (err);
1818 
1819 	return (Z_OK);
1820 }
1821 
1822 int
1823 zonecfg_modify_ipd(zone_dochandle_t handle, struct zone_fstab *oldtabptr,
1824     struct zone_fstab *newtabptr)
1825 {
1826 	int err;
1827 
1828 	if (oldtabptr == NULL || newtabptr == NULL)
1829 		return (Z_INVAL);
1830 
1831 	if ((err = operation_prep(handle)) != Z_OK)
1832 		return (err);
1833 
1834 	if ((err = zonecfg_delete_ipd_core(handle, oldtabptr)) != Z_OK)
1835 		return (err);
1836 
1837 	if ((err = zonecfg_add_ipd_core(handle, newtabptr)) != Z_OK)
1838 		return (err);
1839 
1840 	return (Z_OK);
1841 }
1842 
1843 int
1844 zonecfg_lookup_filesystem(
1845 	zone_dochandle_t handle,
1846 	struct zone_fstab *tabptr)
1847 {
1848 	xmlNodePtr cur, options, firstmatch;
1849 	int err;
1850 	char dirname[MAXPATHLEN], special[MAXPATHLEN], raw[MAXPATHLEN];
1851 	char type[FSTYPSZ];
1852 	char options_str[MAX_MNTOPT_STR];
1853 
1854 	if (tabptr == NULL)
1855 		return (Z_INVAL);
1856 
1857 	if ((err = operation_prep(handle)) != Z_OK)
1858 		return (err);
1859 
1860 	/*
1861 	 * Walk the list of children looking for matches on any properties
1862 	 * specified in the fstab parameter.  If more than one resource
1863 	 * matches, we return Z_INSUFFICIENT_SPEC; if none match, we return
1864 	 * Z_NO_RESOURCE_ID.
1865 	 */
1866 	cur = handle->zone_dh_cur;
1867 	firstmatch = NULL;
1868 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
1869 		if (xmlStrcmp(cur->name, DTD_ELEM_FS))
1870 			continue;
1871 		if (strlen(tabptr->zone_fs_dir) > 0) {
1872 			if ((fetchprop(cur, DTD_ATTR_DIR, dirname,
1873 			    sizeof (dirname)) == Z_OK) &&
1874 			    (strcmp(tabptr->zone_fs_dir, dirname) == 0)) {
1875 				if (firstmatch == NULL)
1876 					firstmatch = cur;
1877 				else
1878 					return (Z_INSUFFICIENT_SPEC);
1879 			}
1880 		}
1881 		if (strlen(tabptr->zone_fs_special) > 0) {
1882 			if ((fetchprop(cur, DTD_ATTR_SPECIAL, special,
1883 			    sizeof (special)) == Z_OK)) {
1884 				if (strcmp(tabptr->zone_fs_special,
1885 				    special) == 0) {
1886 					if (firstmatch == NULL)
1887 						firstmatch = cur;
1888 					else if (firstmatch != cur)
1889 						return (Z_INSUFFICIENT_SPEC);
1890 				} else {
1891 					/*
1892 					 * If another property matched but this
1893 					 * one doesn't then reset firstmatch.
1894 					 */
1895 					if (firstmatch == cur)
1896 						firstmatch = NULL;
1897 				}
1898 			}
1899 		}
1900 		if (strlen(tabptr->zone_fs_raw) > 0) {
1901 			if ((fetchprop(cur, DTD_ATTR_RAW, raw,
1902 			    sizeof (raw)) == Z_OK)) {
1903 				if (strcmp(tabptr->zone_fs_raw, raw) == 0) {
1904 					if (firstmatch == NULL)
1905 						firstmatch = cur;
1906 					else if (firstmatch != cur)
1907 						return (Z_INSUFFICIENT_SPEC);
1908 				} else {
1909 					/*
1910 					 * If another property matched but this
1911 					 * one doesn't then reset firstmatch.
1912 					 */
1913 					if (firstmatch == cur)
1914 						firstmatch = NULL;
1915 				}
1916 			}
1917 		}
1918 		if (strlen(tabptr->zone_fs_type) > 0) {
1919 			if ((fetchprop(cur, DTD_ATTR_TYPE, type,
1920 			    sizeof (type)) == Z_OK)) {
1921 				if (strcmp(tabptr->zone_fs_type, type) == 0) {
1922 					if (firstmatch == NULL)
1923 						firstmatch = cur;
1924 					else if (firstmatch != cur)
1925 						return (Z_INSUFFICIENT_SPEC);
1926 				} else {
1927 					/*
1928 					 * If another property matched but this
1929 					 * one doesn't then reset firstmatch.
1930 					 */
1931 					if (firstmatch == cur)
1932 						firstmatch = NULL;
1933 				}
1934 			}
1935 		}
1936 	}
1937 
1938 	if (firstmatch == NULL)
1939 		return (Z_NO_RESOURCE_ID);
1940 
1941 	cur = firstmatch;
1942 
1943 	if ((err = fetchprop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir,
1944 	    sizeof (tabptr->zone_fs_dir))) != Z_OK)
1945 		return (err);
1946 
1947 	if ((err = fetchprop(cur, DTD_ATTR_SPECIAL, tabptr->zone_fs_special,
1948 	    sizeof (tabptr->zone_fs_special))) != Z_OK)
1949 		return (err);
1950 
1951 	if ((err = fetchprop(cur, DTD_ATTR_RAW, tabptr->zone_fs_raw,
1952 	    sizeof (tabptr->zone_fs_raw))) != Z_OK)
1953 		return (err);
1954 
1955 	if ((err = fetchprop(cur, DTD_ATTR_TYPE, tabptr->zone_fs_type,
1956 	    sizeof (tabptr->zone_fs_type))) != Z_OK)
1957 		return (err);
1958 
1959 	/* options are optional */
1960 	tabptr->zone_fs_options = NULL;
1961 	for (options = cur->xmlChildrenNode; options != NULL;
1962 	    options = options->next) {
1963 		if ((fetchprop(options, DTD_ATTR_NAME, options_str,
1964 		    sizeof (options_str)) != Z_OK))
1965 			break;
1966 		if (zonecfg_add_fs_option(tabptr, options_str) != Z_OK)
1967 			break;
1968 	}
1969 	return (Z_OK);
1970 }
1971 
1972 int
1973 zonecfg_lookup_ipd(zone_dochandle_t handle, struct zone_fstab *tabptr)
1974 {
1975 	xmlNodePtr cur, match;
1976 	int err;
1977 	char dirname[MAXPATHLEN];
1978 
1979 	if (tabptr == NULL)
1980 		return (Z_INVAL);
1981 
1982 	if ((err = operation_prep(handle)) != Z_OK)
1983 		return (err);
1984 
1985 	/*
1986 	 * General algorithm:
1987 	 * Walk the list of children looking for matches on any properties
1988 	 * specified in the fstab parameter.  If more than one resource
1989 	 * matches, we return Z_INSUFFICIENT_SPEC; if none match, we return
1990 	 * Z_NO_RESOURCE_ID.
1991 	 */
1992 	cur = handle->zone_dh_cur;
1993 	match = NULL;
1994 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
1995 		if (xmlStrcmp(cur->name, DTD_ELEM_IPD))
1996 			continue;
1997 		if (strlen(tabptr->zone_fs_dir) > 0) {
1998 			if ((fetchprop(cur, DTD_ATTR_DIR, dirname,
1999 			    sizeof (dirname)) == Z_OK) &&
2000 			    (strcmp(tabptr->zone_fs_dir, dirname) == 0)) {
2001 				if (match == NULL)
2002 					match = cur;
2003 				else
2004 					return (Z_INSUFFICIENT_SPEC);
2005 			}
2006 		}
2007 	}
2008 
2009 	if (match == NULL)
2010 		return (Z_NO_RESOURCE_ID);
2011 
2012 	cur = match;
2013 
2014 	if ((err = fetchprop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir,
2015 	    sizeof (tabptr->zone_fs_dir))) != Z_OK)
2016 		return (err);
2017 
2018 	return (Z_OK);
2019 }
2020 
2021 /*
2022  * Compare two IP addresses in string form.  Allow for the possibility that
2023  * one might have "/<prefix-length>" at the end: allow a match on just the
2024  * IP address (or host name) part.
2025  */
2026 
2027 boolean_t
2028 zonecfg_same_net_address(char *a1, char *a2)
2029 {
2030 	char *slashp, *slashp1, *slashp2;
2031 	int result;
2032 
2033 	if (strcmp(a1, a2) == 0)
2034 		return (B_TRUE);
2035 
2036 	/*
2037 	 * If neither has a slash or both do, they need to match to be
2038 	 * considered the same, but they did not match above, so fail.
2039 	 */
2040 	slashp1 = strchr(a1, '/');
2041 	slashp2 = strchr(a2, '/');
2042 	if ((slashp1 == NULL && slashp2 == NULL) ||
2043 	    (slashp1 != NULL && slashp2 != NULL))
2044 		return (B_FALSE);
2045 
2046 	/*
2047 	 * Only one had a slash: pick that one, zero out the slash, compare
2048 	 * the "address only" strings, restore the slash, and return the
2049 	 * result of the comparison.
2050 	 */
2051 	slashp = (slashp1 == NULL) ? slashp2 : slashp1;
2052 	*slashp = '\0';
2053 	result = strcmp(a1, a2);
2054 	*slashp = '/';
2055 	return ((result == 0));
2056 }
2057 
2058 int
2059 zonecfg_valid_net_address(char *address, struct lifreq *lifr)
2060 {
2061 	struct sockaddr_in *sin4;
2062 	struct sockaddr_in6 *sin6;
2063 	struct addrinfo hints, *result;
2064 	char *slashp = strchr(address, '/');
2065 
2066 	bzero(lifr, sizeof (struct lifreq));
2067 	sin4 = (struct sockaddr_in *)&lifr->lifr_addr;
2068 	sin6 = (struct sockaddr_in6 *)&lifr->lifr_addr;
2069 	if (slashp != NULL)
2070 		*slashp = '\0';
2071 	if (inet_pton(AF_INET, address, &sin4->sin_addr) == 1) {
2072 		sin4->sin_family = AF_INET;
2073 	} else if (inet_pton(AF_INET6, address, &sin6->sin6_addr) == 1) {
2074 		if (slashp == NULL)
2075 			return (Z_IPV6_ADDR_PREFIX_LEN);
2076 		sin6->sin6_family = AF_INET6;
2077 	} else {
2078 		/* "address" may be a host name */
2079 		(void) memset(&hints, 0, sizeof (hints));
2080 		hints.ai_family = PF_INET;
2081 		if (getaddrinfo(address, NULL, &hints, &result) != 0)
2082 			return (Z_BOGUS_ADDRESS);
2083 		sin4->sin_family = result->ai_family;
2084 
2085 		(void) memcpy(&sin4->sin_addr,
2086 		    /* LINTED E_BAD_PTR_CAST_ALIGN */
2087 		    &((struct sockaddr_in *)result->ai_addr)->sin_addr,
2088 		    sizeof (struct in_addr));
2089 
2090 		freeaddrinfo(result);
2091 	}
2092 	return (Z_OK);
2093 }
2094 
2095 boolean_t
2096 zonecfg_ifname_exists(sa_family_t af, char *ifname)
2097 {
2098 	struct lifreq lifr;
2099 	int so;
2100 	int save_errno;
2101 
2102 	(void) memset(&lifr, 0, sizeof (lifr));
2103 	(void) strlcpy(lifr.lifr_name, ifname, sizeof (lifr.lifr_name));
2104 	lifr.lifr_addr.ss_family = af;
2105 	if ((so = socket(af, SOCK_DGRAM, 0)) < 0) {
2106 		/* Odd - can't tell if the ifname exists */
2107 		return (B_FALSE);
2108 	}
2109 	if (ioctl(so, SIOCGLIFFLAGS, (caddr_t)&lifr) < 0) {
2110 		save_errno = errno;
2111 		(void) close(so);
2112 		errno = save_errno;
2113 		return (B_FALSE);
2114 	}
2115 	(void) close(so);
2116 	return (B_TRUE);
2117 }
2118 
2119 /*
2120  * Determines if the physical interface and IP address specified by 'tabptr'
2121  * are in the zone document to which 'handle' refers.  'tabptr' must have an
2122  * interface or an address or both.  If it contains both, then Z_OK is
2123  * returned iff there is exactly one match.  If it contains an interface
2124  * or an address, but not both, then Z_OK is returned iff there is exactly
2125  * one entry with that interface or address.  If there are multiple entries
2126  * matching the query, then Z_INSUFFICIENT_SPEC is returned.  If there
2127  * are no matches, then Z_NO_RESOURCE_ID is returned.
2128  *
2129  * Errors might also be returned if the entry that exactly matches the
2130  * query lacks critical network resource information.
2131  *
2132  * If there is a single exact match, then the matching entry's physical
2133  * interface, IP address, and router information is stored in 'tabptr'.
2134  */
2135 int
2136 zonecfg_lookup_nwif(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
2137 {
2138 	xmlNodePtr cur;
2139 	xmlNodePtr firstmatch;
2140 	boolean_t physfound;
2141 	int err;
2142 	char address[INET6_ADDRSTRLEN];
2143 	char physical[LIFNAMSIZ];
2144 	size_t addrspec;		/* nonzero if tabptr has IP addr */
2145 	size_t physspec;		/* nonzero if tabptr has interface */
2146 
2147 	if (tabptr == NULL)
2148 		return (Z_INVAL);
2149 
2150 	/*
2151 	 * zone_nwif_address and zone_nwif_physical are arrays, so no NULL
2152 	 * checks are necessary.
2153 	 */
2154 	addrspec = strlen(tabptr->zone_nwif_address);
2155 	physspec = strlen(tabptr->zone_nwif_physical);
2156 	assert(addrspec > 0 || physspec > 0);
2157 
2158 	if ((err = operation_prep(handle)) != Z_OK)
2159 		return (err);
2160 
2161 	firstmatch = NULL;
2162 	cur = handle->zone_dh_cur;
2163 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2164 		/* Skip non-net elements */
2165 		if (xmlStrcmp(cur->name, DTD_ELEM_NET))
2166 			continue;
2167 
2168 		/*
2169 		 * If an interface is specified, then first check if the current
2170 		 * element's interface matches the query's interface.
2171 		 */
2172 		if (physspec > 0) {
2173 			physfound = B_FALSE;
2174 			if ((fetchprop(cur, DTD_ATTR_PHYSICAL, physical,
2175 			    sizeof (physical)) == Z_OK) &&
2176 			    (strcmp(tabptr->zone_nwif_physical,
2177 			    physical) == 0)) {
2178 				if (addrspec == 0) {
2179 					if (firstmatch == NULL)
2180 						firstmatch = cur;
2181 					else
2182 						return (Z_INSUFFICIENT_SPEC);
2183 				} else {
2184 					/*
2185 					 * We're also matching based on IP
2186 					 * address, so we can't say that the
2187 					 * current element matches the query
2188 					 * yet.  Indicate that the interfaces
2189 					 * match.
2190 					 */
2191 					physfound = B_TRUE;
2192 				}
2193 			}
2194 		}
2195 		if (addrspec > 0) {
2196 			if ((fetchprop(cur, DTD_ATTR_ADDRESS, address,
2197 			    sizeof (address)) == Z_OK) &&
2198 			    (zonecfg_same_net_address(
2199 			    tabptr->zone_nwif_address, address))) {
2200 				if (physspec == 0) {
2201 					/* We're only matching IP addresses. */
2202 					if (firstmatch == NULL)
2203 						firstmatch = cur;
2204 					else
2205 						return (Z_INSUFFICIENT_SPEC);
2206 				} else if (physfound) {
2207 					/*
2208 					 * Both the interfaces and the addresses
2209 					 * match.
2210 					 */
2211 					if (firstmatch == NULL)
2212 						firstmatch = cur;
2213 					else
2214 						return (Z_INSUFFICIENT_SPEC);
2215 				}
2216 			}
2217 		}
2218 	}
2219 	if (firstmatch == NULL)
2220 		return (Z_NO_RESOURCE_ID);
2221 
2222 	cur = firstmatch;
2223 
2224 	if ((err = fetchprop(cur, DTD_ATTR_PHYSICAL, tabptr->zone_nwif_physical,
2225 	    sizeof (tabptr->zone_nwif_physical))) != Z_OK)
2226 		return (err);
2227 
2228 	if ((err = fetchprop(cur, DTD_ATTR_ADDRESS, tabptr->zone_nwif_address,
2229 	    sizeof (tabptr->zone_nwif_address))) != Z_OK)
2230 		return (err);
2231 
2232 	if ((err = fetchprop(cur, DTD_ATTR_DEFROUTER,
2233 	    tabptr->zone_nwif_defrouter,
2234 	    sizeof (tabptr->zone_nwif_defrouter))) != Z_OK)
2235 		return (err);
2236 
2237 	return (Z_OK);
2238 }
2239 
2240 static int
2241 zonecfg_add_nwif_core(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
2242 {
2243 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
2244 	int err;
2245 
2246 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_NET, NULL);
2247 	if ((err = newprop(newnode, DTD_ATTR_ADDRESS,
2248 	    tabptr->zone_nwif_address)) != Z_OK)
2249 		return (err);
2250 	if ((err = newprop(newnode, DTD_ATTR_PHYSICAL,
2251 	    tabptr->zone_nwif_physical)) != Z_OK)
2252 		return (err);
2253 	/*
2254 	 * Do not add this property when it is not set, for backwards
2255 	 * compatibility and because it is optional.
2256 	 */
2257 	if ((strlen(tabptr->zone_nwif_defrouter) > 0) &&
2258 	    ((err = newprop(newnode, DTD_ATTR_DEFROUTER,
2259 	    tabptr->zone_nwif_defrouter)) != Z_OK))
2260 		return (err);
2261 	return (Z_OK);
2262 }
2263 
2264 int
2265 zonecfg_add_nwif(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
2266 {
2267 	int err;
2268 
2269 	if (tabptr == NULL)
2270 		return (Z_INVAL);
2271 
2272 	if ((err = operation_prep(handle)) != Z_OK)
2273 		return (err);
2274 
2275 	if ((err = zonecfg_add_nwif_core(handle, tabptr)) != Z_OK)
2276 		return (err);
2277 
2278 	return (Z_OK);
2279 }
2280 
2281 static int
2282 zonecfg_delete_nwif_core(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
2283 {
2284 	xmlNodePtr cur = handle->zone_dh_cur;
2285 	boolean_t addr_match, phys_match;
2286 
2287 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2288 		if (xmlStrcmp(cur->name, DTD_ELEM_NET))
2289 			continue;
2290 
2291 		addr_match = match_prop(cur, DTD_ATTR_ADDRESS,
2292 		    tabptr->zone_nwif_address);
2293 		phys_match = match_prop(cur, DTD_ATTR_PHYSICAL,
2294 		    tabptr->zone_nwif_physical);
2295 
2296 		if (addr_match && phys_match) {
2297 			xmlUnlinkNode(cur);
2298 			xmlFreeNode(cur);
2299 			return (Z_OK);
2300 		}
2301 	}
2302 	return (Z_NO_RESOURCE_ID);
2303 }
2304 
2305 int
2306 zonecfg_delete_nwif(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
2307 {
2308 	int err;
2309 
2310 	if (tabptr == NULL)
2311 		return (Z_INVAL);
2312 
2313 	if ((err = operation_prep(handle)) != Z_OK)
2314 		return (err);
2315 
2316 	if ((err = zonecfg_delete_nwif_core(handle, tabptr)) != Z_OK)
2317 		return (err);
2318 
2319 	return (Z_OK);
2320 }
2321 
2322 int
2323 zonecfg_modify_nwif(
2324 	zone_dochandle_t handle,
2325 	struct zone_nwiftab *oldtabptr,
2326 	struct zone_nwiftab *newtabptr)
2327 {
2328 	int err;
2329 
2330 	if (oldtabptr == NULL || newtabptr == NULL)
2331 		return (Z_INVAL);
2332 
2333 	if ((err = operation_prep(handle)) != Z_OK)
2334 		return (err);
2335 
2336 	if ((err = zonecfg_delete_nwif_core(handle, oldtabptr)) != Z_OK)
2337 		return (err);
2338 
2339 	if ((err = zonecfg_add_nwif_core(handle, newtabptr)) != Z_OK)
2340 		return (err);
2341 
2342 	return (Z_OK);
2343 }
2344 
2345 int
2346 zonecfg_lookup_dev(zone_dochandle_t handle, struct zone_devtab *tabptr)
2347 {
2348 	xmlNodePtr cur, firstmatch;
2349 	int err;
2350 	char match[MAXPATHLEN];
2351 
2352 	if (tabptr == NULL)
2353 		return (Z_INVAL);
2354 
2355 	if ((err = operation_prep(handle)) != Z_OK)
2356 		return (err);
2357 
2358 	cur = handle->zone_dh_cur;
2359 	firstmatch = NULL;
2360 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2361 		if (xmlStrcmp(cur->name, DTD_ELEM_DEVICE))
2362 			continue;
2363 		if (strlen(tabptr->zone_dev_match) == 0)
2364 			continue;
2365 
2366 		if ((fetchprop(cur, DTD_ATTR_MATCH, match,
2367 		    sizeof (match)) == Z_OK)) {
2368 			if (strcmp(tabptr->zone_dev_match,
2369 			    match) == 0) {
2370 				if (firstmatch == NULL)
2371 					firstmatch = cur;
2372 				else if (firstmatch != cur)
2373 					return (Z_INSUFFICIENT_SPEC);
2374 			} else {
2375 				/*
2376 				 * If another property matched but this
2377 				 * one doesn't then reset firstmatch.
2378 				 */
2379 				if (firstmatch == cur)
2380 					firstmatch = NULL;
2381 			}
2382 		}
2383 	}
2384 	if (firstmatch == NULL)
2385 		return (Z_NO_RESOURCE_ID);
2386 
2387 	cur = firstmatch;
2388 
2389 	if ((err = fetchprop(cur, DTD_ATTR_MATCH, tabptr->zone_dev_match,
2390 	    sizeof (tabptr->zone_dev_match))) != Z_OK)
2391 		return (err);
2392 
2393 	return (Z_OK);
2394 }
2395 
2396 static int
2397 zonecfg_add_dev_core(zone_dochandle_t handle, struct zone_devtab *tabptr)
2398 {
2399 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
2400 	int err;
2401 
2402 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_DEVICE, NULL);
2403 
2404 	if ((err = newprop(newnode, DTD_ATTR_MATCH,
2405 	    tabptr->zone_dev_match)) != Z_OK)
2406 		return (err);
2407 
2408 	return (Z_OK);
2409 }
2410 
2411 int
2412 zonecfg_add_dev(zone_dochandle_t handle, struct zone_devtab *tabptr)
2413 {
2414 	int err;
2415 
2416 	if (tabptr == NULL)
2417 		return (Z_INVAL);
2418 
2419 	if ((err = operation_prep(handle)) != Z_OK)
2420 		return (err);
2421 
2422 	if ((err = zonecfg_add_dev_core(handle, tabptr)) != Z_OK)
2423 		return (err);
2424 
2425 	return (Z_OK);
2426 }
2427 
2428 static int
2429 zonecfg_delete_dev_core(zone_dochandle_t handle, struct zone_devtab *tabptr)
2430 {
2431 	xmlNodePtr cur = handle->zone_dh_cur;
2432 	int match_match;
2433 
2434 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2435 		if (xmlStrcmp(cur->name, DTD_ELEM_DEVICE))
2436 			continue;
2437 
2438 		match_match = match_prop(cur, DTD_ATTR_MATCH,
2439 		    tabptr->zone_dev_match);
2440 
2441 		if (match_match) {
2442 			xmlUnlinkNode(cur);
2443 			xmlFreeNode(cur);
2444 			return (Z_OK);
2445 		}
2446 	}
2447 	return (Z_NO_RESOURCE_ID);
2448 }
2449 
2450 int
2451 zonecfg_delete_dev(zone_dochandle_t handle, struct zone_devtab *tabptr)
2452 {
2453 	int err;
2454 
2455 	if (tabptr == NULL)
2456 		return (Z_INVAL);
2457 
2458 	if ((err = operation_prep(handle)) != Z_OK)
2459 		return (err);
2460 
2461 	if ((err = zonecfg_delete_dev_core(handle, tabptr)) != Z_OK)
2462 		return (err);
2463 
2464 	return (Z_OK);
2465 }
2466 
2467 int
2468 zonecfg_modify_dev(
2469 	zone_dochandle_t handle,
2470 	struct zone_devtab *oldtabptr,
2471 	struct zone_devtab *newtabptr)
2472 {
2473 	int err;
2474 
2475 	if (oldtabptr == NULL || newtabptr == NULL)
2476 		return (Z_INVAL);
2477 
2478 	if ((err = operation_prep(handle)) != Z_OK)
2479 		return (err);
2480 
2481 	if ((err = zonecfg_delete_dev_core(handle, oldtabptr)) != Z_OK)
2482 		return (err);
2483 
2484 	if ((err = zonecfg_add_dev_core(handle, newtabptr)) != Z_OK)
2485 		return (err);
2486 
2487 	return (Z_OK);
2488 }
2489 
2490 /* Lock to serialize all devwalks */
2491 static pthread_mutex_t zonecfg_devwalk_lock = PTHREAD_MUTEX_INITIALIZER;
2492 /*
2493  * Global variables used to pass data from zonecfg_dev_manifest to the nftw
2494  * call-back (zonecfg_devwalk_cb).  g_devwalk_data is really the void*
2495  * parameter and g_devwalk_cb is really the *cb parameter from
2496  * zonecfg_dev_manifest.
2497  */
2498 typedef struct __g_devwalk_data *g_devwalk_data_t;
2499 static g_devwalk_data_t g_devwalk_data;
2500 static int (*g_devwalk_cb)(const char *, uid_t, gid_t, mode_t, const char *,
2501     void *);
2502 static size_t g_devwalk_skip_prefix;
2503 
2504 /*
2505  * zonecfg_dev_manifest call-back function used during detach to generate the
2506  * dev info in the manifest.
2507  */
2508 static int
2509 get_detach_dev_entry(const char *name, uid_t uid, gid_t gid, mode_t mode,
2510     const char *acl, void *hdl)
2511 {
2512 	zone_dochandle_t handle = (zone_dochandle_t)hdl;
2513 	xmlNodePtr newnode;
2514 	xmlNodePtr cur;
2515 	int err;
2516 	char buf[128];
2517 
2518 	if ((err = operation_prep(handle)) != Z_OK)
2519 		return (err);
2520 
2521 	cur = handle->zone_dh_cur;
2522 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_DEV_PERM, NULL);
2523 	if ((err = newprop(newnode, DTD_ATTR_NAME, (char *)name)) != Z_OK)
2524 		return (err);
2525 	(void) snprintf(buf, sizeof (buf), "%lu", uid);
2526 	if ((err = newprop(newnode, DTD_ATTR_UID, buf)) != Z_OK)
2527 		return (err);
2528 	(void) snprintf(buf, sizeof (buf), "%lu", gid);
2529 	if ((err = newprop(newnode, DTD_ATTR_GID, buf)) != Z_OK)
2530 		return (err);
2531 	(void) snprintf(buf, sizeof (buf), "%o", mode);
2532 	if ((err = newprop(newnode, DTD_ATTR_MODE, buf)) != Z_OK)
2533 		return (err);
2534 	if ((err = newprop(newnode, DTD_ATTR_ACL, (char *)acl)) != Z_OK)
2535 		return (err);
2536 	return (Z_OK);
2537 }
2538 
2539 /*
2540  * This is the nftw call-back function used by zonecfg_dev_manifest.  It is
2541  * responsible for calling the actual call-back.
2542  */
2543 /* ARGSUSED2 */
2544 static int
2545 zonecfg_devwalk_cb(const char *path, const struct stat *st, int f,
2546     struct FTW *ftw)
2547 {
2548 	acl_t *acl;
2549 	char *acl_txt = NULL;
2550 
2551 	/* skip all but character and block devices */
2552 	if (!S_ISBLK(st->st_mode) && !S_ISCHR(st->st_mode))
2553 		return (0);
2554 
2555 	if ((acl_get(path, ACL_NO_TRIVIAL, &acl) == 0) &&
2556 	    acl != NULL) {
2557 		acl_txt = acl_totext(acl, ACL_NORESOLVE);
2558 		acl_free(acl);
2559 	}
2560 
2561 	if (strlen(path) <= g_devwalk_skip_prefix)
2562 		return (0);
2563 
2564 	g_devwalk_cb(path + g_devwalk_skip_prefix, st->st_uid, st->st_gid,
2565 	    st->st_mode & S_IAMB, acl_txt != NULL ? acl_txt : "",
2566 	    g_devwalk_data);
2567 	free(acl_txt);
2568 	return (0);
2569 }
2570 
2571 /*
2572  * Walk the dev tree for the zone specified by hdl and call the
2573  * get_detach_dev_entry call-back function for each entry in the tree.  The
2574  * call-back will be passed the name, uid, gid, mode, acl string and the
2575  * handle input parameter for each dev entry.
2576  *
2577  * Data is passed to get_detach_dev_entry through the global variables
2578  * g_devwalk_data, *g_devwalk_cb, and g_devwalk_skip_prefix.  The
2579  * zonecfg_devwalk_cb function will actually call get_detach_dev_entry.
2580  */
2581 int
2582 zonecfg_dev_manifest(zone_dochandle_t hdl)
2583 {
2584 	char path[MAXPATHLEN];
2585 	int ret;
2586 
2587 	if ((ret = zonecfg_get_zonepath(hdl, path, sizeof (path))) != Z_OK)
2588 		return (ret);
2589 
2590 	if (strlcat(path, "/dev", sizeof (path)) >= sizeof (path))
2591 		return (Z_TOO_BIG);
2592 
2593 	/*
2594 	 * We have to serialize all devwalks in the same process
2595 	 * (which should be fine), since nftw() is so badly designed.
2596 	 */
2597 	(void) pthread_mutex_lock(&zonecfg_devwalk_lock);
2598 
2599 	g_devwalk_skip_prefix = strlen(path) + 1;
2600 	g_devwalk_data = (g_devwalk_data_t)hdl;
2601 	g_devwalk_cb = get_detach_dev_entry;
2602 	(void) nftw(path, zonecfg_devwalk_cb, 0, FTW_PHYS);
2603 
2604 	(void) pthread_mutex_unlock(&zonecfg_devwalk_lock);
2605 	return (Z_OK);
2606 }
2607 
2608 /*
2609  * Update the owner, group, mode and acl on the specified dev (inpath) for
2610  * the zone (hdl).  This function can be used to fix up the dev tree after
2611  * attaching a migrated zone.
2612  */
2613 int
2614 zonecfg_devperms_apply(zone_dochandle_t hdl, const char *inpath, uid_t owner,
2615     gid_t group, mode_t mode, const char *acltxt)
2616 {
2617 	int ret;
2618 	char path[MAXPATHLEN];
2619 	struct stat st;
2620 	acl_t *aclp;
2621 
2622 	if ((ret = zonecfg_get_zonepath(hdl, path, sizeof (path))) != Z_OK)
2623 		return (ret);
2624 
2625 	if (strlcat(path, "/dev/", sizeof (path)) >= sizeof (path))
2626 		return (Z_TOO_BIG);
2627 	if (strlcat(path, inpath, sizeof (path)) >= sizeof (path))
2628 		return (Z_TOO_BIG);
2629 
2630 	if (stat(path, &st) == -1)
2631 		return (Z_INVAL);
2632 
2633 	/* make sure we're only touching device nodes */
2634 	if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode))
2635 		return (Z_INVAL);
2636 
2637 	if (chown(path, owner, group) == -1)
2638 		return (Z_SYSTEM);
2639 
2640 	if (chmod(path, mode) == -1)
2641 		return (Z_SYSTEM);
2642 
2643 	if ((acltxt == NULL) || (strcmp(acltxt, "") == 0))
2644 		return (Z_OK);
2645 
2646 	if (acl_fromtext(acltxt, &aclp) != 0)
2647 		return (Z_SYSTEM);
2648 
2649 	errno = 0;
2650 	if (acl_set(path, aclp) == -1) {
2651 		free(aclp);
2652 		return (Z_SYSTEM);
2653 	}
2654 
2655 	free(aclp);
2656 	return (Z_OK);
2657 }
2658 
2659 /*
2660  * This function finds everything mounted under a zone's rootpath.
2661  * This returns the number of mounts under rootpath, or -1 on error.
2662  * callback is called once per mount found with the first argument
2663  * pointing to the  mount point.
2664  *
2665  * If the callback function returns non-zero zonecfg_find_mounts
2666  * aborts with an error.
2667  */
2668 int
2669 zonecfg_find_mounts(char *rootpath, int (*callback)(const char *, void *),
2670     void *priv) {
2671 	FILE *mnttab;
2672 	struct mnttab m;
2673 	size_t l;
2674 	int zfsl;
2675 	int rv = 0;
2676 	char zfs_path[MAXPATHLEN];
2677 
2678 	assert(rootpath != NULL);
2679 
2680 	if ((zfsl = snprintf(zfs_path, sizeof (zfs_path), "%s/.zfs/", rootpath))
2681 	    >= sizeof (zfs_path))
2682 		return (-1);
2683 
2684 	l = strlen(rootpath);
2685 
2686 	mnttab = fopen("/etc/mnttab", "r");
2687 
2688 	if (mnttab == NULL)
2689 		return (-1);
2690 
2691 	if (ioctl(fileno(mnttab), MNTIOC_SHOWHIDDEN, NULL) < 0)  {
2692 		rv = -1;
2693 		goto out;
2694 	}
2695 
2696 	while (!getmntent(mnttab, &m)) {
2697 		if ((strncmp(rootpath, m.mnt_mountp, l) == 0) &&
2698 		    (m.mnt_mountp[l] == '/') &&
2699 		    (strncmp(zfs_path, m.mnt_mountp, zfsl) != 0)) {
2700 			rv++;
2701 			if (callback == NULL)
2702 				continue;
2703 			if (callback(m.mnt_mountp, priv)) {
2704 				rv = -1;
2705 				goto out;
2706 
2707 			}
2708 		}
2709 	}
2710 
2711 out:
2712 	(void) fclose(mnttab);
2713 	return (rv);
2714 }
2715 
2716 int
2717 zonecfg_lookup_attr(zone_dochandle_t handle, struct zone_attrtab *tabptr)
2718 {
2719 	xmlNodePtr cur, firstmatch;
2720 	int err;
2721 	char name[MAXNAMELEN], type[MAXNAMELEN], value[MAXNAMELEN];
2722 
2723 	if (tabptr == NULL)
2724 		return (Z_INVAL);
2725 
2726 	if ((err = operation_prep(handle)) != Z_OK)
2727 		return (err);
2728 
2729 	cur = handle->zone_dh_cur;
2730 	firstmatch = NULL;
2731 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2732 		if (xmlStrcmp(cur->name, DTD_ELEM_ATTR))
2733 			continue;
2734 		if (strlen(tabptr->zone_attr_name) > 0) {
2735 			if ((fetchprop(cur, DTD_ATTR_NAME, name,
2736 			    sizeof (name)) == Z_OK) &&
2737 			    (strcmp(tabptr->zone_attr_name, name) == 0)) {
2738 				if (firstmatch == NULL)
2739 					firstmatch = cur;
2740 				else
2741 					return (Z_INSUFFICIENT_SPEC);
2742 			}
2743 		}
2744 		if (strlen(tabptr->zone_attr_type) > 0) {
2745 			if ((fetchprop(cur, DTD_ATTR_TYPE, type,
2746 			    sizeof (type)) == Z_OK)) {
2747 				if (strcmp(tabptr->zone_attr_type, type) == 0) {
2748 					if (firstmatch == NULL)
2749 						firstmatch = cur;
2750 					else if (firstmatch != cur)
2751 						return (Z_INSUFFICIENT_SPEC);
2752 				} else {
2753 					/*
2754 					 * If another property matched but this
2755 					 * one doesn't then reset firstmatch.
2756 					 */
2757 					if (firstmatch == cur)
2758 						firstmatch = NULL;
2759 				}
2760 			}
2761 		}
2762 		if (strlen(tabptr->zone_attr_value) > 0) {
2763 			if ((fetchprop(cur, DTD_ATTR_VALUE, value,
2764 			    sizeof (value)) == Z_OK)) {
2765 				if (strcmp(tabptr->zone_attr_value, value) ==
2766 				    0) {
2767 					if (firstmatch == NULL)
2768 						firstmatch = cur;
2769 					else if (firstmatch != cur)
2770 						return (Z_INSUFFICIENT_SPEC);
2771 				} else {
2772 					/*
2773 					 * If another property matched but this
2774 					 * one doesn't then reset firstmatch.
2775 					 */
2776 					if (firstmatch == cur)
2777 						firstmatch = NULL;
2778 				}
2779 			}
2780 		}
2781 	}
2782 	if (firstmatch == NULL)
2783 		return (Z_NO_RESOURCE_ID);
2784 
2785 	cur = firstmatch;
2786 
2787 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_attr_name,
2788 	    sizeof (tabptr->zone_attr_name))) != Z_OK)
2789 		return (err);
2790 
2791 	if ((err = fetchprop(cur, DTD_ATTR_TYPE, tabptr->zone_attr_type,
2792 	    sizeof (tabptr->zone_attr_type))) != Z_OK)
2793 		return (err);
2794 
2795 	if ((err = fetchprop(cur, DTD_ATTR_VALUE, tabptr->zone_attr_value,
2796 	    sizeof (tabptr->zone_attr_value))) != Z_OK)
2797 		return (err);
2798 
2799 	return (Z_OK);
2800 }
2801 
2802 static int
2803 zonecfg_add_attr_core(zone_dochandle_t handle, struct zone_attrtab *tabptr)
2804 {
2805 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
2806 	int err;
2807 
2808 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_ATTR, NULL);
2809 	err = newprop(newnode, DTD_ATTR_NAME, tabptr->zone_attr_name);
2810 	if (err != Z_OK)
2811 		return (err);
2812 	err = newprop(newnode, DTD_ATTR_TYPE, tabptr->zone_attr_type);
2813 	if (err != Z_OK)
2814 		return (err);
2815 	err = newprop(newnode, DTD_ATTR_VALUE, tabptr->zone_attr_value);
2816 	if (err != Z_OK)
2817 		return (err);
2818 	return (Z_OK);
2819 }
2820 
2821 int
2822 zonecfg_add_attr(zone_dochandle_t handle, struct zone_attrtab *tabptr)
2823 {
2824 	int err;
2825 
2826 	if (tabptr == NULL)
2827 		return (Z_INVAL);
2828 
2829 	if ((err = operation_prep(handle)) != Z_OK)
2830 		return (err);
2831 
2832 	if ((err = zonecfg_add_attr_core(handle, tabptr)) != Z_OK)
2833 		return (err);
2834 
2835 	return (Z_OK);
2836 }
2837 
2838 static int
2839 zonecfg_delete_attr_core(zone_dochandle_t handle, struct zone_attrtab *tabptr)
2840 {
2841 	xmlNodePtr cur = handle->zone_dh_cur;
2842 	int name_match, type_match, value_match;
2843 
2844 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
2845 		if (xmlStrcmp(cur->name, DTD_ELEM_ATTR))
2846 			continue;
2847 
2848 		name_match = match_prop(cur, DTD_ATTR_NAME,
2849 		    tabptr->zone_attr_name);
2850 		type_match = match_prop(cur, DTD_ATTR_TYPE,
2851 		    tabptr->zone_attr_type);
2852 		value_match = match_prop(cur, DTD_ATTR_VALUE,
2853 		    tabptr->zone_attr_value);
2854 
2855 		if (name_match && type_match && value_match) {
2856 			xmlUnlinkNode(cur);
2857 			xmlFreeNode(cur);
2858 			return (Z_OK);
2859 		}
2860 	}
2861 	return (Z_NO_RESOURCE_ID);
2862 }
2863 
2864 int
2865 zonecfg_delete_attr(zone_dochandle_t handle, struct zone_attrtab *tabptr)
2866 {
2867 	int err;
2868 
2869 	if (tabptr == NULL)
2870 		return (Z_INVAL);
2871 
2872 	if ((err = operation_prep(handle)) != Z_OK)
2873 		return (err);
2874 
2875 	if ((err = zonecfg_delete_attr_core(handle, tabptr)) != Z_OK)
2876 		return (err);
2877 
2878 	return (Z_OK);
2879 }
2880 
2881 int
2882 zonecfg_modify_attr(
2883 	zone_dochandle_t handle,
2884 	struct zone_attrtab *oldtabptr,
2885 	struct zone_attrtab *newtabptr)
2886 {
2887 	int err;
2888 
2889 	if (oldtabptr == NULL || newtabptr == NULL)
2890 		return (Z_INVAL);
2891 
2892 	if ((err = operation_prep(handle)) != Z_OK)
2893 		return (err);
2894 
2895 	if ((err = zonecfg_delete_attr_core(handle, oldtabptr)) != Z_OK)
2896 		return (err);
2897 
2898 	if ((err = zonecfg_add_attr_core(handle, newtabptr)) != Z_OK)
2899 		return (err);
2900 
2901 	return (Z_OK);
2902 }
2903 
2904 int
2905 zonecfg_get_attr_boolean(const struct zone_attrtab *attr, boolean_t *value)
2906 {
2907 	if (attr == NULL)
2908 		return (Z_INVAL);
2909 
2910 	if (strcmp(attr->zone_attr_type, DTD_ENTITY_BOOLEAN) != 0)
2911 		return (Z_INVAL);
2912 
2913 	if (strcmp(attr->zone_attr_value, DTD_ENTITY_TRUE) == 0) {
2914 		*value = B_TRUE;
2915 		return (Z_OK);
2916 	}
2917 	if (strcmp(attr->zone_attr_value, DTD_ENTITY_FALSE) == 0) {
2918 		*value = B_FALSE;
2919 		return (Z_OK);
2920 	}
2921 	return (Z_INVAL);
2922 }
2923 
2924 int
2925 zonecfg_get_attr_int(const struct zone_attrtab *attr, int64_t *value)
2926 {
2927 	long long result;
2928 	char *endptr;
2929 
2930 	if (attr == NULL)
2931 		return (Z_INVAL);
2932 
2933 	if (strcmp(attr->zone_attr_type, DTD_ENTITY_INT) != 0)
2934 		return (Z_INVAL);
2935 
2936 	errno = 0;
2937 	result = strtoll(attr->zone_attr_value, &endptr, 10);
2938 	if (errno != 0 || *endptr != '\0')
2939 		return (Z_INVAL);
2940 	*value = result;
2941 	return (Z_OK);
2942 }
2943 
2944 int
2945 zonecfg_get_attr_string(const struct zone_attrtab *attr, char *value,
2946     size_t val_sz)
2947 {
2948 	if (attr == NULL)
2949 		return (Z_INVAL);
2950 
2951 	if (strcmp(attr->zone_attr_type, DTD_ENTITY_STRING) != 0)
2952 		return (Z_INVAL);
2953 
2954 	if (strlcpy(value, attr->zone_attr_value, val_sz) >= val_sz)
2955 		return (Z_TOO_BIG);
2956 	return (Z_OK);
2957 }
2958 
2959 int
2960 zonecfg_get_attr_uint(const struct zone_attrtab *attr, uint64_t *value)
2961 {
2962 	unsigned long long result;
2963 	long long neg_result;
2964 	char *endptr;
2965 
2966 	if (attr == NULL)
2967 		return (Z_INVAL);
2968 
2969 	if (strcmp(attr->zone_attr_type, DTD_ENTITY_UINT) != 0)
2970 		return (Z_INVAL);
2971 
2972 	errno = 0;
2973 	result = strtoull(attr->zone_attr_value, &endptr, 10);
2974 	if (errno != 0 || *endptr != '\0')
2975 		return (Z_INVAL);
2976 	errno = 0;
2977 	neg_result = strtoll(attr->zone_attr_value, &endptr, 10);
2978 	/*
2979 	 * Incredibly, strtoull("<negative number>", ...) will not fail but
2980 	 * return whatever (negative) number cast as a u_longlong_t, so we
2981 	 * need to look for this here.
2982 	 */
2983 	if (errno == 0 && neg_result < 0)
2984 		return (Z_INVAL);
2985 	*value = result;
2986 	return (Z_OK);
2987 }
2988 
2989 int
2990 zonecfg_lookup_rctl(zone_dochandle_t handle, struct zone_rctltab *tabptr)
2991 {
2992 	xmlNodePtr cur, val;
2993 	char savedname[MAXNAMELEN];
2994 	struct zone_rctlvaltab *valptr;
2995 	int err;
2996 
2997 	if (tabptr->zone_rctl_name == NULL ||
2998 	    strlen(tabptr->zone_rctl_name) == 0)
2999 		return (Z_INVAL);
3000 
3001 	if ((err = operation_prep(handle)) != Z_OK)
3002 		return (err);
3003 
3004 	cur = handle->zone_dh_cur;
3005 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
3006 		if (xmlStrcmp(cur->name, DTD_ELEM_RCTL))
3007 			continue;
3008 		if ((fetchprop(cur, DTD_ATTR_NAME, savedname,
3009 		    sizeof (savedname)) == Z_OK) &&
3010 		    (strcmp(savedname, tabptr->zone_rctl_name) == 0)) {
3011 			tabptr->zone_rctl_valptr = NULL;
3012 			for (val = cur->xmlChildrenNode; val != NULL;
3013 			    val = val->next) {
3014 				valptr = (struct zone_rctlvaltab *)malloc(
3015 				    sizeof (struct zone_rctlvaltab));
3016 				if (valptr == NULL)
3017 					return (Z_NOMEM);
3018 				if ((fetchprop(val, DTD_ATTR_PRIV,
3019 				    valptr->zone_rctlval_priv,
3020 				    sizeof (valptr->zone_rctlval_priv)) !=
3021 				    Z_OK))
3022 					break;
3023 				if ((fetchprop(val, DTD_ATTR_LIMIT,
3024 				    valptr->zone_rctlval_limit,
3025 				    sizeof (valptr->zone_rctlval_limit)) !=
3026 				    Z_OK))
3027 					break;
3028 				if ((fetchprop(val, DTD_ATTR_ACTION,
3029 				    valptr->zone_rctlval_action,
3030 				    sizeof (valptr->zone_rctlval_action)) !=
3031 				    Z_OK))
3032 					break;
3033 				if (zonecfg_add_rctl_value(tabptr, valptr) !=
3034 				    Z_OK)
3035 					break;
3036 			}
3037 			return (Z_OK);
3038 		}
3039 	}
3040 	return (Z_NO_RESOURCE_ID);
3041 }
3042 
3043 static int
3044 zonecfg_add_rctl_core(zone_dochandle_t handle, struct zone_rctltab *tabptr)
3045 {
3046 	xmlNodePtr newnode, cur = handle->zone_dh_cur, valnode;
3047 	struct zone_rctlvaltab *valptr;
3048 	int err;
3049 
3050 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_RCTL, NULL);
3051 	err = newprop(newnode, DTD_ATTR_NAME, tabptr->zone_rctl_name);
3052 	if (err != Z_OK)
3053 		return (err);
3054 	for (valptr = tabptr->zone_rctl_valptr; valptr != NULL;
3055 	    valptr = valptr->zone_rctlval_next) {
3056 		valnode = xmlNewTextChild(newnode, NULL,
3057 		    DTD_ELEM_RCTLVALUE, NULL);
3058 		err = newprop(valnode, DTD_ATTR_PRIV,
3059 		    valptr->zone_rctlval_priv);
3060 		if (err != Z_OK)
3061 			return (err);
3062 		err = newprop(valnode, DTD_ATTR_LIMIT,
3063 		    valptr->zone_rctlval_limit);
3064 		if (err != Z_OK)
3065 			return (err);
3066 		err = newprop(valnode, DTD_ATTR_ACTION,
3067 		    valptr->zone_rctlval_action);
3068 		if (err != Z_OK)
3069 			return (err);
3070 	}
3071 	return (Z_OK);
3072 }
3073 
3074 int
3075 zonecfg_add_rctl(zone_dochandle_t handle, struct zone_rctltab *tabptr)
3076 {
3077 	int err;
3078 
3079 	if (tabptr == NULL || tabptr->zone_rctl_name == NULL)
3080 		return (Z_INVAL);
3081 
3082 	if ((err = operation_prep(handle)) != Z_OK)
3083 		return (err);
3084 
3085 	if ((err = zonecfg_add_rctl_core(handle, tabptr)) != Z_OK)
3086 		return (err);
3087 
3088 	return (Z_OK);
3089 }
3090 
3091 static int
3092 zonecfg_delete_rctl_core(zone_dochandle_t handle, struct zone_rctltab *tabptr)
3093 {
3094 	xmlNodePtr cur = handle->zone_dh_cur;
3095 	xmlChar *savedname;
3096 	int name_result;
3097 
3098 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
3099 		if (xmlStrcmp(cur->name, DTD_ELEM_RCTL))
3100 			continue;
3101 
3102 		savedname = xmlGetProp(cur, DTD_ATTR_NAME);
3103 		if (savedname == NULL)	/* shouldn't happen */
3104 			continue;
3105 		name_result = xmlStrcmp(savedname,
3106 		    (const xmlChar *) tabptr->zone_rctl_name);
3107 		xmlFree(savedname);
3108 
3109 		if (name_result == 0) {
3110 			xmlUnlinkNode(cur);
3111 			xmlFreeNode(cur);
3112 			return (Z_OK);
3113 		}
3114 	}
3115 	return (Z_NO_RESOURCE_ID);
3116 }
3117 
3118 int
3119 zonecfg_delete_rctl(zone_dochandle_t handle, struct zone_rctltab *tabptr)
3120 {
3121 	int err;
3122 
3123 	if (tabptr == NULL || tabptr->zone_rctl_name == NULL)
3124 		return (Z_INVAL);
3125 
3126 	if ((err = operation_prep(handle)) != Z_OK)
3127 		return (err);
3128 
3129 	if ((err = zonecfg_delete_rctl_core(handle, tabptr)) != Z_OK)
3130 		return (err);
3131 
3132 	return (Z_OK);
3133 }
3134 
3135 int
3136 zonecfg_modify_rctl(
3137 	zone_dochandle_t handle,
3138 	struct zone_rctltab *oldtabptr,
3139 	struct zone_rctltab *newtabptr)
3140 {
3141 	int err;
3142 
3143 	if (oldtabptr == NULL || oldtabptr->zone_rctl_name == NULL ||
3144 	    newtabptr == NULL || newtabptr->zone_rctl_name == NULL)
3145 		return (Z_INVAL);
3146 
3147 	if ((err = operation_prep(handle)) != Z_OK)
3148 		return (err);
3149 
3150 	if ((err = zonecfg_delete_rctl_core(handle, oldtabptr)) != Z_OK)
3151 		return (err);
3152 
3153 	if ((err = zonecfg_add_rctl_core(handle, newtabptr)) != Z_OK)
3154 		return (err);
3155 
3156 	return (Z_OK);
3157 }
3158 
3159 int
3160 zonecfg_add_rctl_value(
3161 	struct zone_rctltab *tabptr,
3162 	struct zone_rctlvaltab *valtabptr)
3163 {
3164 	struct zone_rctlvaltab *last, *old, *new;
3165 	rctlblk_t *rctlblk = alloca(rctlblk_size());
3166 
3167 	last = tabptr->zone_rctl_valptr;
3168 	for (old = last; old != NULL; old = old->zone_rctlval_next)
3169 		last = old;	/* walk to the end of the list */
3170 	new = valtabptr;	/* alloc'd by caller */
3171 	new->zone_rctlval_next = NULL;
3172 	if (zonecfg_construct_rctlblk(valtabptr, rctlblk) != Z_OK)
3173 		return (Z_INVAL);
3174 	if (!zonecfg_valid_rctlblk(rctlblk))
3175 		return (Z_INVAL);
3176 	if (last == NULL)
3177 		tabptr->zone_rctl_valptr = new;
3178 	else
3179 		last->zone_rctlval_next = new;
3180 	return (Z_OK);
3181 }
3182 
3183 int
3184 zonecfg_remove_rctl_value(
3185 	struct zone_rctltab *tabptr,
3186 	struct zone_rctlvaltab *valtabptr)
3187 {
3188 	struct zone_rctlvaltab *last, *this, *next;
3189 
3190 	last = tabptr->zone_rctl_valptr;
3191 	for (this = last; this != NULL; this = this->zone_rctlval_next) {
3192 		if (strcmp(this->zone_rctlval_priv,
3193 		    valtabptr->zone_rctlval_priv) == 0 &&
3194 		    strcmp(this->zone_rctlval_limit,
3195 		    valtabptr->zone_rctlval_limit) == 0 &&
3196 		    strcmp(this->zone_rctlval_action,
3197 		    valtabptr->zone_rctlval_action) == 0) {
3198 			next = this->zone_rctlval_next;
3199 			if (this == tabptr->zone_rctl_valptr)
3200 				tabptr->zone_rctl_valptr = next;
3201 			else
3202 				last->zone_rctlval_next = next;
3203 			free(this);
3204 			return (Z_OK);
3205 		} else
3206 			last = this;
3207 	}
3208 	return (Z_NO_PROPERTY_ID);
3209 }
3210 
3211 void
3212 zonecfg_set_swinv(zone_dochandle_t handle)
3213 {
3214 	handle->zone_dh_sw_inv = B_TRUE;
3215 }
3216 
3217 /*
3218  * Add the pkg to the sw inventory on the handle.
3219  */
3220 int
3221 zonecfg_add_pkg(zone_dochandle_t handle, char *name, char *version)
3222 {
3223 	xmlNodePtr newnode;
3224 	xmlNodePtr cur;
3225 	int err;
3226 
3227 	if ((err = operation_prep(handle)) != Z_OK)
3228 		return (err);
3229 
3230 	cur = handle->zone_dh_cur;
3231 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_PACKAGE, NULL);
3232 	if ((err = newprop(newnode, DTD_ATTR_NAME, name)) != Z_OK)
3233 		return (err);
3234 	if ((err = newprop(newnode, DTD_ATTR_VERSION, version)) != Z_OK)
3235 		return (err);
3236 	return (Z_OK);
3237 }
3238 
3239 int
3240 zonecfg_add_patch(zone_dochandle_t handle, char *id, void **pnode)
3241 {
3242 	xmlNodePtr node = (xmlNodePtr)*pnode;
3243 	xmlNodePtr cur;
3244 	int err;
3245 
3246 	if ((err = operation_prep(handle)) != Z_OK)
3247 		return (err);
3248 
3249 	cur = handle->zone_dh_cur;
3250 	node = xmlNewTextChild(cur, NULL, DTD_ELEM_PATCH, NULL);
3251 	if ((err = newprop(node, DTD_ATTR_ID, id)) != Z_OK)
3252 		return (err);
3253 	*pnode = (void *)node;
3254 	return (Z_OK);
3255 }
3256 
3257 int
3258 zonecfg_add_patch_obs(char *id, void *cur)
3259 {
3260 	xmlNodePtr	node;
3261 	int err;
3262 
3263 	node = xmlNewTextChild((xmlNodePtr)cur, NULL, DTD_ELEM_OBSOLETES, NULL);
3264 	if ((err = newprop(node, DTD_ATTR_ID, id)) != Z_OK)
3265 		return (err);
3266 	return (Z_OK);
3267 }
3268 
3269 char *
3270 zonecfg_strerror(int errnum)
3271 {
3272 	switch (errnum) {
3273 	case Z_OK:
3274 		return (dgettext(TEXT_DOMAIN, "OK"));
3275 	case Z_EMPTY_DOCUMENT:
3276 		return (dgettext(TEXT_DOMAIN, "Empty document"));
3277 	case Z_WRONG_DOC_TYPE:
3278 		return (dgettext(TEXT_DOMAIN, "Wrong document type"));
3279 	case Z_BAD_PROPERTY:
3280 		return (dgettext(TEXT_DOMAIN, "Bad document property"));
3281 	case Z_TEMP_FILE:
3282 		return (dgettext(TEXT_DOMAIN,
3283 		    "Problem creating temporary file"));
3284 	case Z_SAVING_FILE:
3285 		return (dgettext(TEXT_DOMAIN, "Problem saving file"));
3286 	case Z_NO_ENTRY:
3287 		return (dgettext(TEXT_DOMAIN, "No such entry"));
3288 	case Z_BOGUS_ZONE_NAME:
3289 		return (dgettext(TEXT_DOMAIN, "Bogus zone name"));
3290 	case Z_REQD_RESOURCE_MISSING:
3291 		return (dgettext(TEXT_DOMAIN, "Required resource missing"));
3292 	case Z_REQD_PROPERTY_MISSING:
3293 		return (dgettext(TEXT_DOMAIN, "Required property missing"));
3294 	case Z_BAD_HANDLE:
3295 		return (dgettext(TEXT_DOMAIN, "Bad handle"));
3296 	case Z_NOMEM:
3297 		return (dgettext(TEXT_DOMAIN, "Out of memory"));
3298 	case Z_INVAL:
3299 		return (dgettext(TEXT_DOMAIN, "Invalid argument"));
3300 	case Z_ACCES:
3301 		return (dgettext(TEXT_DOMAIN, "Permission denied"));
3302 	case Z_TOO_BIG:
3303 		return (dgettext(TEXT_DOMAIN, "Argument list too long"));
3304 	case Z_MISC_FS:
3305 		return (dgettext(TEXT_DOMAIN,
3306 		    "Miscellaneous file system error"));
3307 	case Z_NO_ZONE:
3308 		return (dgettext(TEXT_DOMAIN, "No such zone configured"));
3309 	case Z_NO_RESOURCE_TYPE:
3310 		return (dgettext(TEXT_DOMAIN, "No such resource type"));
3311 	case Z_NO_RESOURCE_ID:
3312 		return (dgettext(TEXT_DOMAIN, "No such resource with that id"));
3313 	case Z_NO_PROPERTY_TYPE:
3314 		return (dgettext(TEXT_DOMAIN, "No such property type"));
3315 	case Z_NO_PROPERTY_ID:
3316 		return (dgettext(TEXT_DOMAIN, "No such property with that id"));
3317 	case Z_BAD_ZONE_STATE:
3318 		return (dgettext(TEXT_DOMAIN,
3319 		    "Zone state is invalid for the requested operation"));
3320 	case Z_INVALID_DOCUMENT:
3321 		return (dgettext(TEXT_DOMAIN, "Invalid document"));
3322 	case Z_NAME_IN_USE:
3323 		return (dgettext(TEXT_DOMAIN, "Zone name already in use"));
3324 	case Z_NO_SUCH_ID:
3325 		return (dgettext(TEXT_DOMAIN, "No such zone ID"));
3326 	case Z_UPDATING_INDEX:
3327 		return (dgettext(TEXT_DOMAIN, "Problem updating index file"));
3328 	case Z_LOCKING_FILE:
3329 		return (dgettext(TEXT_DOMAIN, "Locking index file"));
3330 	case Z_UNLOCKING_FILE:
3331 		return (dgettext(TEXT_DOMAIN, "Unlocking index file"));
3332 	case Z_INSUFFICIENT_SPEC:
3333 		return (dgettext(TEXT_DOMAIN, "Insufficient specification"));
3334 	case Z_RESOLVED_PATH:
3335 		return (dgettext(TEXT_DOMAIN, "Resolved path mismatch"));
3336 	case Z_IPV6_ADDR_PREFIX_LEN:
3337 		return (dgettext(TEXT_DOMAIN,
3338 		    "IPv6 address missing required prefix length"));
3339 	case Z_BOGUS_ADDRESS:
3340 		return (dgettext(TEXT_DOMAIN,
3341 		    "Neither an IPv4 nor an IPv6 address nor a host name"));
3342 	case Z_PRIV_PROHIBITED:
3343 		return (dgettext(TEXT_DOMAIN,
3344 		    "Specified privilege is prohibited"));
3345 	case Z_PRIV_REQUIRED:
3346 		return (dgettext(TEXT_DOMAIN,
3347 		    "Required privilege is missing"));
3348 	case Z_PRIV_UNKNOWN:
3349 		return (dgettext(TEXT_DOMAIN,
3350 		    "Specified privilege is unknown"));
3351 	case Z_BRAND_ERROR:
3352 		return (dgettext(TEXT_DOMAIN,
3353 		    "Brand-specific error"));
3354 	case Z_INCOMPATIBLE:
3355 		return (dgettext(TEXT_DOMAIN, "Incompatible settings"));
3356 	case Z_ALIAS_DISALLOW:
3357 		return (dgettext(TEXT_DOMAIN,
3358 		    "An incompatible rctl already exists for this property"));
3359 	case Z_CLEAR_DISALLOW:
3360 		return (dgettext(TEXT_DOMAIN,
3361 		    "Clearing this property is not allowed"));
3362 	case Z_POOL:
3363 		return (dgettext(TEXT_DOMAIN, "libpool(3LIB) error"));
3364 	case Z_POOLS_NOT_ACTIVE:
3365 		return (dgettext(TEXT_DOMAIN, "Pools facility not active; "
3366 		    "zone will not be bound to pool"));
3367 	case Z_POOL_ENABLE:
3368 		return (dgettext(TEXT_DOMAIN,
3369 		    "Could not enable pools facility"));
3370 	case Z_NO_POOL:
3371 		return (dgettext(TEXT_DOMAIN,
3372 		    "Pool not found; using default pool"));
3373 	case Z_POOL_CREATE:
3374 		return (dgettext(TEXT_DOMAIN,
3375 		    "Could not create a temporary pool"));
3376 	case Z_POOL_BIND:
3377 		return (dgettext(TEXT_DOMAIN, "Could not bind zone to pool"));
3378 	default:
3379 		return (dgettext(TEXT_DOMAIN, "Unknown error"));
3380 	}
3381 }
3382 
3383 /*
3384  * Note that the zonecfg_setXent() and zonecfg_endXent() calls are all the
3385  * same, as they just turn around and call zonecfg_setent() / zonecfg_endent().
3386  */
3387 
3388 static int
3389 zonecfg_setent(zone_dochandle_t handle)
3390 {
3391 	xmlNodePtr cur;
3392 	int err;
3393 
3394 	if (handle == NULL)
3395 		return (Z_INVAL);
3396 
3397 	if ((err = operation_prep(handle)) != Z_OK) {
3398 		handle->zone_dh_cur = NULL;
3399 		return (err);
3400 	}
3401 	cur = handle->zone_dh_cur;
3402 	cur = cur->xmlChildrenNode;
3403 	handle->zone_dh_cur = cur;
3404 	return (Z_OK);
3405 }
3406 
3407 static int
3408 zonecfg_endent(zone_dochandle_t handle)
3409 {
3410 	if (handle == NULL)
3411 		return (Z_INVAL);
3412 
3413 	handle->zone_dh_cur = handle->zone_dh_top;
3414 	return (Z_OK);
3415 }
3416 
3417 /*
3418  * Do the work required to manipulate a process through libproc.
3419  * If grab_process() returns no errors (0), then release_process()
3420  * must eventually be called.
3421  *
3422  * Return values:
3423  *      0 Successful creation of agent thread
3424  *      1 Error grabbing
3425  *      2 Error creating agent
3426  */
3427 static int
3428 grab_process(pr_info_handle_t *p)
3429 {
3430 	int ret;
3431 
3432 	if ((p->pr = Pgrab(p->pid, 0, &ret)) != NULL) {
3433 
3434 		if (Psetflags(p->pr, PR_RLC) != 0) {
3435 			Prelease(p->pr, 0);
3436 			return (1);
3437 		}
3438 		if (Pcreate_agent(p->pr) == 0) {
3439 			return (0);
3440 
3441 		} else {
3442 			Prelease(p->pr, 0);
3443 			return (2);
3444 		}
3445 	} else {
3446 		return (1);
3447 	}
3448 }
3449 
3450 /*
3451  * Release the specified process. This destroys the agent
3452  * and releases the process. If the process is NULL, nothing
3453  * is done. This function should only be called if grab_process()
3454  * has previously been called and returned success.
3455  *
3456  * This function is Pgrab-safe.
3457  */
3458 static void
3459 release_process(struct ps_prochandle *Pr)
3460 {
3461 	if (Pr == NULL)
3462 		return;
3463 
3464 	Pdestroy_agent(Pr);
3465 	Prelease(Pr, 0);
3466 }
3467 
3468 static boolean_t
3469 grab_zone_proc(char *zonename, pr_info_handle_t *p)
3470 {
3471 	DIR *dirp;
3472 	struct dirent *dentp;
3473 	zoneid_t zoneid;
3474 	int pid_self;
3475 	psinfo_t psinfo;
3476 
3477 	if (zone_get_id(zonename, &zoneid) != 0)
3478 		return (B_FALSE);
3479 
3480 	pid_self = getpid();
3481 
3482 	if ((dirp = opendir("/proc")) == NULL)
3483 		return (B_FALSE);
3484 
3485 	while (dentp = readdir(dirp)) {
3486 		p->pid = atoi(dentp->d_name);
3487 
3488 		/* Skip self */
3489 		if (p->pid == pid_self)
3490 			continue;
3491 
3492 		if (proc_get_psinfo(p->pid, &psinfo) != 0)
3493 			continue;
3494 
3495 		if (psinfo.pr_zoneid != zoneid)
3496 			continue;
3497 
3498 		/* attempt to grab process */
3499 		if (grab_process(p) != 0)
3500 			continue;
3501 
3502 		if (pr_getzoneid(p->pr) != zoneid) {
3503 			release_process(p->pr);
3504 			continue;
3505 		}
3506 
3507 		(void) closedir(dirp);
3508 		return (B_TRUE);
3509 	}
3510 
3511 	(void) closedir(dirp);
3512 	return (B_FALSE);
3513 }
3514 
3515 static boolean_t
3516 get_priv_rctl(struct ps_prochandle *pr, char *name, rctlblk_t *rblk)
3517 {
3518 	if (pr_getrctl(pr, name, NULL, rblk, RCTL_FIRST))
3519 		return (B_FALSE);
3520 
3521 	if (rctlblk_get_privilege(rblk) == RCPRIV_PRIVILEGED)
3522 		return (B_TRUE);
3523 
3524 	while (pr_getrctl(pr, name, rblk, rblk, RCTL_NEXT) == 0) {
3525 		if (rctlblk_get_privilege(rblk) == RCPRIV_PRIVILEGED)
3526 			return (B_TRUE);
3527 	}
3528 
3529 	return (B_FALSE);
3530 }
3531 
3532 /*
3533  * Apply the current rctl settings to the specified, running zone.
3534  */
3535 int
3536 zonecfg_apply_rctls(char *zone_name, zone_dochandle_t handle)
3537 {
3538 	int err;
3539 	int res = Z_OK;
3540 	rctlblk_t *rblk;
3541 	pr_info_handle_t p;
3542 	struct zone_rctltab rctl;
3543 
3544 	if ((err = zonecfg_setrctlent(handle)) != Z_OK)
3545 		return (err);
3546 
3547 	if ((rblk = (rctlblk_t *)malloc(rctlblk_size())) == NULL) {
3548 		(void) zonecfg_endrctlent(handle);
3549 		return (Z_NOMEM);
3550 	}
3551 
3552 	if (!grab_zone_proc(zone_name, &p)) {
3553 		(void) zonecfg_endrctlent(handle);
3554 		free(rblk);
3555 		return (Z_SYSTEM);
3556 	}
3557 
3558 	while (zonecfg_getrctlent(handle, &rctl) == Z_OK) {
3559 		char *rname;
3560 		struct zone_rctlvaltab *valptr;
3561 
3562 		rname = rctl.zone_rctl_name;
3563 
3564 		/* first delete all current privileged settings for this rctl */
3565 		while (get_priv_rctl(p.pr, rname, rblk)) {
3566 			if (pr_setrctl(p.pr, rname, NULL, rblk, RCTL_DELETE) !=
3567 			    0) {
3568 				res = Z_SYSTEM;
3569 				goto done;
3570 			}
3571 		}
3572 
3573 		/* now set each new value for the rctl */
3574 		for (valptr = rctl.zone_rctl_valptr; valptr != NULL;
3575 		    valptr = valptr->zone_rctlval_next) {
3576 			if ((err = zonecfg_construct_rctlblk(valptr, rblk))
3577 			    != Z_OK) {
3578 				res = errno = err;
3579 				goto done;
3580 			}
3581 
3582 			if (pr_setrctl(p.pr, rname, NULL, rblk, RCTL_INSERT)) {
3583 				res = Z_SYSTEM;
3584 				goto done;
3585 			}
3586 		}
3587 	}
3588 
3589 done:
3590 	release_process(p.pr);
3591 	free(rblk);
3592 	(void) zonecfg_endrctlent(handle);
3593 
3594 	return (res);
3595 }
3596 
3597 static const xmlChar *
3598 nm_to_dtd(char *nm)
3599 {
3600 	if (strcmp(nm, "device") == 0)
3601 		return (DTD_ELEM_DEVICE);
3602 	if (strcmp(nm, "fs") == 0)
3603 		return (DTD_ELEM_FS);
3604 	if (strcmp(nm, "inherit-pkg-dir") == 0)
3605 		return (DTD_ELEM_IPD);
3606 	if (strcmp(nm, "net") == 0)
3607 		return (DTD_ELEM_NET);
3608 	if (strcmp(nm, "attr") == 0)
3609 		return (DTD_ELEM_ATTR);
3610 	if (strcmp(nm, "rctl") == 0)
3611 		return (DTD_ELEM_RCTL);
3612 	if (strcmp(nm, "dataset") == 0)
3613 		return (DTD_ELEM_DATASET);
3614 
3615 	return (NULL);
3616 }
3617 
3618 int
3619 zonecfg_num_resources(zone_dochandle_t handle, char *rsrc)
3620 {
3621 	int num = 0;
3622 	const xmlChar *dtd;
3623 	xmlNodePtr cur;
3624 
3625 	if ((dtd = nm_to_dtd(rsrc)) == NULL)
3626 		return (num);
3627 
3628 	if (zonecfg_setent(handle) != Z_OK)
3629 		return (num);
3630 
3631 	for (cur = handle->zone_dh_cur; cur != NULL; cur = cur->next)
3632 		if (xmlStrcmp(cur->name, dtd) == 0)
3633 			num++;
3634 
3635 	(void) zonecfg_endent(handle);
3636 
3637 	return (num);
3638 }
3639 
3640 int
3641 zonecfg_del_all_resources(zone_dochandle_t handle, char *rsrc)
3642 {
3643 	int err;
3644 	const xmlChar *dtd;
3645 	xmlNodePtr cur;
3646 
3647 	if ((dtd = nm_to_dtd(rsrc)) == NULL)
3648 		return (Z_NO_RESOURCE_TYPE);
3649 
3650 	if ((err = zonecfg_setent(handle)) != Z_OK)
3651 		return (err);
3652 
3653 	cur = handle->zone_dh_cur;
3654 	while (cur != NULL) {
3655 		xmlNodePtr tmp;
3656 
3657 		if (xmlStrcmp(cur->name, dtd)) {
3658 			cur = cur->next;
3659 			continue;
3660 		}
3661 
3662 		tmp = cur->next;
3663 		xmlUnlinkNode(cur);
3664 		xmlFreeNode(cur);
3665 		cur = tmp;
3666 	}
3667 
3668 	(void) zonecfg_endent(handle);
3669 	return (Z_OK);
3670 }
3671 
3672 static boolean_t
3673 valid_uint(char *s, uint64_t *n)
3674 {
3675 	char *endp;
3676 
3677 	/* strtoull accepts '-'?! so we want to flag that as an error */
3678 	if (strchr(s, '-') != NULL)
3679 		return (B_FALSE);
3680 
3681 	errno = 0;
3682 	*n = strtoull(s, &endp, 10);
3683 
3684 	if (errno != 0 || *endp != '\0')
3685 		return (B_FALSE);
3686 	return (B_TRUE);
3687 }
3688 
3689 /*
3690  * Convert a string representing a number (possibly a fraction) into an integer.
3691  * The string can have a modifier (K, M, G or T).   The modifiers are treated
3692  * as powers of two (not 10).
3693  */
3694 int
3695 zonecfg_str_to_bytes(char *str, uint64_t *bytes)
3696 {
3697 	long double val;
3698 	char *unitp;
3699 	uint64_t scale;
3700 
3701 	if ((val = strtold(str, &unitp)) < 0)
3702 		return (-1);
3703 
3704 	/* remove any leading white space from units string */
3705 	while (isspace(*unitp) != 0)
3706 		++unitp;
3707 
3708 	/* if no units explicitly set, error */
3709 	if (unitp == NULL || *unitp == '\0') {
3710 		scale = 1;
3711 	} else {
3712 		int i;
3713 		char *units[] = {"K", "M", "G", "T", NULL};
3714 
3715 		scale = 1024;
3716 
3717 		/* update scale based on units */
3718 		for (i = 0; units[i] != NULL; i++) {
3719 			if (strcasecmp(unitp, units[i]) == 0)
3720 				break;
3721 			scale <<= 10;
3722 		}
3723 
3724 		if (units[i] == NULL)
3725 			return (-1);
3726 	}
3727 
3728 	*bytes = (uint64_t)(val * scale);
3729 	return (0);
3730 }
3731 
3732 boolean_t
3733 zonecfg_valid_ncpus(char *lowstr, char *highstr)
3734 {
3735 	uint64_t low, high;
3736 
3737 	if (!valid_uint(lowstr, &low) || !valid_uint(highstr, &high) ||
3738 	    low < 1 || low > high)
3739 		return (B_FALSE);
3740 
3741 	return (B_TRUE);
3742 }
3743 
3744 boolean_t
3745 zonecfg_valid_importance(char *impstr)
3746 {
3747 	uint64_t num;
3748 
3749 	if (!valid_uint(impstr, &num))
3750 		return (B_FALSE);
3751 
3752 	return (B_TRUE);
3753 }
3754 
3755 boolean_t
3756 zonecfg_valid_alias_limit(char *name, char *limitstr, uint64_t *limit)
3757 {
3758 	int i;
3759 
3760 	for (i = 0; aliases[i].shortname != NULL; i++)
3761 		if (strcmp(name, aliases[i].shortname) == 0)
3762 			break;
3763 
3764 	if (aliases[i].shortname == NULL)
3765 		return (B_FALSE);
3766 
3767 	if (!valid_uint(limitstr, limit) || *limit < aliases[i].low_limit)
3768 		return (B_FALSE);
3769 
3770 	return (B_TRUE);
3771 }
3772 
3773 boolean_t
3774 zonecfg_valid_memlimit(char *memstr, uint64_t *mem_val)
3775 {
3776 	if (zonecfg_str_to_bytes(memstr, mem_val) != 0)
3777 		return (B_FALSE);
3778 
3779 	return (B_TRUE);
3780 }
3781 
3782 static int
3783 zerr_pool(char *pool_err, int err_size, int res)
3784 {
3785 	(void) strlcpy(pool_err, pool_strerror(pool_error()), err_size);
3786 	return (res);
3787 }
3788 
3789 static int
3790 create_tmp_pset(char *pool_err, int err_size, pool_conf_t *pconf, pool_t *pool,
3791     char *name, int min, int max)
3792 {
3793 	pool_resource_t *res;
3794 	pool_elem_t *elem;
3795 	pool_value_t *val;
3796 
3797 	if ((res = pool_resource_create(pconf, "pset", name)) == NULL)
3798 		return (zerr_pool(pool_err, err_size, Z_POOL));
3799 
3800 	if (pool_associate(pconf, pool, res) != PO_SUCCESS)
3801 		return (zerr_pool(pool_err, err_size, Z_POOL));
3802 
3803 	if ((elem = pool_resource_to_elem(pconf, res)) == NULL)
3804 		return (zerr_pool(pool_err, err_size, Z_POOL));
3805 
3806 	if ((val = pool_value_alloc()) == NULL)
3807 		return (zerr_pool(pool_err, err_size, Z_POOL));
3808 
3809 	/* set the maximum number of cpus for the pset */
3810 	pool_value_set_uint64(val, (uint64_t)max);
3811 
3812 	if (pool_put_property(pconf, elem, "pset.max", val) != PO_SUCCESS) {
3813 		pool_value_free(val);
3814 		return (zerr_pool(pool_err, err_size, Z_POOL));
3815 	}
3816 
3817 	/* set the minimum number of cpus for the pset */
3818 	pool_value_set_uint64(val, (uint64_t)min);
3819 
3820 	if (pool_put_property(pconf, elem, "pset.min", val) != PO_SUCCESS) {
3821 		pool_value_free(val);
3822 		return (zerr_pool(pool_err, err_size, Z_POOL));
3823 	}
3824 
3825 	pool_value_free(val);
3826 
3827 	return (Z_OK);
3828 }
3829 
3830 static int
3831 create_tmp_pool(char *pool_err, int err_size, pool_conf_t *pconf, char *name,
3832     struct zone_psettab *pset_tab)
3833 {
3834 	pool_t *pool;
3835 	int res = Z_OK;
3836 
3837 	/* create a temporary pool configuration */
3838 	if (pool_conf_open(pconf, NULL, PO_TEMP) != PO_SUCCESS) {
3839 		res = zerr_pool(pool_err, err_size, Z_POOL);
3840 		return (res);
3841 	}
3842 
3843 	if ((pool = pool_create(pconf, name)) == NULL) {
3844 		res = zerr_pool(pool_err, err_size, Z_POOL_CREATE);
3845 		goto done;
3846 	}
3847 
3848 	/* set pool importance */
3849 	if (pset_tab->zone_importance[0] != '\0') {
3850 		pool_elem_t *elem;
3851 		pool_value_t *val;
3852 
3853 		if ((elem = pool_to_elem(pconf, pool)) == NULL) {
3854 			res = zerr_pool(pool_err, err_size, Z_POOL);
3855 			goto done;
3856 		}
3857 
3858 		if ((val = pool_value_alloc()) == NULL) {
3859 			res = zerr_pool(pool_err, err_size, Z_POOL);
3860 			goto done;
3861 		}
3862 
3863 		pool_value_set_int64(val,
3864 		    (int64_t)atoi(pset_tab->zone_importance));
3865 
3866 		if (pool_put_property(pconf, elem, "pool.importance", val)
3867 		    != PO_SUCCESS) {
3868 			res = zerr_pool(pool_err, err_size, Z_POOL);
3869 			pool_value_free(val);
3870 			goto done;
3871 		}
3872 
3873 		pool_value_free(val);
3874 	}
3875 
3876 	if ((res = create_tmp_pset(pool_err, err_size, pconf, pool, name,
3877 	    atoi(pset_tab->zone_ncpu_min),
3878 	    atoi(pset_tab->zone_ncpu_max))) != Z_OK)
3879 		goto done;
3880 
3881 	/* validation */
3882 	if (pool_conf_status(pconf) == POF_INVALID) {
3883 		res = zerr_pool(pool_err, err_size, Z_POOL);
3884 		goto done;
3885 	}
3886 
3887 	/*
3888 	 * This validation is the one we expect to fail if the user specified
3889 	 * an invalid configuration (too many cpus) for this system.
3890 	 */
3891 	if (pool_conf_validate(pconf, POV_RUNTIME) != PO_SUCCESS) {
3892 		res = zerr_pool(pool_err, err_size, Z_POOL_CREATE);
3893 		goto done;
3894 	}
3895 
3896 	/*
3897 	 * Commit the dynamic configuration but not the pool configuration
3898 	 * file.
3899 	 */
3900 	if (pool_conf_commit(pconf, 1) != PO_SUCCESS)
3901 		res = zerr_pool(pool_err, err_size, Z_POOL);
3902 
3903 done:
3904 	(void) pool_conf_close(pconf);
3905 	return (res);
3906 }
3907 
3908 static int
3909 get_running_tmp_pset(pool_conf_t *pconf, pool_t *pool, pool_resource_t *pset,
3910     struct zone_psettab *pset_tab)
3911 {
3912 	int nfound = 0;
3913 	pool_elem_t *pe;
3914 	pool_value_t *pv = pool_value_alloc();
3915 	uint64_t val_uint;
3916 
3917 	if (pool != NULL) {
3918 		pe = pool_to_elem(pconf, pool);
3919 		if (pool_get_property(pconf, pe, "pool.importance", pv)
3920 		    != POC_INVAL) {
3921 			int64_t val_int;
3922 
3923 			(void) pool_value_get_int64(pv, &val_int);
3924 			(void) snprintf(pset_tab->zone_importance,
3925 			    sizeof (pset_tab->zone_importance), "%d", val_int);
3926 			nfound++;
3927 		}
3928 	}
3929 
3930 	if (pset != NULL) {
3931 		pe = pool_resource_to_elem(pconf, pset);
3932 		if (pool_get_property(pconf, pe, "pset.min", pv) != POC_INVAL) {
3933 			(void) pool_value_get_uint64(pv, &val_uint);
3934 			(void) snprintf(pset_tab->zone_ncpu_min,
3935 			    sizeof (pset_tab->zone_ncpu_min), "%u", val_uint);
3936 			nfound++;
3937 		}
3938 
3939 		if (pool_get_property(pconf, pe, "pset.max", pv) != POC_INVAL) {
3940 			(void) pool_value_get_uint64(pv, &val_uint);
3941 			(void) snprintf(pset_tab->zone_ncpu_max,
3942 			    sizeof (pset_tab->zone_ncpu_max), "%u", val_uint);
3943 			nfound++;
3944 		}
3945 	}
3946 
3947 	pool_value_free(pv);
3948 
3949 	if (nfound == 3)
3950 		return (PO_SUCCESS);
3951 
3952 	return (PO_FAIL);
3953 }
3954 
3955 /*
3956  * Determine if a tmp pool is configured and if so, if the configuration is
3957  * still valid or if it has been changed since the tmp pool was created.
3958  * If the tmp pool configuration is no longer valid, delete the tmp pool.
3959  *
3960  * Set *valid=B_TRUE if there is an existing, valid tmp pool configuration.
3961  */
3962 static int
3963 verify_del_tmp_pool(pool_conf_t *pconf, char *tmp_name, char *pool_err,
3964     int err_size, struct zone_psettab *pset_tab, boolean_t *exists)
3965 {
3966 	int res = Z_OK;
3967 	pool_t *pool;
3968 	pool_resource_t *pset;
3969 	struct zone_psettab pset_current;
3970 
3971 	*exists = B_FALSE;
3972 
3973 	if (pool_conf_open(pconf, pool_dynamic_location(), PO_RDWR)
3974 	    != PO_SUCCESS) {
3975 		res = zerr_pool(pool_err, err_size, Z_POOL);
3976 		return (res);
3977 	}
3978 
3979 	pool = pool_get_pool(pconf, tmp_name);
3980 	pset = pool_get_resource(pconf, "pset", tmp_name);
3981 
3982 	if (pool == NULL && pset == NULL) {
3983 		/* no tmp pool configured */
3984 		goto done;
3985 	}
3986 
3987 	/*
3988 	 * If an existing tmp pool for this zone is configured with the proper
3989 	 * settings, then the tmp pool is valid.
3990 	 */
3991 	if (get_running_tmp_pset(pconf, pool, pset, &pset_current)
3992 	    == PO_SUCCESS &&
3993 	    strcmp(pset_tab->zone_ncpu_min,
3994 	    pset_current.zone_ncpu_min) == 0 &&
3995 	    strcmp(pset_tab->zone_ncpu_max,
3996 	    pset_current.zone_ncpu_max) == 0 &&
3997 	    strcmp(pset_tab->zone_importance,
3998 	    pset_current.zone_importance) == 0) {
3999 		*exists = B_TRUE;
4000 
4001 	} else {
4002 		/*
4003 		 * An out-of-date tmp pool configuration exists.  Delete it
4004 		 * so that we can create the correct tmp pool config.
4005 		 */
4006 		if (pset != NULL &&
4007 		    pool_resource_destroy(pconf, pset) != PO_SUCCESS) {
4008 			res = zerr_pool(pool_err, err_size, Z_POOL);
4009 			goto done;
4010 		}
4011 
4012 		if (pool != NULL &&
4013 		    pool_destroy(pconf, pool) != PO_SUCCESS) {
4014 			res = zerr_pool(pool_err, err_size, Z_POOL);
4015 			goto done;
4016 		}
4017 
4018 		/* commit dynamic config */
4019 		if (pool_conf_commit(pconf, 0) != PO_SUCCESS)
4020 			res = zerr_pool(pool_err, err_size, Z_POOL);
4021 	}
4022 
4023 done:
4024 	(void) pool_conf_close(pconf);
4025 
4026 	return (res);
4027 }
4028 
4029 /*
4030  * Destroy any existing tmp pool.
4031  */
4032 int
4033 zonecfg_destroy_tmp_pool(char *zone_name, char *pool_err, int err_size)
4034 {
4035 	int status;
4036 	int res = Z_OK;
4037 	pool_conf_t *pconf;
4038 	pool_t *pool;
4039 	pool_resource_t *pset;
4040 	char tmp_name[MAX_TMP_POOL_NAME];
4041 
4042 	/* if pools not enabled then nothing to do */
4043 	if (pool_get_status(&status) != PO_SUCCESS || status != POOL_ENABLED)
4044 		return (Z_OK);
4045 
4046 	if ((pconf = pool_conf_alloc()) == NULL)
4047 		return (zerr_pool(pool_err, err_size, Z_POOL));
4048 
4049 	(void) snprintf(tmp_name, sizeof (tmp_name), TMP_POOL_NAME, zone_name);
4050 
4051 	if (pool_conf_open(pconf, pool_dynamic_location(), PO_RDWR)
4052 	    != PO_SUCCESS) {
4053 		res = zerr_pool(pool_err, err_size, Z_POOL);
4054 		pool_conf_free(pconf);
4055 		return (res);
4056 	}
4057 
4058 	pool = pool_get_pool(pconf, tmp_name);
4059 	pset = pool_get_resource(pconf, "pset", tmp_name);
4060 
4061 	if (pool == NULL && pset == NULL) {
4062 		/* nothing to destroy, we're done */
4063 		goto done;
4064 	}
4065 
4066 	if (pset != NULL && pool_resource_destroy(pconf, pset) != PO_SUCCESS) {
4067 		res = zerr_pool(pool_err, err_size, Z_POOL);
4068 		goto done;
4069 	}
4070 
4071 	if (pool != NULL && pool_destroy(pconf, pool) != PO_SUCCESS) {
4072 		res = zerr_pool(pool_err, err_size, Z_POOL);
4073 		goto done;
4074 	}
4075 
4076 	/* commit dynamic config */
4077 	if (pool_conf_commit(pconf, 0) != PO_SUCCESS)
4078 		res = zerr_pool(pool_err, err_size, Z_POOL);
4079 
4080 done:
4081 	(void) pool_conf_close(pconf);
4082 	pool_conf_free(pconf);
4083 
4084 	return (res);
4085 }
4086 
4087 /*
4088  * Attempt to bind to a tmp pool for this zone.  If there is no tmp pool
4089  * configured, we just return Z_OK.
4090  *
4091  * We either attempt to create the tmp pool for this zone or rebind to an
4092  * existing tmp pool for this zone.
4093  *
4094  * Rebinding is used when a zone with a tmp pool reboots so that we don't have
4095  * to recreate the tmp pool.  To do this we need to be sure we work correctly
4096  * for the following cases:
4097  *
4098  *	- there is an existing, properly configured tmp pool.
4099  *	- zonecfg added tmp pool after zone was booted, must now create.
4100  *	- zonecfg updated tmp pool config after zone was booted, in this case
4101  *	  we destroy the old tmp pool and create a new one.
4102  */
4103 int
4104 zonecfg_bind_tmp_pool(zone_dochandle_t handle, zoneid_t zoneid, char *pool_err,
4105     int err_size)
4106 {
4107 	struct zone_psettab pset_tab;
4108 	int err;
4109 	int status;
4110 	pool_conf_t *pconf;
4111 	boolean_t exists;
4112 	char zone_name[ZONENAME_MAX];
4113 	char tmp_name[MAX_TMP_POOL_NAME];
4114 
4115 	(void) getzonenamebyid(zoneid, zone_name, sizeof (zone_name));
4116 
4117 	err = zonecfg_lookup_pset(handle, &pset_tab);
4118 
4119 	/* if no temporary pool configured, we're done */
4120 	if (err == Z_NO_ENTRY)
4121 		return (Z_OK);
4122 
4123 	/*
4124 	 * importance might not have a value but we need to validate it here,
4125 	 * so set the default.
4126 	 */
4127 	if (pset_tab.zone_importance[0] == '\0')
4128 		(void) strlcpy(pset_tab.zone_importance, "1",
4129 		    sizeof (pset_tab.zone_importance));
4130 
4131 	/* if pools not enabled, enable them now */
4132 	if (pool_get_status(&status) != PO_SUCCESS || status != POOL_ENABLED) {
4133 		if (pool_set_status(POOL_ENABLED) != PO_SUCCESS)
4134 			return (Z_POOL_ENABLE);
4135 	}
4136 
4137 	if ((pconf = pool_conf_alloc()) == NULL)
4138 		return (zerr_pool(pool_err, err_size, Z_POOL));
4139 
4140 	(void) snprintf(tmp_name, sizeof (tmp_name), TMP_POOL_NAME, zone_name);
4141 
4142 	/*
4143 	 * Check if a valid tmp pool/pset already exists.  If so, we just
4144 	 * reuse it.
4145 	 */
4146 	if ((err = verify_del_tmp_pool(pconf, tmp_name, pool_err, err_size,
4147 	    &pset_tab, &exists)) != Z_OK) {
4148 		pool_conf_free(pconf);
4149 		return (err);
4150 	}
4151 
4152 	if (!exists)
4153 		err = create_tmp_pool(pool_err, err_size, pconf, tmp_name,
4154 		    &pset_tab);
4155 
4156 	pool_conf_free(pconf);
4157 
4158 	if (err != Z_OK)
4159 		return (err);
4160 
4161 	/* Bind the zone to the pool. */
4162 	if (pool_set_binding(tmp_name, P_ZONEID, zoneid) != PO_SUCCESS)
4163 		return (zerr_pool(pool_err, err_size, Z_POOL_BIND));
4164 
4165 	return (Z_OK);
4166 }
4167 
4168 /*
4169  * Attempt to bind to a permanent pool for this zone.  If there is no
4170  * permanent pool configured, we just return Z_OK.
4171  */
4172 int
4173 zonecfg_bind_pool(zone_dochandle_t handle, zoneid_t zoneid, char *pool_err,
4174     int err_size)
4175 {
4176 	pool_conf_t *poolconf;
4177 	pool_t *pool;
4178 	char poolname[MAXPATHLEN];
4179 	int status;
4180 	int error;
4181 
4182 	/*
4183 	 * Find the pool mentioned in the zone configuration, and bind to it.
4184 	 */
4185 	error = zonecfg_get_pool(handle, poolname, sizeof (poolname));
4186 	if (error == Z_NO_ENTRY || (error == Z_OK && strlen(poolname) == 0)) {
4187 		/*
4188 		 * The property is not set on the zone, so the pool
4189 		 * should be bound to the default pool.  But that's
4190 		 * already done by the kernel, so we can just return.
4191 		 */
4192 		return (Z_OK);
4193 	}
4194 	if (error != Z_OK) {
4195 		/*
4196 		 * Not an error, even though it shouldn't be happening.
4197 		 */
4198 		return (Z_OK);
4199 	}
4200 	/*
4201 	 * Don't do anything if pools aren't enabled.
4202 	 */
4203 	if (pool_get_status(&status) != PO_SUCCESS || status != POOL_ENABLED)
4204 		return (Z_POOLS_NOT_ACTIVE);
4205 
4206 	/*
4207 	 * Try to provide a sane error message if the requested pool doesn't
4208 	 * exist.
4209 	 */
4210 	if ((poolconf = pool_conf_alloc()) == NULL)
4211 		return (zerr_pool(pool_err, err_size, Z_POOL));
4212 
4213 	if (pool_conf_open(poolconf, pool_dynamic_location(), PO_RDONLY) !=
4214 	    PO_SUCCESS) {
4215 		pool_conf_free(poolconf);
4216 		return (zerr_pool(pool_err, err_size, Z_POOL));
4217 	}
4218 	pool = pool_get_pool(poolconf, poolname);
4219 	(void) pool_conf_close(poolconf);
4220 	pool_conf_free(poolconf);
4221 	if (pool == NULL)
4222 		return (Z_NO_POOL);
4223 
4224 	/*
4225 	 * Bind the zone to the pool.
4226 	 */
4227 	if (pool_set_binding(poolname, P_ZONEID, zoneid) != PO_SUCCESS) {
4228 		/* if bind fails, return poolname for the error msg */
4229 		(void) strlcpy(pool_err, poolname, err_size);
4230 		return (Z_POOL_BIND);
4231 	}
4232 
4233 	return (Z_OK);
4234 }
4235 
4236 
4237 static boolean_t
4238 svc_enabled(char *svc_name)
4239 {
4240 	scf_simple_prop_t	*prop;
4241 	boolean_t		found = B_FALSE;
4242 
4243 	prop = scf_simple_prop_get(NULL, svc_name, SCF_PG_GENERAL,
4244 	    SCF_PROPERTY_ENABLED);
4245 
4246 	if (scf_simple_prop_numvalues(prop) == 1 &&
4247 	    *scf_simple_prop_next_boolean(prop) != 0)
4248 		found = B_TRUE;
4249 
4250 	scf_simple_prop_free(prop);
4251 
4252 	return (found);
4253 }
4254 
4255 /*
4256  * If the zone has capped-memory, make sure the rcap service is enabled.
4257  */
4258 int
4259 zonecfg_enable_rcapd(char *err, int size)
4260 {
4261 	if (!svc_enabled(RCAP_SERVICE) &&
4262 	    smf_enable_instance(RCAP_SERVICE, 0) == -1) {
4263 		(void) strlcpy(err, scf_strerror(scf_error()), size);
4264 		return (Z_SYSTEM);
4265 	}
4266 
4267 	return (Z_OK);
4268 }
4269 
4270 /*
4271  * Return true if pset has cpu range specified and poold is not enabled.
4272  */
4273 boolean_t
4274 zonecfg_warn_poold(zone_dochandle_t handle)
4275 {
4276 	struct zone_psettab pset_tab;
4277 	int min, max;
4278 	int err;
4279 
4280 	err = zonecfg_lookup_pset(handle, &pset_tab);
4281 
4282 	/* if no temporary pool configured, we're done */
4283 	if (err == Z_NO_ENTRY)
4284 		return (B_FALSE);
4285 
4286 	min = atoi(pset_tab.zone_ncpu_min);
4287 	max = atoi(pset_tab.zone_ncpu_max);
4288 
4289 	/* range not specified, no need for poold */
4290 	if (min == max)
4291 		return (B_FALSE);
4292 
4293 	/* we have a range, check if poold service is enabled */
4294 	if (svc_enabled(POOLD_SERVICE))
4295 		return (B_FALSE);
4296 
4297 	return (B_TRUE);
4298 }
4299 
4300 static int
4301 get_pool_sched_class(char *poolname, char *class, int clsize)
4302 {
4303 	int status;
4304 	pool_conf_t *poolconf;
4305 	pool_t *pool;
4306 	pool_elem_t *pe;
4307 	pool_value_t *pv = pool_value_alloc();
4308 	const char *sched_str;
4309 
4310 	if (pool_get_status(&status) != PO_SUCCESS || status != POOL_ENABLED)
4311 		return (Z_NO_POOL);
4312 
4313 	if ((poolconf = pool_conf_alloc()) == NULL)
4314 		return (Z_NO_POOL);
4315 
4316 	if (pool_conf_open(poolconf, pool_dynamic_location(), PO_RDONLY) !=
4317 	    PO_SUCCESS) {
4318 		pool_conf_free(poolconf);
4319 		return (Z_NO_POOL);
4320 	}
4321 
4322 	if ((pool = pool_get_pool(poolconf, poolname)) == NULL) {
4323 		(void) pool_conf_close(poolconf);
4324 		pool_conf_free(poolconf);
4325 		return (Z_NO_POOL);
4326 	}
4327 
4328 	pe = pool_to_elem(poolconf, pool);
4329 	if (pool_get_property(poolconf, pe, "pool.scheduler", pv)
4330 	    != POC_INVAL) {
4331 		(void) pool_value_get_string(pv, &sched_str);
4332 		if (strlcpy(class, sched_str, clsize) >= clsize)
4333 			return (Z_TOO_BIG);
4334 	}
4335 
4336 	(void) pool_conf_close(poolconf);
4337 	pool_conf_free(poolconf);
4338 	return (Z_OK);
4339 }
4340 
4341 /*
4342  * Get the default scheduling class for the zone.  This will either be the
4343  * class set on the zone's pool or the system default scheduling class.
4344  */
4345 int
4346 zonecfg_get_dflt_sched_class(zone_dochandle_t handle, char *class, int clsize)
4347 {
4348 	char poolname[MAXPATHLEN];
4349 
4350 	if (zonecfg_get_pool(handle, poolname, sizeof (poolname)) == Z_OK) {
4351 		/* check if the zone's pool specified a sched class */
4352 		if (get_pool_sched_class(poolname, class, clsize) == Z_OK)
4353 			return (Z_OK);
4354 	}
4355 
4356 	if (priocntl(0, 0, PC_GETDFLCL, class, (uint64_t)clsize) == -1)
4357 		return (Z_TOO_BIG);
4358 
4359 	return (Z_OK);
4360 }
4361 
4362 int
4363 zonecfg_setfsent(zone_dochandle_t handle)
4364 {
4365 	return (zonecfg_setent(handle));
4366 }
4367 
4368 int
4369 zonecfg_getfsent(zone_dochandle_t handle, struct zone_fstab *tabptr)
4370 {
4371 	xmlNodePtr cur, options;
4372 	char options_str[MAX_MNTOPT_STR];
4373 	int err;
4374 
4375 	if (handle == NULL)
4376 		return (Z_INVAL);
4377 
4378 	if ((cur = handle->zone_dh_cur) == NULL)
4379 		return (Z_NO_ENTRY);
4380 
4381 	for (; cur != NULL; cur = cur->next)
4382 		if (!xmlStrcmp(cur->name, DTD_ELEM_FS))
4383 			break;
4384 	if (cur == NULL) {
4385 		handle->zone_dh_cur = handle->zone_dh_top;
4386 		return (Z_NO_ENTRY);
4387 	}
4388 
4389 	if ((err = fetchprop(cur, DTD_ATTR_SPECIAL, tabptr->zone_fs_special,
4390 	    sizeof (tabptr->zone_fs_special))) != Z_OK) {
4391 		handle->zone_dh_cur = handle->zone_dh_top;
4392 		return (err);
4393 	}
4394 
4395 	if ((err = fetchprop(cur, DTD_ATTR_RAW, tabptr->zone_fs_raw,
4396 	    sizeof (tabptr->zone_fs_raw))) != Z_OK) {
4397 		handle->zone_dh_cur = handle->zone_dh_top;
4398 		return (err);
4399 	}
4400 
4401 	if ((err = fetchprop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir,
4402 	    sizeof (tabptr->zone_fs_dir))) != Z_OK) {
4403 		handle->zone_dh_cur = handle->zone_dh_top;
4404 		return (err);
4405 	}
4406 
4407 	if ((err = fetchprop(cur, DTD_ATTR_TYPE, tabptr->zone_fs_type,
4408 	    sizeof (tabptr->zone_fs_type))) != Z_OK) {
4409 		handle->zone_dh_cur = handle->zone_dh_top;
4410 		return (err);
4411 	}
4412 
4413 	/* OK for options to be NULL */
4414 	tabptr->zone_fs_options = NULL;
4415 	for (options = cur->xmlChildrenNode; options != NULL;
4416 	    options = options->next) {
4417 		if (fetchprop(options, DTD_ATTR_NAME, options_str,
4418 		    sizeof (options_str)) != Z_OK)
4419 			break;
4420 		if (zonecfg_add_fs_option(tabptr, options_str) != Z_OK)
4421 			break;
4422 	}
4423 
4424 	handle->zone_dh_cur = cur->next;
4425 	return (Z_OK);
4426 }
4427 
4428 int
4429 zonecfg_endfsent(zone_dochandle_t handle)
4430 {
4431 	return (zonecfg_endent(handle));
4432 }
4433 
4434 int
4435 zonecfg_setipdent(zone_dochandle_t handle)
4436 {
4437 	return (zonecfg_setent(handle));
4438 }
4439 
4440 int
4441 zonecfg_getipdent(zone_dochandle_t handle, struct zone_fstab *tabptr)
4442 {
4443 	xmlNodePtr cur;
4444 	int err;
4445 
4446 	if (handle == NULL)
4447 		return (Z_INVAL);
4448 
4449 	if ((cur = handle->zone_dh_cur) == NULL)
4450 		return (Z_NO_ENTRY);
4451 
4452 	for (; cur != NULL; cur = cur->next)
4453 		if (!xmlStrcmp(cur->name, DTD_ELEM_IPD))
4454 			break;
4455 	if (cur == NULL) {
4456 		handle->zone_dh_cur = handle->zone_dh_top;
4457 		return (Z_NO_ENTRY);
4458 	}
4459 
4460 	if ((err = fetchprop(cur, DTD_ATTR_DIR, tabptr->zone_fs_dir,
4461 	    sizeof (tabptr->zone_fs_dir))) != Z_OK) {
4462 		handle->zone_dh_cur = handle->zone_dh_top;
4463 		return (err);
4464 	}
4465 
4466 	handle->zone_dh_cur = cur->next;
4467 	return (Z_OK);
4468 }
4469 
4470 int
4471 zonecfg_endipdent(zone_dochandle_t handle)
4472 {
4473 	return (zonecfg_endent(handle));
4474 }
4475 
4476 int
4477 zonecfg_setnwifent(zone_dochandle_t handle)
4478 {
4479 	return (zonecfg_setent(handle));
4480 }
4481 
4482 int
4483 zonecfg_getnwifent(zone_dochandle_t handle, struct zone_nwiftab *tabptr)
4484 {
4485 	xmlNodePtr cur;
4486 	int err;
4487 
4488 	if (handle == NULL)
4489 		return (Z_INVAL);
4490 
4491 	if ((cur = handle->zone_dh_cur) == NULL)
4492 		return (Z_NO_ENTRY);
4493 
4494 	for (; cur != NULL; cur = cur->next)
4495 		if (!xmlStrcmp(cur->name, DTD_ELEM_NET))
4496 			break;
4497 	if (cur == NULL) {
4498 		handle->zone_dh_cur = handle->zone_dh_top;
4499 		return (Z_NO_ENTRY);
4500 	}
4501 
4502 	if ((err = fetchprop(cur, DTD_ATTR_ADDRESS, tabptr->zone_nwif_address,
4503 	    sizeof (tabptr->zone_nwif_address))) != Z_OK) {
4504 		handle->zone_dh_cur = handle->zone_dh_top;
4505 		return (err);
4506 	}
4507 
4508 	if ((err = fetchprop(cur, DTD_ATTR_PHYSICAL, tabptr->zone_nwif_physical,
4509 	    sizeof (tabptr->zone_nwif_physical))) != Z_OK) {
4510 		handle->zone_dh_cur = handle->zone_dh_top;
4511 		return (err);
4512 	}
4513 
4514 	if ((err = fetchprop(cur, DTD_ATTR_DEFROUTER,
4515 	    tabptr->zone_nwif_defrouter,
4516 	    sizeof (tabptr->zone_nwif_defrouter))) != Z_OK) {
4517 		handle->zone_dh_cur = handle->zone_dh_top;
4518 		return (err);
4519 	}
4520 
4521 	handle->zone_dh_cur = cur->next;
4522 	return (Z_OK);
4523 }
4524 
4525 int
4526 zonecfg_endnwifent(zone_dochandle_t handle)
4527 {
4528 	return (zonecfg_endent(handle));
4529 }
4530 
4531 int
4532 zonecfg_setdevent(zone_dochandle_t handle)
4533 {
4534 	return (zonecfg_setent(handle));
4535 }
4536 
4537 int
4538 zonecfg_getdevent(zone_dochandle_t handle, struct zone_devtab *tabptr)
4539 {
4540 	xmlNodePtr cur;
4541 	int err;
4542 
4543 	if (handle == NULL)
4544 		return (Z_INVAL);
4545 
4546 	if ((cur = handle->zone_dh_cur) == NULL)
4547 		return (Z_NO_ENTRY);
4548 
4549 	for (; cur != NULL; cur = cur->next)
4550 		if (!xmlStrcmp(cur->name, DTD_ELEM_DEVICE))
4551 			break;
4552 	if (cur == NULL) {
4553 		handle->zone_dh_cur = handle->zone_dh_top;
4554 		return (Z_NO_ENTRY);
4555 	}
4556 
4557 	if ((err = fetchprop(cur, DTD_ATTR_MATCH, tabptr->zone_dev_match,
4558 	    sizeof (tabptr->zone_dev_match))) != Z_OK) {
4559 		handle->zone_dh_cur = handle->zone_dh_top;
4560 		return (err);
4561 	}
4562 
4563 	handle->zone_dh_cur = cur->next;
4564 	return (Z_OK);
4565 }
4566 
4567 int
4568 zonecfg_enddevent(zone_dochandle_t handle)
4569 {
4570 	return (zonecfg_endent(handle));
4571 }
4572 
4573 int
4574 zonecfg_setrctlent(zone_dochandle_t handle)
4575 {
4576 	return (zonecfg_setent(handle));
4577 }
4578 
4579 int
4580 zonecfg_getrctlent(zone_dochandle_t handle, struct zone_rctltab *tabptr)
4581 {
4582 	xmlNodePtr cur, val;
4583 	struct zone_rctlvaltab *valptr;
4584 	int err;
4585 
4586 	if (handle == NULL)
4587 		return (Z_INVAL);
4588 
4589 	if ((cur = handle->zone_dh_cur) == NULL)
4590 		return (Z_NO_ENTRY);
4591 
4592 	for (; cur != NULL; cur = cur->next)
4593 		if (!xmlStrcmp(cur->name, DTD_ELEM_RCTL))
4594 			break;
4595 	if (cur == NULL) {
4596 		handle->zone_dh_cur = handle->zone_dh_top;
4597 		return (Z_NO_ENTRY);
4598 	}
4599 
4600 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_rctl_name,
4601 	    sizeof (tabptr->zone_rctl_name))) != Z_OK) {
4602 		handle->zone_dh_cur = handle->zone_dh_top;
4603 		return (err);
4604 	}
4605 
4606 	tabptr->zone_rctl_valptr = NULL;
4607 	for (val = cur->xmlChildrenNode; val != NULL; val = val->next) {
4608 		valptr = (struct zone_rctlvaltab *)malloc(
4609 		    sizeof (struct zone_rctlvaltab));
4610 		if (valptr == NULL)
4611 			return (Z_NOMEM);
4612 		if (fetchprop(val, DTD_ATTR_PRIV, valptr->zone_rctlval_priv,
4613 		    sizeof (valptr->zone_rctlval_priv)) != Z_OK)
4614 			break;
4615 		if (fetchprop(val, DTD_ATTR_LIMIT, valptr->zone_rctlval_limit,
4616 		    sizeof (valptr->zone_rctlval_limit)) != Z_OK)
4617 			break;
4618 		if (fetchprop(val, DTD_ATTR_ACTION, valptr->zone_rctlval_action,
4619 		    sizeof (valptr->zone_rctlval_action)) != Z_OK)
4620 			break;
4621 		if (zonecfg_add_rctl_value(tabptr, valptr) != Z_OK)
4622 			break;
4623 	}
4624 
4625 	handle->zone_dh_cur = cur->next;
4626 	return (Z_OK);
4627 }
4628 
4629 int
4630 zonecfg_endrctlent(zone_dochandle_t handle)
4631 {
4632 	return (zonecfg_endent(handle));
4633 }
4634 
4635 int
4636 zonecfg_setattrent(zone_dochandle_t handle)
4637 {
4638 	return (zonecfg_setent(handle));
4639 }
4640 
4641 int
4642 zonecfg_getattrent(zone_dochandle_t handle, struct zone_attrtab *tabptr)
4643 {
4644 	xmlNodePtr cur;
4645 	int err;
4646 
4647 	if (handle == NULL)
4648 		return (Z_INVAL);
4649 
4650 	if ((cur = handle->zone_dh_cur) == NULL)
4651 		return (Z_NO_ENTRY);
4652 
4653 	for (; cur != NULL; cur = cur->next)
4654 		if (!xmlStrcmp(cur->name, DTD_ELEM_ATTR))
4655 			break;
4656 	if (cur == NULL) {
4657 		handle->zone_dh_cur = handle->zone_dh_top;
4658 		return (Z_NO_ENTRY);
4659 	}
4660 
4661 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_attr_name,
4662 	    sizeof (tabptr->zone_attr_name))) != Z_OK) {
4663 		handle->zone_dh_cur = handle->zone_dh_top;
4664 		return (err);
4665 	}
4666 
4667 	if ((err = fetchprop(cur, DTD_ATTR_TYPE, tabptr->zone_attr_type,
4668 	    sizeof (tabptr->zone_attr_type))) != Z_OK) {
4669 		handle->zone_dh_cur = handle->zone_dh_top;
4670 		return (err);
4671 	}
4672 
4673 	if ((err = fetchprop(cur, DTD_ATTR_VALUE, tabptr->zone_attr_value,
4674 	    sizeof (tabptr->zone_attr_value))) != Z_OK) {
4675 		handle->zone_dh_cur = handle->zone_dh_top;
4676 		return (err);
4677 	}
4678 
4679 	handle->zone_dh_cur = cur->next;
4680 	return (Z_OK);
4681 }
4682 
4683 int
4684 zonecfg_endattrent(zone_dochandle_t handle)
4685 {
4686 	return (zonecfg_endent(handle));
4687 }
4688 
4689 /*
4690  * The privileges available on the system and described in privileges(5)
4691  * fall into four categories with respect to non-global zones:
4692  *
4693  *      Default set of privileges considered safe for all non-global
4694  *      zones.  These privileges are "safe" in the sense that a
4695  *      privileged process in the zone cannot affect processes in any
4696  *      other zone on the system.
4697  *
4698  *      Set of privileges not currently permitted within a non-global
4699  *      zone.  These privileges are considered by default, "unsafe,"
4700  *      and include ones which affect global resources (such as the
4701  *      system clock or physical memory) or are overly broad and cover
4702  *      more than one mechanism in the system.  In other cases, there
4703  *      has not been sufficient virtualization in the parts of the
4704  *      system the privilege covers to allow its use within a
4705  *      non-global zone.
4706  *
4707  *      Set of privileges required in order to get a zone booted and
4708  *      init(1M) started.  These cannot be removed from the zone's
4709  *      privilege set.
4710  *
4711  * All other privileges are optional and are potentially useful for
4712  * processes executing inside a non-global zone.
4713  *
4714  * When privileges are added to the system, a determination needs to be
4715  * made as to which category the privilege belongs to.  Ideally,
4716  * privileges should be fine-grained enough and the mechanisms they cover
4717  * virtualized enough so that they can be made available to non-global
4718  * zones.
4719  */
4720 
4721 /*
4722  * Define some of the tokens that priv_str_to_set(3C) recognizes.  Since
4723  * the privilege string separator can be any character, although it is
4724  * usually a comma character, define these here as well in the event that
4725  * they change or are augmented in the future.
4726  */
4727 #define	BASIC_TOKEN		"basic"
4728 #define	DEFAULT_TOKEN		"default"
4729 #define	ZONE_TOKEN		"zone"
4730 #define	TOKEN_PRIV_CHAR		','
4731 #define	TOKEN_PRIV_STR		","
4732 
4733 typedef struct priv_node {
4734 	struct priv_node	*pn_next;	/* Next privilege */
4735 	char			*pn_priv;	/* Privileges name */
4736 } priv_node_t;
4737 
4738 /* Privileges lists can differ across brands */
4739 typedef struct priv_lists {
4740 	/* Privileges considered safe for all non-global zones of a brand */
4741 	struct priv_node	*pl_default;
4742 
4743 	/* Privileges not permitted for all non-global zones of a brand */
4744 	struct priv_node	*pl_prohibited;
4745 
4746 	/* Privileges required for all non-global zones of a brand */
4747 	struct priv_node	*pl_required;
4748 
4749 	/*
4750 	 * ip-type of the zone these privileges lists apply to.
4751 	 * It is used to pass ip-type to the callback function,
4752 	 * priv_lists_cb, which has no way of getting the ip-type.
4753 	 */
4754 	const char		*pl_iptype;
4755 } priv_lists_t;
4756 
4757 static int
4758 priv_lists_cb(void *data, priv_iter_t *priv_iter)
4759 {
4760 	priv_lists_t *plp = (priv_lists_t *)data;
4761 	priv_node_t *pnp;
4762 
4763 	/* Skip this privilege if ip-type does not match */
4764 	if ((strcmp(priv_iter->pi_iptype, "all") != 0) &&
4765 	    (strcmp(priv_iter->pi_iptype, plp->pl_iptype) != 0))
4766 		return (0);
4767 
4768 	/* Allocate a new priv list node. */
4769 	if ((pnp = malloc(sizeof (*pnp))) == NULL)
4770 		return (-1);
4771 	if ((pnp->pn_priv = strdup(priv_iter->pi_name)) == NULL) {
4772 		free(pnp);
4773 		return (-1);
4774 	}
4775 
4776 	/* Insert the new priv list node into the right list */
4777 	if (strcmp(priv_iter->pi_set, "default") == 0) {
4778 		pnp->pn_next = plp->pl_default;
4779 		plp->pl_default = pnp;
4780 	} else if (strcmp(priv_iter->pi_set, "prohibited") == 0) {
4781 		pnp->pn_next = plp->pl_prohibited;
4782 		plp->pl_prohibited = pnp;
4783 	} else if (strcmp(priv_iter->pi_set, "required") == 0) {
4784 		pnp->pn_next = plp->pl_required;
4785 		plp->pl_required = pnp;
4786 	} else {
4787 		free(pnp->pn_priv);
4788 		free(pnp);
4789 		return (-1);
4790 	}
4791 	return (0);
4792 }
4793 
4794 static void
4795 priv_lists_destroy(priv_lists_t *plp)
4796 {
4797 	priv_node_t *pnp;
4798 
4799 	assert(plp != NULL);
4800 
4801 	while ((pnp = plp->pl_default) != NULL) {
4802 		plp->pl_default = pnp->pn_next;
4803 		free(pnp->pn_priv);
4804 		free(pnp);
4805 	}
4806 	while ((pnp = plp->pl_prohibited) != NULL) {
4807 		plp->pl_prohibited = pnp->pn_next;
4808 		free(pnp->pn_priv);
4809 		free(pnp);
4810 	}
4811 	while ((pnp = plp->pl_required) != NULL) {
4812 		plp->pl_required = pnp->pn_next;
4813 		free(pnp->pn_priv);
4814 		free(pnp);
4815 	}
4816 	free(plp);
4817 }
4818 
4819 static int
4820 priv_lists_create(zone_dochandle_t handle, priv_lists_t **plpp,
4821     const char *curr_iptype)
4822 {
4823 	priv_lists_t *plp;
4824 	brand_handle_t bh;
4825 	char brand[MAXNAMELEN];
4826 
4827 	if (handle != NULL) {
4828 		if (zonecfg_get_brand(handle, brand, sizeof (brand)) != 0)
4829 			return (Z_BRAND_ERROR);
4830 	} else {
4831 		(void) strlcpy(brand, NATIVE_BRAND_NAME, MAXNAMELEN);
4832 	}
4833 
4834 	if ((bh = brand_open(brand)) == NULL)
4835 		return (Z_BRAND_ERROR);
4836 
4837 	if ((plp = calloc(1, sizeof (priv_lists_t))) == NULL) {
4838 		brand_close(bh);
4839 		return (Z_NOMEM);
4840 	}
4841 
4842 	plp->pl_iptype = curr_iptype;
4843 
4844 	/* construct the privilege lists */
4845 	if (brand_config_iter_privilege(bh, priv_lists_cb, plp) != 0) {
4846 		priv_lists_destroy(plp);
4847 		brand_close(bh);
4848 		return (Z_BRAND_ERROR);
4849 	}
4850 
4851 	brand_close(bh);
4852 	*plpp = plp;
4853 	return (Z_OK);
4854 }
4855 
4856 static int
4857 get_default_privset(priv_set_t *privs, priv_lists_t *plp)
4858 {
4859 	priv_node_t *pnp;
4860 	priv_set_t *basic;
4861 
4862 	basic = priv_str_to_set(BASIC_TOKEN, TOKEN_PRIV_STR, NULL);
4863 	if (basic == NULL)
4864 		return (errno == ENOMEM ? Z_NOMEM : Z_INVAL);
4865 
4866 	priv_union(basic, privs);
4867 	priv_freeset(basic);
4868 
4869 	for (pnp = plp->pl_default; pnp != NULL; pnp = pnp->pn_next) {
4870 		if (priv_addset(privs, pnp->pn_priv) != 0)
4871 			return (Z_INVAL);
4872 	}
4873 
4874 	return (Z_OK);
4875 }
4876 
4877 int
4878 zonecfg_default_privset(priv_set_t *privs, const char *curr_iptype)
4879 {
4880 	priv_lists_t *plp;
4881 	int ret;
4882 
4883 	if ((ret = priv_lists_create(NULL, &plp, curr_iptype)) != Z_OK)
4884 		return (ret);
4885 	ret = get_default_privset(privs, plp);
4886 	priv_lists_destroy(plp);
4887 	return (ret);
4888 }
4889 
4890 void
4891 append_priv_token(char *priv, char *str, size_t strlen)
4892 {
4893 	if (*str != '\0')
4894 		(void) strlcat(str, TOKEN_PRIV_STR, strlen);
4895 	(void) strlcat(str, priv, strlen);
4896 }
4897 
4898 /*
4899  * Verify that the supplied string is a valid privilege limit set for a
4900  * non-global zone.  This string must not only be acceptable to
4901  * priv_str_to_set(3C) which parses it, but it also must resolve to a
4902  * privilege set that includes certain required privileges and lacks
4903  * certain prohibited privileges.
4904  */
4905 static int
4906 verify_privset(char *privbuf, priv_set_t *privs, char **privname,
4907     boolean_t add_default, priv_lists_t *plp)
4908 {
4909 	priv_node_t *pnp;
4910 	char *tmp, *cp, *lasts;
4911 	size_t len;
4912 	priv_set_t *mergeset;
4913 	const char *token;
4914 
4915 	/*
4916 	 * The verification of the privilege string occurs in several
4917 	 * phases.  In the first phase, the supplied string is scanned for
4918 	 * the ZONE_TOKEN token which is not support as part of the
4919 	 * "limitpriv" property.
4920 	 *
4921 	 * Duplicate the supplied privilege string since strtok_r(3C)
4922 	 * tokenizes its input by null-terminating the tokens.
4923 	 */
4924 	if ((tmp = strdup(privbuf)) == NULL)
4925 		return (Z_NOMEM);
4926 	for (cp = strtok_r(tmp, TOKEN_PRIV_STR, &lasts); cp != NULL;
4927 	    cp = strtok_r(NULL, TOKEN_PRIV_STR, &lasts)) {
4928 		if (strcmp(cp, ZONE_TOKEN) == 0) {
4929 			free(tmp);
4930 			if ((*privname = strdup(ZONE_TOKEN)) == NULL)
4931 				return (Z_NOMEM);
4932 			else
4933 				return (Z_PRIV_UNKNOWN);
4934 		}
4935 	}
4936 	free(tmp);
4937 
4938 	if (add_default) {
4939 		/*
4940 		 * If DEFAULT_TOKEN was specified, a string needs to be
4941 		 * built containing the privileges from the default, safe
4942 		 * set along with those of the "limitpriv" property.
4943 		 */
4944 		len = strlen(privbuf) + sizeof (BASIC_TOKEN) + 2;
4945 
4946 		for (pnp = plp->pl_default; pnp != NULL; pnp = pnp->pn_next)
4947 			len += strlen(pnp->pn_priv) + 1;
4948 		tmp = alloca(len);
4949 		*tmp = '\0';
4950 
4951 		append_priv_token(BASIC_TOKEN, tmp, len);
4952 		for (pnp = plp->pl_default; pnp != NULL; pnp = pnp->pn_next)
4953 			append_priv_token(pnp->pn_priv, tmp, len);
4954 		(void) strlcat(tmp, TOKEN_PRIV_STR, len);
4955 		(void) strlcat(tmp, privbuf, len);
4956 	} else {
4957 		tmp = privbuf;
4958 	}
4959 
4960 
4961 	/*
4962 	 * In the next phase, attempt to convert the merged privilege
4963 	 * string into a privilege set.  In the case of an error, either
4964 	 * there was a memory allocation failure or there was an invalid
4965 	 * privilege token in the string.  In either case, return an
4966 	 * appropriate error code but in the event of an invalid token,
4967 	 * allocate a string containing its name and return that back to
4968 	 * the caller.
4969 	 */
4970 	mergeset = priv_str_to_set(tmp, TOKEN_PRIV_STR, &token);
4971 	if (mergeset == NULL) {
4972 		if (token == NULL)
4973 			return (Z_NOMEM);
4974 		if ((cp = strchr(token, TOKEN_PRIV_CHAR)) != NULL)
4975 			*cp = '\0';
4976 		if ((*privname = strdup(token)) == NULL)
4977 			return (Z_NOMEM);
4978 		else
4979 			return (Z_PRIV_UNKNOWN);
4980 	}
4981 
4982 	/*
4983 	 * Next, verify that none of the prohibited zone privileges are
4984 	 * present in the merged privilege set.
4985 	 */
4986 	for (pnp = plp->pl_prohibited; pnp != NULL; pnp = pnp->pn_next) {
4987 		if (priv_ismember(mergeset, pnp->pn_priv)) {
4988 			priv_freeset(mergeset);
4989 			if ((*privname = strdup(pnp->pn_priv)) == NULL)
4990 				return (Z_NOMEM);
4991 			else
4992 				return (Z_PRIV_PROHIBITED);
4993 		}
4994 	}
4995 
4996 	/*
4997 	 * Finally, verify that all of the required zone privileges are
4998 	 * present in the merged privilege set.
4999 	 */
5000 	for (pnp = plp->pl_required; pnp != NULL; pnp = pnp->pn_next) {
5001 		if (!priv_ismember(mergeset, pnp->pn_priv)) {
5002 			priv_freeset(mergeset);
5003 			if ((*privname = strdup(pnp->pn_priv)) == NULL)
5004 				return (Z_NOMEM);
5005 			else
5006 				return (Z_PRIV_REQUIRED);
5007 		}
5008 	}
5009 
5010 	priv_copyset(mergeset, privs);
5011 	priv_freeset(mergeset);
5012 	return (Z_OK);
5013 }
5014 
5015 /*
5016  * Fill in the supplied privilege set with either the default, safe set of
5017  * privileges suitable for a non-global zone, or one based on the
5018  * "limitpriv" property in the zone's configuration.
5019  *
5020  * In the event of an invalid privilege specification in the
5021  * configuration, a string is allocated and returned containing the
5022  * "privilege" causing the issue.  It is the caller's responsibility to
5023  * free this memory when it is done with it.
5024  */
5025 int
5026 zonecfg_get_privset(zone_dochandle_t handle, priv_set_t *privs,
5027     char **privname)
5028 {
5029 	priv_lists_t *plp;
5030 	char *cp, *limitpriv = NULL;
5031 	int err, limitlen;
5032 	zone_iptype_t iptype;
5033 	const char *curr_iptype;
5034 
5035 	/*
5036 	 * Attempt to lookup the "limitpriv" property.  If it does not
5037 	 * exist or matches the string DEFAULT_TOKEN exactly, then the
5038 	 * default, safe privilege set is returned.
5039 	 */
5040 	if ((err = zonecfg_get_limitpriv(handle, &limitpriv)) != Z_OK)
5041 		return (err);
5042 
5043 	if ((err = zonecfg_get_iptype(handle, &iptype)) != Z_OK)
5044 		return (err);
5045 
5046 	switch (iptype) {
5047 	case ZS_SHARED:
5048 		curr_iptype = "shared";
5049 		break;
5050 	case ZS_EXCLUSIVE:
5051 		curr_iptype = "exclusive";
5052 		break;
5053 	}
5054 
5055 	if ((err = priv_lists_create(handle, &plp, curr_iptype)) != Z_OK)
5056 		return (err);
5057 
5058 	limitlen = strlen(limitpriv);
5059 	if (limitlen == 0 || strcmp(limitpriv, DEFAULT_TOKEN) == 0) {
5060 		free(limitpriv);
5061 		err = get_default_privset(privs, plp);
5062 		priv_lists_destroy(plp);
5063 		return (err);
5064 	}
5065 
5066 	/*
5067 	 * Check if the string DEFAULT_TOKEN is the first token in a list
5068 	 * of privileges.
5069 	 */
5070 	cp = strchr(limitpriv, TOKEN_PRIV_CHAR);
5071 	if (cp != NULL &&
5072 	    strncmp(limitpriv, DEFAULT_TOKEN, cp - limitpriv) == 0)
5073 		err = verify_privset(cp + 1, privs, privname, B_TRUE, plp);
5074 	else
5075 		err = verify_privset(limitpriv, privs, privname, B_FALSE, plp);
5076 
5077 	free(limitpriv);
5078 	priv_lists_destroy(plp);
5079 	return (err);
5080 }
5081 
5082 int
5083 zone_get_zonepath(char *zone_name, char *zonepath, size_t rp_sz)
5084 {
5085 	zone_dochandle_t handle;
5086 	boolean_t found = B_FALSE;
5087 	struct zoneent *ze;
5088 	FILE *cookie;
5089 	int err;
5090 	char *cp;
5091 
5092 	if (zone_name == NULL)
5093 		return (Z_INVAL);
5094 
5095 	(void) strlcpy(zonepath, zonecfg_root, rp_sz);
5096 	cp = zonepath + strlen(zonepath);
5097 	while (cp > zonepath && cp[-1] == '/')
5098 		*--cp = '\0';
5099 
5100 	if (strcmp(zone_name, GLOBAL_ZONENAME) == 0) {
5101 		if (zonepath[0] == '\0')
5102 			(void) strlcpy(zonepath, "/", rp_sz);
5103 		return (Z_OK);
5104 	}
5105 
5106 	/*
5107 	 * First check the index file.  Because older versions did not have
5108 	 * a copy of the zone path, allow for it to be zero length, in which
5109 	 * case we ignore this result and fall back to the XML files.
5110 	 */
5111 	cookie = setzoneent();
5112 	while ((ze = getzoneent_private(cookie)) != NULL) {
5113 		if (strcmp(ze->zone_name, zone_name) == 0) {
5114 			found = B_TRUE;
5115 			if (ze->zone_path[0] != '\0')
5116 				(void) strlcpy(cp, ze->zone_path,
5117 				    rp_sz - (cp - zonepath));
5118 		}
5119 		free(ze);
5120 		if (found)
5121 			break;
5122 	}
5123 	endzoneent(cookie);
5124 	if (found && *cp != '\0')
5125 		return (Z_OK);
5126 
5127 	/* Fall back to the XML files. */
5128 	if ((handle = zonecfg_init_handle()) == NULL)
5129 		return (Z_NOMEM);
5130 
5131 	/*
5132 	 * Check the snapshot first: if a zone is running, its zonepath
5133 	 * may have changed.
5134 	 */
5135 	if (zonecfg_get_snapshot_handle(zone_name, handle) != Z_OK) {
5136 		if ((err = zonecfg_get_handle(zone_name, handle)) != Z_OK) {
5137 			zonecfg_fini_handle(handle);
5138 			return (err);
5139 		}
5140 	}
5141 	err = zonecfg_get_zonepath(handle, zonepath, rp_sz);
5142 	zonecfg_fini_handle(handle);
5143 	return (err);
5144 }
5145 
5146 int
5147 zone_get_rootpath(char *zone_name, char *rootpath, size_t rp_sz)
5148 {
5149 	int err;
5150 
5151 	/* This function makes sense for non-global zones only. */
5152 	if (strcmp(zone_name, GLOBAL_ZONENAME) == 0)
5153 		return (Z_BOGUS_ZONE_NAME);
5154 	if ((err = zone_get_zonepath(zone_name, rootpath, rp_sz)) != Z_OK)
5155 		return (err);
5156 	if (strlcat(rootpath, "/root", rp_sz) >= rp_sz)
5157 		return (Z_TOO_BIG);
5158 	return (Z_OK);
5159 }
5160 
5161 int
5162 zone_get_brand(char *zone_name, char *brandname, size_t rp_sz)
5163 {
5164 	int err;
5165 	zone_dochandle_t handle;
5166 	char myzone[MAXNAMELEN];
5167 	int myzoneid = getzoneid();
5168 
5169 	/*
5170 	 * If we are not in the global zone, then we don't have the zone
5171 	 * .xml files with the brand name available.  Thus, we are going to
5172 	 * have to ask the kernel for the information.
5173 	 */
5174 	if (myzoneid != GLOBAL_ZONEID) {
5175 		if (is_system_labeled()) {
5176 			(void) strlcpy(brandname, NATIVE_BRAND_NAME, rp_sz);
5177 			return (Z_OK);
5178 		}
5179 		if (zone_getattr(myzoneid, ZONE_ATTR_NAME, myzone,
5180 		    sizeof (myzone)) < 0)
5181 			return (Z_NO_ZONE);
5182 		if (strncmp(zone_name, myzone, MAXNAMELEN) != NULL)
5183 			return (Z_NO_ZONE);
5184 		err = zone_getattr(myzoneid, ZONE_ATTR_BRAND, brandname, rp_sz);
5185 		if (err < 0)
5186 			return ((errno == EFAULT) ? Z_TOO_BIG : Z_INVAL);
5187 		return (Z_OK);
5188 	}
5189 
5190 	if (strcmp(zone_name, "global") == NULL) {
5191 		(void) strlcpy(brandname, NATIVE_BRAND_NAME, rp_sz);
5192 		return (Z_OK);
5193 	}
5194 	if ((handle = zonecfg_init_handle()) == NULL)
5195 		return (Z_NOMEM);
5196 
5197 	err = zonecfg_get_handle((char *)zone_name, handle);
5198 	if (err == Z_OK)
5199 		err = zonecfg_get_brand(handle, brandname, rp_sz);
5200 
5201 	zonecfg_fini_handle(handle);
5202 	return (err);
5203 }
5204 
5205 /*
5206  * Return the appropriate root for the active /dev.
5207  * For normal zone, the path is $ZONEPATH/root;
5208  * for scratch zone, the dev path is $ZONEPATH/lu.
5209  */
5210 int
5211 zone_get_devroot(char *zone_name, char *devroot, size_t rp_sz)
5212 {
5213 	int err;
5214 	char *suffix;
5215 	zone_state_t state;
5216 
5217 	/* This function makes sense for non-global zones only. */
5218 	if (strcmp(zone_name, GLOBAL_ZONENAME) == 0)
5219 		return (Z_BOGUS_ZONE_NAME);
5220 	if ((err = zone_get_zonepath(zone_name, devroot, rp_sz)) != Z_OK)
5221 		return (err);
5222 
5223 	if (zone_get_state(zone_name, &state) == Z_OK &&
5224 	    state == ZONE_STATE_MOUNTED)
5225 		suffix = "/lu";
5226 	else
5227 		suffix = "/root";
5228 	if (strlcat(devroot, suffix, rp_sz) >= rp_sz)
5229 		return (Z_TOO_BIG);
5230 	return (Z_OK);
5231 }
5232 
5233 static zone_state_t
5234 kernel_state_to_user_state(zoneid_t zoneid, zone_status_t kernel_state)
5235 {
5236 	char zoneroot[MAXPATHLEN];
5237 	size_t zlen;
5238 
5239 	assert(kernel_state <= ZONE_MAX_STATE);
5240 	switch (kernel_state) {
5241 		case ZONE_IS_UNINITIALIZED:
5242 		case ZONE_IS_INITIALIZED:
5243 			/* The kernel will not return these two states */
5244 			return (ZONE_STATE_READY);
5245 		case ZONE_IS_READY:
5246 			/*
5247 			 * If the zone's root is mounted on $ZONEPATH/lu, then
5248 			 * it's a mounted scratch zone.
5249 			 */
5250 			if (zone_getattr(zoneid, ZONE_ATTR_ROOT, zoneroot,
5251 			    sizeof (zoneroot)) >= 0) {
5252 				zlen = strlen(zoneroot);
5253 				if (zlen > 3 &&
5254 				    strcmp(zoneroot + zlen - 3, "/lu") == 0)
5255 					return (ZONE_STATE_MOUNTED);
5256 			}
5257 			return (ZONE_STATE_READY);
5258 		case ZONE_IS_BOOTING:
5259 		case ZONE_IS_RUNNING:
5260 			return (ZONE_STATE_RUNNING);
5261 		case ZONE_IS_SHUTTING_DOWN:
5262 		case ZONE_IS_EMPTY:
5263 			return (ZONE_STATE_SHUTTING_DOWN);
5264 		case ZONE_IS_DOWN:
5265 		case ZONE_IS_DYING:
5266 		case ZONE_IS_DEAD:
5267 		default:
5268 			return (ZONE_STATE_DOWN);
5269 	}
5270 	/* NOTREACHED */
5271 }
5272 
5273 int
5274 zone_get_state(char *zone_name, zone_state_t *state_num)
5275 {
5276 	zone_status_t status;
5277 	zoneid_t zone_id;
5278 	struct zoneent *ze;
5279 	boolean_t found = B_FALSE;
5280 	FILE *cookie;
5281 	char kernzone[ZONENAME_MAX];
5282 	FILE *fp;
5283 
5284 	if (zone_name == NULL)
5285 		return (Z_INVAL);
5286 
5287 	/*
5288 	 * If we're looking at an alternate root, then we need to query the
5289 	 * kernel using the scratch zone name.
5290 	 */
5291 	zone_id = -1;
5292 	if (*zonecfg_root != '\0' && !zonecfg_is_scratch(zone_name)) {
5293 		if ((fp = zonecfg_open_scratch("", B_FALSE)) != NULL) {
5294 			if (zonecfg_find_scratch(fp, zone_name, zonecfg_root,
5295 			    kernzone, sizeof (kernzone)) == 0)
5296 				zone_id = getzoneidbyname(kernzone);
5297 			zonecfg_close_scratch(fp);
5298 		}
5299 	} else {
5300 		zone_id = getzoneidbyname(zone_name);
5301 	}
5302 
5303 	/* check to see if zone is running */
5304 	if (zone_id != -1 &&
5305 	    zone_getattr(zone_id, ZONE_ATTR_STATUS, &status,
5306 	    sizeof (status)) >= 0) {
5307 		*state_num = kernel_state_to_user_state(zone_id, status);
5308 		return (Z_OK);
5309 	}
5310 
5311 	cookie = setzoneent();
5312 	while ((ze = getzoneent_private(cookie)) != NULL) {
5313 		if (strcmp(ze->zone_name, zone_name) == 0) {
5314 			found = B_TRUE;
5315 			*state_num = ze->zone_state;
5316 		}
5317 		free(ze);
5318 		if (found)
5319 			break;
5320 	}
5321 	endzoneent(cookie);
5322 	return ((found) ? Z_OK : Z_NO_ZONE);
5323 }
5324 
5325 int
5326 zone_set_state(char *zone, zone_state_t state)
5327 {
5328 	struct zoneent ze;
5329 
5330 	if (state != ZONE_STATE_CONFIGURED && state != ZONE_STATE_INSTALLED &&
5331 	    state != ZONE_STATE_INCOMPLETE)
5332 		return (Z_INVAL);
5333 
5334 	bzero(&ze, sizeof (ze));
5335 	(void) strlcpy(ze.zone_name, zone, sizeof (ze.zone_name));
5336 	ze.zone_state = state;
5337 	(void) strlcpy(ze.zone_path, "", sizeof (ze.zone_path));
5338 	return (putzoneent(&ze, PZE_MODIFY));
5339 }
5340 
5341 /*
5342  * Get id (if any) for specified zone.  There are four possible outcomes:
5343  * - If the string corresponds to the numeric id of an active (booted)
5344  *   zone, sets *zip to the zone id and returns 0.
5345  * - If the string corresponds to the name of an active (booted) zone,
5346  *   sets *zip to the zone id and returns 0.
5347  * - If the string is a name in the configuration but is not booted,
5348  *   sets *zip to ZONE_ID_UNDEFINED and returns 0.
5349  * - Otherwise, leaves *zip unchanged and returns -1.
5350  *
5351  * This function acts as an auxiliary filter on the function of the same
5352  * name in libc; the linker binds to this version if libzonecfg exists,
5353  * and the libc version if it doesn't.  Any changes to this version of
5354  * the function should probably be reflected in the libc version as well.
5355  */
5356 int
5357 zone_get_id(const char *str, zoneid_t *zip)
5358 {
5359 	zone_dochandle_t hdl;
5360 	zoneid_t zoneid;
5361 	char *cp;
5362 	int err;
5363 
5364 	/* first try looking for active zone by id */
5365 	errno = 0;
5366 	zoneid = (zoneid_t)strtol(str, &cp, 0);
5367 	if (errno == 0 && cp != str && *cp == '\0' &&
5368 	    getzonenamebyid(zoneid, NULL, 0) != -1) {
5369 		*zip = zoneid;
5370 		return (0);
5371 	}
5372 
5373 	/* then look for active zone by name */
5374 	if ((zoneid = getzoneidbyname(str)) != -1) {
5375 		*zip = zoneid;
5376 		return (0);
5377 	}
5378 
5379 	/* if in global zone, try looking up name in configuration database */
5380 	if (getzoneid() != GLOBAL_ZONEID ||
5381 	    (hdl = zonecfg_init_handle()) == NULL)
5382 		return (-1);
5383 
5384 	if (zonecfg_get_handle(str, hdl) == Z_OK) {
5385 		/* zone exists but isn't active */
5386 		*zip = ZONE_ID_UNDEFINED;
5387 		err = 0;
5388 	} else {
5389 		err = -1;
5390 	}
5391 
5392 	zonecfg_fini_handle(hdl);
5393 	return (err);
5394 }
5395 
5396 char *
5397 zone_state_str(zone_state_t state_num)
5398 {
5399 	switch (state_num) {
5400 	case ZONE_STATE_CONFIGURED:
5401 		return (ZONE_STATE_STR_CONFIGURED);
5402 	case ZONE_STATE_INCOMPLETE:
5403 		return (ZONE_STATE_STR_INCOMPLETE);
5404 	case ZONE_STATE_INSTALLED:
5405 		return (ZONE_STATE_STR_INSTALLED);
5406 	case ZONE_STATE_READY:
5407 		return (ZONE_STATE_STR_READY);
5408 	case ZONE_STATE_MOUNTED:
5409 		return (ZONE_STATE_STR_MOUNTED);
5410 	case ZONE_STATE_RUNNING:
5411 		return (ZONE_STATE_STR_RUNNING);
5412 	case ZONE_STATE_SHUTTING_DOWN:
5413 		return (ZONE_STATE_STR_SHUTTING_DOWN);
5414 	case ZONE_STATE_DOWN:
5415 		return (ZONE_STATE_STR_DOWN);
5416 	default:
5417 		return ("unknown");
5418 	}
5419 }
5420 
5421 /*
5422  * Given a UUID value, find an associated zone name.  This is intended to be
5423  * used by callers who set up some 'default' name (corresponding to the
5424  * expected name for the zone) in the zonename buffer, and thus the function
5425  * doesn't touch this buffer on failure.
5426  */
5427 int
5428 zonecfg_get_name_by_uuid(const uuid_t uuidin, char *zonename, size_t namelen)
5429 {
5430 	FILE *fp;
5431 	struct zoneent *ze;
5432 	uchar_t *uuid;
5433 
5434 	/*
5435 	 * A small amount of subterfuge via casts is necessary here because
5436 	 * libuuid doesn't use const correctly, but we don't want to export
5437 	 * this brokenness to our clients.
5438 	 */
5439 	uuid = (uchar_t *)uuidin;
5440 	if (uuid_is_null(uuid))
5441 		return (Z_NO_ZONE);
5442 	if ((fp = setzoneent()) == NULL)
5443 		return (Z_NO_ZONE);
5444 	while ((ze = getzoneent_private(fp)) != NULL) {
5445 		if (uuid_compare(uuid, ze->zone_uuid) == 0)
5446 			break;
5447 		free(ze);
5448 	}
5449 	endzoneent(fp);
5450 	if (ze != NULL) {
5451 		(void) strlcpy(zonename, ze->zone_name, namelen);
5452 		free(ze);
5453 		return (Z_OK);
5454 	} else {
5455 		return (Z_NO_ZONE);
5456 	}
5457 }
5458 
5459 /*
5460  * Given a zone name, get its UUID.  Returns a "NULL" UUID value if the zone
5461  * exists but the file doesn't have a value set yet.  Returns an error if the
5462  * zone cannot be located.
5463  */
5464 int
5465 zonecfg_get_uuid(const char *zonename, uuid_t uuid)
5466 {
5467 	FILE *fp;
5468 	struct zoneent *ze;
5469 
5470 	if ((fp = setzoneent()) == NULL)
5471 		return (Z_NO_ZONE);
5472 	while ((ze = getzoneent_private(fp)) != NULL) {
5473 		if (strcmp(ze->zone_name, zonename) == 0)
5474 			break;
5475 		free(ze);
5476 	}
5477 	endzoneent(fp);
5478 	if (ze != NULL) {
5479 		uuid_copy(uuid, ze->zone_uuid);
5480 		free(ze);
5481 		return (Z_OK);
5482 	} else {
5483 		return (Z_NO_ZONE);
5484 	}
5485 }
5486 
5487 /*
5488  * File-system convenience functions.
5489  */
5490 boolean_t
5491 zonecfg_valid_fs_type(const char *type)
5492 {
5493 	/*
5494 	 * We already know which FS types don't work.
5495 	 */
5496 	if (strcmp(type, "proc") == 0 ||
5497 	    strcmp(type, "mntfs") == 0 ||
5498 	    strcmp(type, "autofs") == 0 ||
5499 	    strncmp(type, "nfs", sizeof ("nfs") - 1) == 0 ||
5500 	    strcmp(type, "cachefs") == 0)
5501 		return (B_FALSE);
5502 	/*
5503 	 * The caller may do more detailed verification to make sure other
5504 	 * aspects of this filesystem type make sense.
5505 	 */
5506 	return (B_TRUE);
5507 }
5508 
5509 /*
5510  * Generally uninteresting rctl convenience functions.
5511  */
5512 
5513 int
5514 zonecfg_construct_rctlblk(const struct zone_rctlvaltab *rctlval,
5515     rctlblk_t *rctlblk)
5516 {
5517 	unsigned long long ull;
5518 	char *endp;
5519 	rctl_priv_t priv;
5520 	rctl_qty_t limit;
5521 	uint_t action;
5522 
5523 	/* Get the privilege */
5524 	if (strcmp(rctlval->zone_rctlval_priv, "basic") == 0) {
5525 		priv = RCPRIV_BASIC;
5526 	} else if (strcmp(rctlval->zone_rctlval_priv, "privileged") == 0) {
5527 		priv = RCPRIV_PRIVILEGED;
5528 	} else {
5529 		/* Invalid privilege */
5530 		return (Z_INVAL);
5531 	}
5532 
5533 	/* deal with negative input; strtoull(3c) doesn't do what we want */
5534 	if (rctlval->zone_rctlval_limit[0] == '-')
5535 		return (Z_INVAL);
5536 	/* Get the limit */
5537 	errno = 0;
5538 	ull = strtoull(rctlval->zone_rctlval_limit, &endp, 0);
5539 	if (errno != 0 || *endp != '\0') {
5540 		/* parse failed */
5541 		return (Z_INVAL);
5542 	}
5543 	limit = (rctl_qty_t)ull;
5544 
5545 	/* Get the action */
5546 	if (strcmp(rctlval->zone_rctlval_action, "none") == 0) {
5547 		action = RCTL_LOCAL_NOACTION;
5548 	} else if (strcmp(rctlval->zone_rctlval_action, "signal") == 0) {
5549 		action = RCTL_LOCAL_SIGNAL;
5550 	} else if (strcmp(rctlval->zone_rctlval_action, "deny") == 0) {
5551 		action = RCTL_LOCAL_DENY;
5552 	} else {
5553 		/* Invalid Action */
5554 		return (Z_INVAL);
5555 	}
5556 	rctlblk_set_local_action(rctlblk, action, 0);
5557 	rctlblk_set_privilege(rctlblk, priv);
5558 	rctlblk_set_value(rctlblk, limit);
5559 	return (Z_OK);
5560 }
5561 
5562 static int
5563 rctl_check(const char *rctlname, void *arg)
5564 {
5565 	const char *attrname = arg;
5566 
5567 	/*
5568 	 * Returning 1 here is our signal to zonecfg_is_rctl() that it is
5569 	 * indeed an rctl name recognized by the system.
5570 	 */
5571 	return (strcmp(rctlname, attrname) == 0 ? 1 : 0);
5572 }
5573 
5574 boolean_t
5575 zonecfg_is_rctl(const char *name)
5576 {
5577 	return (rctl_walk(rctl_check, (void *)name) == 1);
5578 }
5579 
5580 boolean_t
5581 zonecfg_valid_rctlname(const char *name)
5582 {
5583 	const char *c;
5584 
5585 	if (strncmp(name, "zone.", sizeof ("zone.") - 1) != 0)
5586 		return (B_FALSE);
5587 	if (strlen(name) == sizeof ("zone.") - 1)
5588 		return (B_FALSE);
5589 	for (c = name + sizeof ("zone.") - 1; *c != '\0'; c++) {
5590 		if (!isalpha(*c) && *c != '-')
5591 			return (B_FALSE);
5592 	}
5593 	return (B_TRUE);
5594 }
5595 
5596 boolean_t
5597 zonecfg_valid_rctlblk(const rctlblk_t *rctlblk)
5598 {
5599 	rctl_priv_t priv = rctlblk_get_privilege((rctlblk_t *)rctlblk);
5600 	uint_t action = rctlblk_get_local_action((rctlblk_t *)rctlblk, NULL);
5601 
5602 	if (priv != RCPRIV_PRIVILEGED)
5603 		return (B_FALSE);
5604 	if (action != RCTL_LOCAL_NOACTION && action != RCTL_LOCAL_DENY)
5605 		return (B_FALSE);
5606 	return (B_TRUE);
5607 }
5608 
5609 boolean_t
5610 zonecfg_valid_rctl(const char *name, const rctlblk_t *rctlblk)
5611 {
5612 	rctlblk_t *current, *next;
5613 	rctl_qty_t limit = rctlblk_get_value((rctlblk_t *)rctlblk);
5614 	uint_t action = rctlblk_get_local_action((rctlblk_t *)rctlblk, NULL);
5615 	uint_t global_flags;
5616 
5617 	if (!zonecfg_valid_rctlblk(rctlblk))
5618 		return (B_FALSE);
5619 	if (!zonecfg_valid_rctlname(name))
5620 		return (B_FALSE);
5621 
5622 	current = alloca(rctlblk_size());
5623 	if (getrctl(name, NULL, current, RCTL_FIRST) != 0)
5624 		return (B_TRUE);	/* not an rctl on this system */
5625 	/*
5626 	 * Make sure the proposed value isn't greater than the current system
5627 	 * value.
5628 	 */
5629 	next = alloca(rctlblk_size());
5630 	while (rctlblk_get_privilege(current) != RCPRIV_SYSTEM) {
5631 		rctlblk_t *tmp;
5632 
5633 		if (getrctl(name, current, next, RCTL_NEXT) != 0)
5634 			return (B_FALSE);	/* shouldn't happen */
5635 		tmp = current;
5636 		current = next;
5637 		next = tmp;
5638 	}
5639 	if (limit > rctlblk_get_value(current))
5640 		return (B_FALSE);
5641 
5642 	/*
5643 	 * Make sure the proposed action is allowed.
5644 	 */
5645 	global_flags = rctlblk_get_global_flags(current);
5646 	if ((global_flags & RCTL_GLOBAL_DENY_NEVER) &&
5647 	    action == RCTL_LOCAL_DENY)
5648 		return (B_FALSE);
5649 	if ((global_flags & RCTL_GLOBAL_DENY_ALWAYS) &&
5650 	    action == RCTL_LOCAL_NOACTION)
5651 		return (B_FALSE);
5652 
5653 	return (B_TRUE);
5654 }
5655 
5656 /*
5657  * There is always a race condition between reading the initial copy of
5658  * a zones state and its state changing.  We address this by providing
5659  * zonecfg_notify_critical_enter and zonecfg_noticy_critical_exit functions.
5660  * When zonecfg_critical_enter is called, sets the state field to LOCKED
5661  * and aquires biglock. Biglock protects against other threads executing
5662  * critical_enter and the state field protects against state changes during
5663  * the critical period.
5664  *
5665  * If any state changes occur, zn_cb will set the failed field of the znotify
5666  * structure.  This will cause the critical_exit function to re-lock the
5667  * channel and return an error. Since evsnts may be delayed, the critical_exit
5668  * function "flushes" the queue by putting an event on the queue and waiting for
5669  * zn_cb to notify critical_exit that it received the ping event.
5670  */
5671 static const char *
5672 string_get_tok(const char *in, char delim, int num)
5673 {
5674 	int i = 0;
5675 
5676 	for (; i < num; in++) {
5677 		if (*in == delim)
5678 			i++;
5679 		if (*in == 0)
5680 			return (NULL);
5681 	}
5682 	return (in);
5683 }
5684 
5685 static boolean_t
5686 is_ping(sysevent_t *ev)
5687 {
5688 	if (strcmp(sysevent_get_subclass_name(ev),
5689 	    ZONE_EVENT_PING_SUBCLASS) == 0) {
5690 		return (B_TRUE);
5691 	} else {
5692 		return (B_FALSE);
5693 	}
5694 }
5695 
5696 static boolean_t
5697 is_my_ping(sysevent_t *ev)
5698 {
5699 	const char *sender;
5700 	char mypid[sizeof (pid_t) * 3 + 1];
5701 
5702 	(void) snprintf(mypid, sizeof (mypid), "%i", getpid());
5703 	sender = string_get_tok(sysevent_get_pub(ev), ':', 3);
5704 	if (sender == NULL)
5705 		return (B_FALSE);
5706 	if (strcmp(sender, mypid) != 0)
5707 		return (B_FALSE);
5708 	return (B_TRUE);
5709 }
5710 
5711 static int
5712 do_callback(struct znotify *zevtchan, sysevent_t *ev)
5713 {
5714 	nvlist_t *l;
5715 	int zid;
5716 	char *zonename;
5717 	char *newstate;
5718 	char *oldstate;
5719 	int ret;
5720 	hrtime_t when;
5721 
5722 	if (strcmp(sysevent_get_subclass_name(ev),
5723 	    ZONE_EVENT_STATUS_SUBCLASS) == 0) {
5724 
5725 		if (sysevent_get_attr_list(ev, &l) != 0) {
5726 			if (errno == ENOMEM) {
5727 				zevtchan->zn_failure_count++;
5728 				return (EAGAIN);
5729 			}
5730 			return (0);
5731 		}
5732 		ret = 0;
5733 
5734 		if ((nvlist_lookup_string(l, ZONE_CB_NAME, &zonename) == 0) &&
5735 		    (nvlist_lookup_string(l, ZONE_CB_NEWSTATE, &newstate)
5736 		    == 0) &&
5737 		    (nvlist_lookup_string(l, ZONE_CB_OLDSTATE, &oldstate)
5738 		    == 0) &&
5739 		    (nvlist_lookup_uint64(l, ZONE_CB_TIMESTAMP,
5740 		    (uint64_t *)&when) == 0) &&
5741 		    (nvlist_lookup_int32(l, ZONE_CB_ZONEID, &zid) == 0)) {
5742 			ret = zevtchan->zn_callback(zonename, zid, newstate,
5743 			    oldstate, when, zevtchan->zn_private);
5744 		}
5745 
5746 		zevtchan->zn_failure_count = 0;
5747 		nvlist_free(l);
5748 		return (ret);
5749 	} else {
5750 		/*
5751 		 * We have received an event in an unknown subclass. Ignore.
5752 		 */
5753 		zevtchan->zn_failure_count = 0;
5754 		return (0);
5755 	}
5756 }
5757 
5758 static int
5759 zn_cb(sysevent_t *ev, void *p)
5760 {
5761 	struct znotify *zevtchan = p;
5762 	int error;
5763 
5764 	(void) pthread_mutex_lock(&(zevtchan->zn_mutex));
5765 
5766 	if (is_ping(ev) && !is_my_ping(ev)) {
5767 		(void) pthread_mutex_unlock((&zevtchan->zn_mutex));
5768 		return (0);
5769 	}
5770 
5771 	if (zevtchan->zn_state == ZN_LOCKED) {
5772 		assert(!is_ping(ev));
5773 		zevtchan->zn_failed = B_TRUE;
5774 		(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5775 		return (0);
5776 	}
5777 
5778 	if (zevtchan->zn_state == ZN_PING_INFLIGHT) {
5779 		if (is_ping(ev)) {
5780 			zevtchan->zn_state = ZN_PING_RECEIVED;
5781 			(void) pthread_cond_signal(&(zevtchan->zn_cond));
5782 			(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5783 			return (0);
5784 		} else {
5785 			zevtchan->zn_failed = B_TRUE;
5786 			(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5787 			return (0);
5788 		}
5789 	}
5790 
5791 	if (zevtchan->zn_state == ZN_UNLOCKED) {
5792 
5793 		error = do_callback(zevtchan, ev);
5794 		(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5795 		/*
5796 		 * Every ENOMEM failure causes do_callback to increment
5797 		 * zn_failure_count and every success causes it to
5798 		 * set zn_failure_count to zero.  If we got EAGAIN,
5799 		 * we will sleep for zn_failure_count seconds and return
5800 		 * EAGAIN to gpec to try again.
5801 		 *
5802 		 * After 55 seconds, or 10 try's we give up and drop the
5803 		 * event.
5804 		 */
5805 		if (error == EAGAIN) {
5806 			if (zevtchan->zn_failure_count > ZONE_CB_RETRY_COUNT) {
5807 				return (0);
5808 			}
5809 			(void) sleep(zevtchan->zn_failure_count);
5810 		}
5811 		return (error);
5812 	}
5813 
5814 	if (zevtchan->zn_state == ZN_PING_RECEIVED) {
5815 		(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5816 		return (0);
5817 	}
5818 
5819 	abort();
5820 	return (0);
5821 }
5822 
5823 void
5824 zonecfg_notify_critical_enter(void *h)
5825 {
5826 	struct znotify *zevtchan = h;
5827 
5828 	(void) pthread_mutex_lock(&(zevtchan->zn_bigmutex));
5829 	zevtchan->zn_state = ZN_LOCKED;
5830 }
5831 
5832 int
5833 zonecfg_notify_critical_exit(void * h)
5834 {
5835 
5836 	struct znotify *zevtchan = h;
5837 
5838 	if (zevtchan->zn_state == ZN_UNLOCKED)
5839 		return (0);
5840 
5841 	(void) pthread_mutex_lock(&(zevtchan->zn_mutex));
5842 	zevtchan->zn_state = ZN_PING_INFLIGHT;
5843 
5844 	(void) sysevent_evc_publish(zevtchan->zn_eventchan,
5845 	    ZONE_EVENT_STATUS_CLASS,
5846 	    ZONE_EVENT_PING_SUBCLASS, ZONE_EVENT_PING_PUBLISHER,
5847 	    zevtchan->zn_subscriber_id, NULL, EVCH_SLEEP);
5848 
5849 	while (zevtchan->zn_state != ZN_PING_RECEIVED) {
5850 		(void) pthread_cond_wait(&(zevtchan->zn_cond),
5851 		    &(zevtchan->zn_mutex));
5852 	}
5853 
5854 	if (zevtchan->zn_failed == B_TRUE) {
5855 		zevtchan->zn_state = ZN_LOCKED;
5856 		zevtchan->zn_failed = B_FALSE;
5857 		(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5858 		return (1);
5859 	}
5860 
5861 	zevtchan->zn_state = ZN_UNLOCKED;
5862 	(void) pthread_mutex_unlock(&(zevtchan->zn_mutex));
5863 	(void) pthread_mutex_unlock(&(zevtchan->zn_bigmutex));
5864 	return (0);
5865 }
5866 
5867 void
5868 zonecfg_notify_critical_abort(void *h)
5869 {
5870 	struct znotify *zevtchan = h;
5871 
5872 	zevtchan->zn_state = ZN_UNLOCKED;
5873 	zevtchan->zn_failed = B_FALSE;
5874 	/*
5875 	 * Don't do anything about zn_lock. If it is held, it could only be
5876 	 * held by zn_cb and it will be unlocked soon.
5877 	 */
5878 	(void) pthread_mutex_unlock(&(zevtchan->zn_bigmutex));
5879 }
5880 
5881 void *
5882 zonecfg_notify_bind(int(*func)(const char *zonename, zoneid_t zid,
5883     const char *newstate, const char *oldstate, hrtime_t when, void *p),
5884     void *p)
5885 {
5886 	struct znotify *zevtchan;
5887 	int i = 1;
5888 	int r;
5889 
5890 	zevtchan = malloc(sizeof (struct znotify));
5891 
5892 	if (zevtchan == NULL)
5893 		return (NULL);
5894 
5895 	zevtchan->zn_private = p;
5896 	zevtchan->zn_callback = func;
5897 	zevtchan->zn_state = ZN_UNLOCKED;
5898 	zevtchan->zn_failed = B_FALSE;
5899 
5900 	if (pthread_mutex_init(&(zevtchan->zn_mutex), NULL))
5901 		goto out3;
5902 	if (pthread_cond_init(&(zevtchan->zn_cond), NULL)) {
5903 		(void) pthread_mutex_destroy(&(zevtchan->zn_mutex));
5904 		goto out3;
5905 	}
5906 	if (pthread_mutex_init(&(zevtchan->zn_bigmutex), NULL)) {
5907 		(void) pthread_mutex_destroy(&(zevtchan->zn_mutex));
5908 		(void) pthread_cond_destroy(&(zevtchan->zn_cond));
5909 		goto out3;
5910 	}
5911 
5912 	if (sysevent_evc_bind(ZONE_EVENT_CHANNEL, &(zevtchan->zn_eventchan),
5913 	    0) != 0)
5914 		goto out2;
5915 
5916 	do {
5917 		/*
5918 		 * At 4 digits the subscriber ID gets too long and we have
5919 		 * no chance of successfully registering.
5920 		 */
5921 		if (i > 999)
5922 			goto out1;
5923 
5924 		(void) sprintf(zevtchan->zn_subscriber_id, "zone_%li_%i",
5925 		    getpid() % 999999l, i);
5926 
5927 		r = sysevent_evc_subscribe(zevtchan->zn_eventchan,
5928 		    zevtchan->zn_subscriber_id, ZONE_EVENT_STATUS_CLASS, zn_cb,
5929 		    zevtchan, 0);
5930 
5931 		i++;
5932 
5933 	} while (r);
5934 
5935 	return (zevtchan);
5936 out1:
5937 	sysevent_evc_unbind(zevtchan->zn_eventchan);
5938 out2:
5939 	(void) pthread_mutex_destroy(&zevtchan->zn_mutex);
5940 	(void) pthread_cond_destroy(&zevtchan->zn_cond);
5941 	(void) pthread_mutex_destroy(&(zevtchan->zn_bigmutex));
5942 out3:
5943 	free(zevtchan);
5944 
5945 	return (NULL);
5946 }
5947 
5948 void
5949 zonecfg_notify_unbind(void *handle)
5950 {
5951 
5952 	int ret;
5953 
5954 	sysevent_evc_unbind(((struct znotify *)handle)->zn_eventchan);
5955 	/*
5956 	 * Check that all evc threads have gone away. This should be
5957 	 * enforced by sysevent_evc_unbind.
5958 	 */
5959 	ret = pthread_mutex_trylock(&((struct znotify *)handle)->zn_mutex);
5960 
5961 	if (ret)
5962 		abort();
5963 
5964 	(void) pthread_mutex_unlock(&((struct znotify *)handle)->zn_mutex);
5965 	(void) pthread_mutex_destroy(&((struct znotify *)handle)->zn_mutex);
5966 	(void) pthread_cond_destroy(&((struct znotify *)handle)->zn_cond);
5967 	(void) pthread_mutex_destroy(&((struct znotify *)handle)->zn_bigmutex);
5968 
5969 	free(handle);
5970 }
5971 
5972 static int
5973 zonecfg_add_ds_core(zone_dochandle_t handle, struct zone_dstab *tabptr)
5974 {
5975 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
5976 	int err;
5977 
5978 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_DATASET, NULL);
5979 	if ((err = newprop(newnode, DTD_ATTR_NAME,
5980 	    tabptr->zone_dataset_name)) != Z_OK)
5981 		return (err);
5982 	return (Z_OK);
5983 }
5984 
5985 int
5986 zonecfg_add_ds(zone_dochandle_t handle, struct zone_dstab *tabptr)
5987 {
5988 	int err;
5989 
5990 	if (tabptr == NULL)
5991 		return (Z_INVAL);
5992 
5993 	if ((err = operation_prep(handle)) != Z_OK)
5994 		return (err);
5995 
5996 	if ((err = zonecfg_add_ds_core(handle, tabptr)) != Z_OK)
5997 		return (err);
5998 
5999 	return (Z_OK);
6000 }
6001 
6002 static int
6003 zonecfg_delete_ds_core(zone_dochandle_t handle, struct zone_dstab *tabptr)
6004 {
6005 	xmlNodePtr cur = handle->zone_dh_cur;
6006 
6007 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6008 		if (xmlStrcmp(cur->name, DTD_ELEM_DATASET))
6009 			continue;
6010 
6011 		if (match_prop(cur, DTD_ATTR_NAME,
6012 		    tabptr->zone_dataset_name)) {
6013 			xmlUnlinkNode(cur);
6014 			xmlFreeNode(cur);
6015 			return (Z_OK);
6016 		}
6017 	}
6018 	return (Z_NO_RESOURCE_ID);
6019 }
6020 
6021 int
6022 zonecfg_delete_ds(zone_dochandle_t handle, struct zone_dstab *tabptr)
6023 {
6024 	int err;
6025 
6026 	if (tabptr == NULL)
6027 		return (Z_INVAL);
6028 
6029 	if ((err = operation_prep(handle)) != Z_OK)
6030 		return (err);
6031 
6032 	if ((err = zonecfg_delete_ds_core(handle, tabptr)) != Z_OK)
6033 		return (err);
6034 
6035 	return (Z_OK);
6036 }
6037 
6038 int
6039 zonecfg_modify_ds(
6040 	zone_dochandle_t handle,
6041 	struct zone_dstab *oldtabptr,
6042 	struct zone_dstab *newtabptr)
6043 {
6044 	int err;
6045 
6046 	if (oldtabptr == NULL || newtabptr == NULL)
6047 		return (Z_INVAL);
6048 
6049 	if ((err = operation_prep(handle)) != Z_OK)
6050 		return (err);
6051 
6052 	if ((err = zonecfg_delete_ds_core(handle, oldtabptr)) != Z_OK)
6053 		return (err);
6054 
6055 	if ((err = zonecfg_add_ds_core(handle, newtabptr)) != Z_OK)
6056 		return (err);
6057 
6058 	return (Z_OK);
6059 }
6060 
6061 int
6062 zonecfg_lookup_ds(zone_dochandle_t handle, struct zone_dstab *tabptr)
6063 {
6064 	xmlNodePtr cur, firstmatch;
6065 	int err;
6066 	char dataset[MAXNAMELEN];
6067 
6068 	if (tabptr == NULL)
6069 		return (Z_INVAL);
6070 
6071 	if ((err = operation_prep(handle)) != Z_OK)
6072 		return (err);
6073 
6074 	cur = handle->zone_dh_cur;
6075 	firstmatch = NULL;
6076 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6077 		if (xmlStrcmp(cur->name, DTD_ELEM_DATASET))
6078 			continue;
6079 		if (strlen(tabptr->zone_dataset_name) > 0) {
6080 			if ((fetchprop(cur, DTD_ATTR_NAME, dataset,
6081 			    sizeof (dataset)) == Z_OK) &&
6082 			    (strcmp(tabptr->zone_dataset_name,
6083 			    dataset) == 0)) {
6084 				if (firstmatch == NULL)
6085 					firstmatch = cur;
6086 				else
6087 					return (Z_INSUFFICIENT_SPEC);
6088 			}
6089 		}
6090 	}
6091 	if (firstmatch == NULL)
6092 		return (Z_NO_RESOURCE_ID);
6093 
6094 	cur = firstmatch;
6095 
6096 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_dataset_name,
6097 	    sizeof (tabptr->zone_dataset_name))) != Z_OK)
6098 		return (err);
6099 
6100 	return (Z_OK);
6101 }
6102 
6103 int
6104 zonecfg_setdsent(zone_dochandle_t handle)
6105 {
6106 	return (zonecfg_setent(handle));
6107 }
6108 
6109 int
6110 zonecfg_getdsent(zone_dochandle_t handle, struct zone_dstab *tabptr)
6111 {
6112 	xmlNodePtr cur;
6113 	int err;
6114 
6115 	if (handle == NULL)
6116 		return (Z_INVAL);
6117 
6118 	if ((cur = handle->zone_dh_cur) == NULL)
6119 		return (Z_NO_ENTRY);
6120 
6121 	for (; cur != NULL; cur = cur->next)
6122 		if (!xmlStrcmp(cur->name, DTD_ELEM_DATASET))
6123 			break;
6124 	if (cur == NULL) {
6125 		handle->zone_dh_cur = handle->zone_dh_top;
6126 		return (Z_NO_ENTRY);
6127 	}
6128 
6129 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_dataset_name,
6130 	    sizeof (tabptr->zone_dataset_name))) != Z_OK) {
6131 		handle->zone_dh_cur = handle->zone_dh_top;
6132 		return (err);
6133 	}
6134 
6135 	handle->zone_dh_cur = cur->next;
6136 	return (Z_OK);
6137 }
6138 
6139 int
6140 zonecfg_enddsent(zone_dochandle_t handle)
6141 {
6142 	return (zonecfg_endent(handle));
6143 }
6144 
6145 /*
6146  * Support for aliased rctls; that is, rctls that have simplified names in
6147  * zonecfg.  For example, max-lwps is an alias for a well defined zone.max-lwps
6148  * rctl.  If there are multiple existing values for one of these rctls or if
6149  * there is a single value that does not match the well defined template (i.e.
6150  * it has a different action) then we cannot treat the rctl as having an alias
6151  * so we return Z_ALIAS_DISALLOW.  That means that the rctl cannot be
6152  * managed in zonecfg via an alias and that the standard rctl syntax must be
6153  * used.
6154  *
6155  * The possible return values are:
6156  *	Z_NO_PROPERTY_ID - invalid alias name
6157  *	Z_ALIAS_DISALLOW - pre-existing, incompatible rctl definition
6158  *	Z_NO_ENTRY - no rctl is configured for this alias
6159  *	Z_OK - we got a valid rctl for the specified alias
6160  */
6161 int
6162 zonecfg_get_aliased_rctl(zone_dochandle_t handle, char *name, uint64_t *rval)
6163 {
6164 	boolean_t found = B_FALSE;
6165 	boolean_t found_val = B_FALSE;
6166 	xmlNodePtr cur, val;
6167 	char savedname[MAXNAMELEN];
6168 	struct zone_rctlvaltab rctl;
6169 	int i;
6170 	int err;
6171 
6172 	for (i = 0; aliases[i].shortname != NULL; i++)
6173 		if (strcmp(name, aliases[i].shortname) == 0)
6174 			break;
6175 
6176 	if (aliases[i].shortname == NULL)
6177 		return (Z_NO_PROPERTY_ID);
6178 
6179 	if ((err = operation_prep(handle)) != Z_OK)
6180 		return (err);
6181 
6182 	cur = handle->zone_dh_cur;
6183 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6184 		if (xmlStrcmp(cur->name, DTD_ELEM_RCTL) != 0)
6185 			continue;
6186 		if ((fetchprop(cur, DTD_ATTR_NAME, savedname,
6187 		    sizeof (savedname)) == Z_OK) &&
6188 		    (strcmp(savedname, aliases[i].realname) == 0)) {
6189 
6190 			/*
6191 			 * If we already saw one of these, we can't have an
6192 			 * alias since we just found another.
6193 			 */
6194 			if (found)
6195 				return (Z_ALIAS_DISALLOW);
6196 			found = B_TRUE;
6197 
6198 			for (val = cur->xmlChildrenNode; val != NULL;
6199 			    val = val->next) {
6200 				/*
6201 				 * If we already have one value, we can't have
6202 				 * an alias since we just found another.
6203 				 */
6204 				if (found_val)
6205 					return (Z_ALIAS_DISALLOW);
6206 				found_val = B_TRUE;
6207 
6208 				if ((fetchprop(val, DTD_ATTR_PRIV,
6209 				    rctl.zone_rctlval_priv,
6210 				    sizeof (rctl.zone_rctlval_priv)) != Z_OK))
6211 					break;
6212 				if ((fetchprop(val, DTD_ATTR_LIMIT,
6213 				    rctl.zone_rctlval_limit,
6214 				    sizeof (rctl.zone_rctlval_limit)) != Z_OK))
6215 					break;
6216 				if ((fetchprop(val, DTD_ATTR_ACTION,
6217 				    rctl.zone_rctlval_action,
6218 				    sizeof (rctl.zone_rctlval_action)) != Z_OK))
6219 					break;
6220 			}
6221 
6222 			/* check priv and action match the expected vals */
6223 			if (strcmp(rctl.zone_rctlval_priv,
6224 			    aliases[i].priv) != 0 ||
6225 			    strcmp(rctl.zone_rctlval_action,
6226 			    aliases[i].action) != 0)
6227 				return (Z_ALIAS_DISALLOW);
6228 		}
6229 	}
6230 
6231 	if (found) {
6232 		*rval = strtoull(rctl.zone_rctlval_limit, NULL, 10);
6233 		return (Z_OK);
6234 	}
6235 
6236 	return (Z_NO_ENTRY);
6237 }
6238 
6239 int
6240 zonecfg_rm_aliased_rctl(zone_dochandle_t handle, char *name)
6241 {
6242 	int i;
6243 	uint64_t val;
6244 	struct zone_rctltab rctltab;
6245 
6246 	/*
6247 	 * First check that we have a valid aliased rctl to remove.
6248 	 * This will catch an rctl entry with non-standard values or
6249 	 * multiple rctl values for this name.  We need to ignore those
6250 	 * rctl entries.
6251 	 */
6252 	if (zonecfg_get_aliased_rctl(handle, name, &val) != Z_OK)
6253 		return (Z_OK);
6254 
6255 	for (i = 0; aliases[i].shortname != NULL; i++)
6256 		if (strcmp(name, aliases[i].shortname) == 0)
6257 			break;
6258 
6259 	if (aliases[i].shortname == NULL)
6260 		return (Z_NO_RESOURCE_ID);
6261 
6262 	(void) strlcpy(rctltab.zone_rctl_name, aliases[i].realname,
6263 	    sizeof (rctltab.zone_rctl_name));
6264 
6265 	return (zonecfg_delete_rctl(handle, &rctltab));
6266 }
6267 
6268 boolean_t
6269 zonecfg_aliased_rctl_ok(zone_dochandle_t handle, char *name)
6270 {
6271 	uint64_t tmp_val;
6272 
6273 	switch (zonecfg_get_aliased_rctl(handle, name, &tmp_val)) {
6274 	case Z_OK:
6275 		/*FALLTHRU*/
6276 	case Z_NO_ENTRY:
6277 		return (B_TRUE);
6278 	default:
6279 		return (B_FALSE);
6280 	}
6281 }
6282 
6283 int
6284 zonecfg_set_aliased_rctl(zone_dochandle_t handle, char *name, uint64_t val)
6285 {
6286 	int i;
6287 	int err;
6288 	struct zone_rctltab rctltab;
6289 	struct zone_rctlvaltab *rctlvaltab;
6290 	char buf[128];
6291 
6292 	if (!zonecfg_aliased_rctl_ok(handle, name))
6293 		return (Z_ALIAS_DISALLOW);
6294 
6295 	for (i = 0; aliases[i].shortname != NULL; i++)
6296 		if (strcmp(name, aliases[i].shortname) == 0)
6297 			break;
6298 
6299 	if (aliases[i].shortname == NULL)
6300 		return (Z_NO_RESOURCE_ID);
6301 
6302 	/* remove any pre-existing definition for this rctl */
6303 	(void) zonecfg_rm_aliased_rctl(handle, name);
6304 
6305 	(void) strlcpy(rctltab.zone_rctl_name, aliases[i].realname,
6306 	    sizeof (rctltab.zone_rctl_name));
6307 
6308 	rctltab.zone_rctl_valptr = NULL;
6309 
6310 	if ((rctlvaltab = calloc(1, sizeof (struct zone_rctlvaltab))) == NULL)
6311 		return (Z_NOMEM);
6312 
6313 	(void) snprintf(buf, sizeof (buf), "%llu", (long long)val);
6314 
6315 	(void) strlcpy(rctlvaltab->zone_rctlval_priv, aliases[i].priv,
6316 	    sizeof (rctlvaltab->zone_rctlval_priv));
6317 	(void) strlcpy(rctlvaltab->zone_rctlval_limit, buf,
6318 	    sizeof (rctlvaltab->zone_rctlval_limit));
6319 	(void) strlcpy(rctlvaltab->zone_rctlval_action, aliases[i].action,
6320 	    sizeof (rctlvaltab->zone_rctlval_action));
6321 
6322 	rctlvaltab->zone_rctlval_next = NULL;
6323 
6324 	if ((err = zonecfg_add_rctl_value(&rctltab, rctlvaltab)) != Z_OK)
6325 		return (err);
6326 
6327 	return (zonecfg_add_rctl(handle, &rctltab));
6328 }
6329 
6330 static int
6331 delete_tmp_pool(zone_dochandle_t handle)
6332 {
6333 	int err;
6334 	xmlNodePtr cur = handle->zone_dh_cur;
6335 
6336 	if ((err = operation_prep(handle)) != Z_OK)
6337 		return (err);
6338 
6339 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6340 		if (xmlStrcmp(cur->name, DTD_ELEM_TMPPOOL) == 0) {
6341 			xmlUnlinkNode(cur);
6342 			xmlFreeNode(cur);
6343 			return (Z_OK);
6344 		}
6345 	}
6346 
6347 	return (Z_NO_RESOURCE_ID);
6348 }
6349 
6350 static int
6351 modify_tmp_pool(zone_dochandle_t handle, char *pool_importance)
6352 {
6353 	int err;
6354 	xmlNodePtr cur = handle->zone_dh_cur;
6355 	xmlNodePtr newnode;
6356 
6357 	err = delete_tmp_pool(handle);
6358 	if (err != Z_OK && err != Z_NO_RESOURCE_ID)
6359 		return (err);
6360 
6361 	if (*pool_importance != '\0') {
6362 		if ((err = operation_prep(handle)) != Z_OK)
6363 			return (err);
6364 
6365 		newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_TMPPOOL, NULL);
6366 		if ((err = newprop(newnode, DTD_ATTR_IMPORTANCE,
6367 		    pool_importance)) != Z_OK)
6368 			return (err);
6369 	}
6370 
6371 	return (Z_OK);
6372 }
6373 
6374 static int
6375 add_pset_core(zone_dochandle_t handle, struct zone_psettab *tabptr)
6376 {
6377 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
6378 	int err;
6379 
6380 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_PSET, NULL);
6381 	if ((err = newprop(newnode, DTD_ATTR_NCPU_MIN,
6382 	    tabptr->zone_ncpu_min)) != Z_OK)
6383 		return (err);
6384 	if ((err = newprop(newnode, DTD_ATTR_NCPU_MAX,
6385 	    tabptr->zone_ncpu_max)) != Z_OK)
6386 		return (err);
6387 
6388 	if ((err = modify_tmp_pool(handle, tabptr->zone_importance)) != Z_OK)
6389 		return (err);
6390 
6391 	return (Z_OK);
6392 }
6393 
6394 int
6395 zonecfg_add_pset(zone_dochandle_t handle, struct zone_psettab *tabptr)
6396 {
6397 	int err;
6398 
6399 	if (tabptr == NULL)
6400 		return (Z_INVAL);
6401 
6402 	if ((err = operation_prep(handle)) != Z_OK)
6403 		return (err);
6404 
6405 	if ((err = add_pset_core(handle, tabptr)) != Z_OK)
6406 		return (err);
6407 
6408 	return (Z_OK);
6409 }
6410 
6411 int
6412 zonecfg_delete_pset(zone_dochandle_t handle)
6413 {
6414 	int err;
6415 	int res = Z_NO_RESOURCE_ID;
6416 	xmlNodePtr cur = handle->zone_dh_cur;
6417 
6418 	if ((err = operation_prep(handle)) != Z_OK)
6419 		return (err);
6420 
6421 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6422 		if (xmlStrcmp(cur->name, DTD_ELEM_PSET) == 0) {
6423 			xmlUnlinkNode(cur);
6424 			xmlFreeNode(cur);
6425 			res = Z_OK;
6426 			break;
6427 		}
6428 	}
6429 
6430 	/*
6431 	 * Once we have msets, we should check that a mset
6432 	 * do not exist before we delete the tmp_pool data.
6433 	 */
6434 	err = delete_tmp_pool(handle);
6435 	if (err != Z_OK && err != Z_NO_RESOURCE_ID)
6436 		return (err);
6437 
6438 	return (res);
6439 }
6440 
6441 int
6442 zonecfg_modify_pset(zone_dochandle_t handle, struct zone_psettab *tabptr)
6443 {
6444 	int err;
6445 
6446 	if (tabptr == NULL)
6447 		return (Z_INVAL);
6448 
6449 	if ((err = zonecfg_delete_pset(handle)) != Z_OK)
6450 		return (err);
6451 
6452 	if ((err = add_pset_core(handle, tabptr)) != Z_OK)
6453 		return (err);
6454 
6455 	return (Z_OK);
6456 }
6457 
6458 int
6459 zonecfg_lookup_pset(zone_dochandle_t handle, struct zone_psettab *tabptr)
6460 {
6461 	xmlNodePtr cur;
6462 	int err;
6463 	int res = Z_NO_ENTRY;
6464 
6465 	if (tabptr == NULL)
6466 		return (Z_INVAL);
6467 
6468 	if ((err = operation_prep(handle)) != Z_OK)
6469 		return (err);
6470 
6471 	/* this is an optional component */
6472 	tabptr->zone_importance[0] = '\0';
6473 
6474 	cur = handle->zone_dh_cur;
6475 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6476 		if (xmlStrcmp(cur->name, DTD_ELEM_PSET) == 0) {
6477 			if ((err = fetchprop(cur, DTD_ATTR_NCPU_MIN,
6478 			    tabptr->zone_ncpu_min,
6479 			    sizeof (tabptr->zone_ncpu_min))) != Z_OK) {
6480 				handle->zone_dh_cur = handle->zone_dh_top;
6481 				return (err);
6482 			}
6483 
6484 			if ((err = fetchprop(cur, DTD_ATTR_NCPU_MAX,
6485 			    tabptr->zone_ncpu_max,
6486 			    sizeof (tabptr->zone_ncpu_max))) != Z_OK) {
6487 				handle->zone_dh_cur = handle->zone_dh_top;
6488 				return (err);
6489 			}
6490 
6491 			res = Z_OK;
6492 
6493 		} else if (xmlStrcmp(cur->name, DTD_ELEM_TMPPOOL) == 0) {
6494 			if ((err = fetchprop(cur, DTD_ATTR_IMPORTANCE,
6495 			    tabptr->zone_importance,
6496 			    sizeof (tabptr->zone_importance))) != Z_OK) {
6497 				handle->zone_dh_cur = handle->zone_dh_top;
6498 				return (err);
6499 			}
6500 		}
6501 	}
6502 
6503 	return (res);
6504 }
6505 
6506 int
6507 zonecfg_getpsetent(zone_dochandle_t handle, struct zone_psettab *tabptr)
6508 {
6509 	int err;
6510 
6511 	if ((err = zonecfg_setent(handle)) != Z_OK)
6512 		return (err);
6513 
6514 	err = zonecfg_lookup_pset(handle, tabptr);
6515 
6516 	(void) zonecfg_endent(handle);
6517 
6518 	return (err);
6519 }
6520 
6521 static int
6522 add_mcap(zone_dochandle_t handle, struct zone_mcaptab *tabptr)
6523 {
6524 	xmlNodePtr newnode, cur = handle->zone_dh_cur;
6525 	int err;
6526 
6527 	newnode = xmlNewTextChild(cur, NULL, DTD_ELEM_MCAP, NULL);
6528 	if ((err = newprop(newnode, DTD_ATTR_PHYSCAP, tabptr->zone_physmem_cap))
6529 	    != Z_OK)
6530 		return (err);
6531 
6532 	return (Z_OK);
6533 }
6534 
6535 int
6536 zonecfg_delete_mcap(zone_dochandle_t handle)
6537 {
6538 	int err;
6539 	xmlNodePtr cur = handle->zone_dh_cur;
6540 
6541 	if ((err = operation_prep(handle)) != Z_OK)
6542 		return (err);
6543 
6544 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6545 		if (xmlStrcmp(cur->name, DTD_ELEM_MCAP) != 0)
6546 			continue;
6547 
6548 		xmlUnlinkNode(cur);
6549 		xmlFreeNode(cur);
6550 		return (Z_OK);
6551 	}
6552 	return (Z_NO_RESOURCE_ID);
6553 }
6554 
6555 int
6556 zonecfg_modify_mcap(zone_dochandle_t handle, struct zone_mcaptab *tabptr)
6557 {
6558 	int err;
6559 
6560 	if (tabptr == NULL)
6561 		return (Z_INVAL);
6562 
6563 	err = zonecfg_delete_mcap(handle);
6564 	/* it is ok if there is no mcap entry */
6565 	if (err != Z_OK && err != Z_NO_RESOURCE_ID)
6566 		return (err);
6567 
6568 	if ((err = add_mcap(handle, tabptr)) != Z_OK)
6569 		return (err);
6570 
6571 	return (Z_OK);
6572 }
6573 
6574 int
6575 zonecfg_lookup_mcap(zone_dochandle_t handle, struct zone_mcaptab *tabptr)
6576 {
6577 	xmlNodePtr cur;
6578 	int err;
6579 
6580 	if (tabptr == NULL)
6581 		return (Z_INVAL);
6582 
6583 	if ((err = operation_prep(handle)) != Z_OK)
6584 		return (err);
6585 
6586 	cur = handle->zone_dh_cur;
6587 	for (cur = cur->xmlChildrenNode; cur != NULL; cur = cur->next) {
6588 		if (xmlStrcmp(cur->name, DTD_ELEM_MCAP) != 0)
6589 			continue;
6590 		if ((err = fetchprop(cur, DTD_ATTR_PHYSCAP,
6591 		    tabptr->zone_physmem_cap,
6592 		    sizeof (tabptr->zone_physmem_cap))) != Z_OK) {
6593 			handle->zone_dh_cur = handle->zone_dh_top;
6594 			return (err);
6595 		}
6596 
6597 		return (Z_OK);
6598 	}
6599 
6600 	return (Z_NO_ENTRY);
6601 }
6602 
6603 static int
6604 getmcapent_core(zone_dochandle_t handle, struct zone_mcaptab *tabptr)
6605 {
6606 	xmlNodePtr cur;
6607 	int err;
6608 
6609 	if (handle == NULL)
6610 		return (Z_INVAL);
6611 
6612 	if ((cur = handle->zone_dh_cur) == NULL)
6613 		return (Z_NO_ENTRY);
6614 
6615 	for (; cur != NULL; cur = cur->next)
6616 		if (xmlStrcmp(cur->name, DTD_ELEM_MCAP) == 0)
6617 			break;
6618 	if (cur == NULL) {
6619 		handle->zone_dh_cur = handle->zone_dh_top;
6620 		return (Z_NO_ENTRY);
6621 	}
6622 
6623 	if ((err = fetchprop(cur, DTD_ATTR_PHYSCAP, tabptr->zone_physmem_cap,
6624 	    sizeof (tabptr->zone_physmem_cap))) != Z_OK) {
6625 		handle->zone_dh_cur = handle->zone_dh_top;
6626 		return (err);
6627 	}
6628 
6629 	handle->zone_dh_cur = cur->next;
6630 	return (Z_OK);
6631 }
6632 
6633 int
6634 zonecfg_getmcapent(zone_dochandle_t handle, struct zone_mcaptab *tabptr)
6635 {
6636 	int err;
6637 
6638 	if ((err = zonecfg_setent(handle)) != Z_OK)
6639 		return (err);
6640 
6641 	err = getmcapent_core(handle, tabptr);
6642 
6643 	(void) zonecfg_endent(handle);
6644 
6645 	return (err);
6646 }
6647 
6648 /*
6649  * Get the full tree of pkg/patch metadata in a set of nested AVL trees.
6650  * pkgs_avl is an AVL tree of pkgs.  Each pkg element contains a
6651  * zpe_patches_avl member which holds an AVL tree of patches for that pkg.
6652  * The patch elements have the same zpe_patches_avl member, each of which can
6653  * hold an AVL tree of patches that are obsoleted by the patch.
6654  *
6655  * The zone xml data contains DTD_ELEM_PACKAGE elements, followed by
6656  * DTD_ELEM_PATCH elements.  The DTD_ELEM_PATCH patch element applies to the
6657  * DTD_ELEM_PACKAGE that precedes it.  The DTD_ELEM_PATCH element may have
6658  * child DTD_ELEM_OBSOLETES nodes associated with it.  The DTD_ELEM_PACKAGE
6659  * really should have had the DTD_ELEM_PATCH elements as children but it
6660  * was not defined that way initially so we are stuck with the DTD definition
6661  * now.  However, we can safely assume the ordering for compatibility.
6662  */
6663 int
6664 zonecfg_getpkgdata(zone_dochandle_t handle, uu_avl_pool_t *pkg_pool,
6665     uu_avl_t *pkgs_avl)
6666 {
6667 	xmlNodePtr cur;
6668 	int res;
6669 	zone_pkg_entry_t *pkg;
6670 	char name[MAXNAMELEN];
6671 	char version[ZONE_PKG_VERSMAX];
6672 
6673 	if (handle == NULL)
6674 		return (Z_INVAL);
6675 
6676 	if ((res = zonecfg_setent(handle)) != Z_OK)
6677 		return (res);
6678 
6679 	if ((cur = handle->zone_dh_cur) == NULL) {
6680 		res = Z_NO_ENTRY;
6681 		goto done;
6682 	}
6683 
6684 	for (; cur != NULL; cur = cur->next) {
6685 		if (xmlStrcmp(cur->name, DTD_ELEM_PACKAGE) == 0) {
6686 			uu_avl_index_t where;
6687 
6688 			if ((res = fetchprop(cur, DTD_ATTR_NAME, name,
6689 			    sizeof (name))) != Z_OK)
6690 				goto done;
6691 
6692 			if ((res = fetchprop(cur, DTD_ATTR_VERSION, version,
6693 			    sizeof (version))) != Z_OK)
6694 				goto done;
6695 
6696 			if ((pkg = (zone_pkg_entry_t *)
6697 			    malloc(sizeof (zone_pkg_entry_t))) == NULL) {
6698 				res = Z_NOMEM;
6699 				goto done;
6700 			}
6701 
6702 			if ((pkg->zpe_name = strdup(name)) == NULL) {
6703 				free(pkg);
6704 				res = Z_NOMEM;
6705 				goto done;
6706 			}
6707 
6708 			if ((pkg->zpe_vers = strdup(version)) == NULL) {
6709 				free(pkg->zpe_name);
6710 				free(pkg);
6711 				res = Z_NOMEM;
6712 				goto done;
6713 			}
6714 
6715 			pkg->zpe_patches_avl = NULL;
6716 
6717 			uu_avl_node_init(pkg, &pkg->zpe_entry, pkg_pool);
6718 			if (uu_avl_find(pkgs_avl, pkg, NULL, &where) != NULL) {
6719 				free(pkg->zpe_name);
6720 				free(pkg->zpe_vers);
6721 				free(pkg);
6722 			} else {
6723 				uu_avl_insert(pkgs_avl, pkg, where);
6724 			}
6725 
6726 		} else if (xmlStrcmp(cur->name, DTD_ELEM_PATCH) == 0) {
6727 			zone_pkg_entry_t *patch;
6728 			uu_avl_index_t where;
6729 			char *p;
6730 			char *dashp = NULL;
6731 			xmlNodePtr child;
6732 
6733 			if ((res = fetchprop(cur, DTD_ATTR_ID, name,
6734 			    sizeof (name))) != Z_OK)
6735 				goto done;
6736 
6737 			if ((patch = (zone_pkg_entry_t *)
6738 			    malloc(sizeof (zone_pkg_entry_t))) == NULL) {
6739 				res = Z_NOMEM;
6740 				goto done;
6741 			}
6742 
6743 			if ((p = strchr(name, '-')) != NULL) {
6744 				dashp = p;
6745 				*p++ = '\0';
6746 			} else {
6747 				p = "";
6748 			}
6749 
6750 			if ((patch->zpe_name = strdup(name)) == NULL) {
6751 				free(patch);
6752 				res = Z_NOMEM;
6753 				goto done;
6754 			}
6755 
6756 			if ((patch->zpe_vers = strdup(p)) == NULL) {
6757 				free(patch->zpe_name);
6758 				free(patch);
6759 				res = Z_NOMEM;
6760 				goto done;
6761 			}
6762 
6763 			if (dashp != NULL)
6764 				*dashp = '-';
6765 
6766 			patch->zpe_patches_avl = NULL;
6767 
6768 			if (pkg->zpe_patches_avl == NULL) {
6769 				pkg->zpe_patches_avl = uu_avl_create(pkg_pool,
6770 				    NULL, UU_DEFAULT);
6771 				if (pkg->zpe_patches_avl == NULL) {
6772 					free(patch->zpe_name);
6773 					free(patch->zpe_vers);
6774 					free(patch);
6775 					res = Z_NOMEM;
6776 					goto done;
6777 				}
6778 			}
6779 
6780 			uu_avl_node_init(patch, &patch->zpe_entry, pkg_pool);
6781 			if (uu_avl_find(pkg->zpe_patches_avl, patch, NULL,
6782 			    &where) != NULL) {
6783 				free(patch->zpe_name);
6784 				free(patch->zpe_vers);
6785 				free(patch);
6786 			} else {
6787 				uu_avl_insert(pkg->zpe_patches_avl, patch,
6788 				    where);
6789 			}
6790 
6791 			/* Add any patches this patch obsoletes. */
6792 			for (child = cur->xmlChildrenNode; child != NULL;
6793 			    child = child->next) {
6794 				zone_pkg_entry_t *obs;
6795 
6796 				if (xmlStrcmp(child->name, DTD_ELEM_OBSOLETES)
6797 				    != 0)
6798 					continue;
6799 
6800 				if ((res = fetchprop(child, DTD_ATTR_ID,
6801 				    name, sizeof (name))) != Z_OK)
6802 					goto done;
6803 
6804 				if ((obs = (zone_pkg_entry_t *)malloc(
6805 				    sizeof (zone_pkg_entry_t))) == NULL) {
6806 					res = Z_NOMEM;
6807 					goto done;
6808 				}
6809 
6810 				if ((obs->zpe_name = strdup(name)) == NULL) {
6811 					free(obs);
6812 					res = Z_NOMEM;
6813 					goto done;
6814 				}
6815 				/*
6816 				 * The version doesn't matter for obsoleted
6817 				 * patches.
6818 				 */
6819 				obs->zpe_vers = NULL;
6820 				obs->zpe_patches_avl = NULL;
6821 
6822 				/*
6823 				 * If this is the first obsolete patch, add an
6824 				 * AVL tree to the parent patch element.
6825 				 */
6826 				if (patch->zpe_patches_avl == NULL) {
6827 					patch->zpe_patches_avl =
6828 					    uu_avl_create(pkg_pool, NULL,
6829 					    UU_DEFAULT);
6830 					if (patch->zpe_patches_avl == NULL) {
6831 						free(obs->zpe_name);
6832 						free(obs);
6833 						res = Z_NOMEM;
6834 						goto done;
6835 					}
6836 				}
6837 
6838 				/* Insert obsolete patch into the AVL tree. */
6839 				uu_avl_node_init(obs, &obs->zpe_entry,
6840 				    pkg_pool);
6841 				if (uu_avl_find(patch->zpe_patches_avl, obs,
6842 				    NULL, &where) != NULL) {
6843 					free(obs->zpe_name);
6844 					free(obs);
6845 				} else {
6846 					uu_avl_insert(patch->zpe_patches_avl,
6847 					    obs, where);
6848 				}
6849 			}
6850 		}
6851 	}
6852 
6853 done:
6854 	(void) zonecfg_endent(handle);
6855 	return (res);
6856 }
6857 
6858 int
6859 zonecfg_setdevperment(zone_dochandle_t handle)
6860 {
6861 	return (zonecfg_setent(handle));
6862 }
6863 
6864 int
6865 zonecfg_getdevperment(zone_dochandle_t handle, struct zone_devpermtab *tabptr)
6866 {
6867 	xmlNodePtr cur;
6868 	int err;
6869 	char buf[128];
6870 
6871 	tabptr->zone_devperm_acl = NULL;
6872 
6873 	if (handle == NULL)
6874 		return (Z_INVAL);
6875 
6876 	if ((cur = handle->zone_dh_cur) == NULL)
6877 		return (Z_NO_ENTRY);
6878 
6879 	for (; cur != NULL; cur = cur->next)
6880 		if (!xmlStrcmp(cur->name, DTD_ELEM_DEV_PERM))
6881 			break;
6882 	if (cur == NULL) {
6883 		handle->zone_dh_cur = handle->zone_dh_top;
6884 		return (Z_NO_ENTRY);
6885 	}
6886 
6887 	if ((err = fetchprop(cur, DTD_ATTR_NAME, tabptr->zone_devperm_name,
6888 	    sizeof (tabptr->zone_devperm_name))) != Z_OK) {
6889 		handle->zone_dh_cur = handle->zone_dh_top;
6890 		return (err);
6891 	}
6892 
6893 	if ((err = fetchprop(cur, DTD_ATTR_UID, buf, sizeof (buf))) != Z_OK) {
6894 		handle->zone_dh_cur = handle->zone_dh_top;
6895 		return (err);
6896 	}
6897 	tabptr->zone_devperm_uid = (uid_t)atol(buf);
6898 
6899 	if ((err = fetchprop(cur, DTD_ATTR_GID, buf, sizeof (buf))) != Z_OK) {
6900 		handle->zone_dh_cur = handle->zone_dh_top;
6901 		return (err);
6902 	}
6903 	tabptr->zone_devperm_gid = (gid_t)atol(buf);
6904 
6905 	if ((err = fetchprop(cur, DTD_ATTR_MODE, buf, sizeof (buf))) != Z_OK) {
6906 		handle->zone_dh_cur = handle->zone_dh_top;
6907 		return (err);
6908 	}
6909 	tabptr->zone_devperm_mode = (mode_t)strtol(buf, (char **)NULL, 8);
6910 
6911 	if ((err = fetch_alloc_prop(cur, DTD_ATTR_ACL,
6912 	    &(tabptr->zone_devperm_acl))) != Z_OK) {
6913 		handle->zone_dh_cur = handle->zone_dh_top;
6914 		return (err);
6915 	}
6916 
6917 	handle->zone_dh_cur = cur->next;
6918 	return (Z_OK);
6919 }
6920 
6921 int
6922 zonecfg_enddevperment(zone_dochandle_t handle)
6923 {
6924 	return (zonecfg_endent(handle));
6925 }
6926 
6927 /* PRINTFLIKE1 */
6928 static void
6929 zerror(const char *zone_name, const char *fmt, ...)
6930 {
6931 	va_list alist;
6932 
6933 	va_start(alist, fmt);
6934 	(void) fprintf(stderr, "zone '%s': ", zone_name);
6935 	(void) vfprintf(stderr, fmt, alist);
6936 	(void) fprintf(stderr, "\n");
6937 	va_end(alist);
6938 }
6939 
6940 static void
6941 zperror(const char *str)
6942 {
6943 	(void) fprintf(stderr, "%s: %s\n", str, strerror(errno));
6944 }
6945 
6946 /*
6947  * The following three routines implement a simple locking mechanism to
6948  * ensure that only one instance of zoneadm at a time is able to manipulate
6949  * a given zone.  The lock is built on top of an fcntl(2) lock of
6950  * [<altroot>]/var/run/zones/<zonename>.zoneadm.lock.  If a zoneadm instance
6951  * can grab that lock, it is allowed to manipulate the zone.
6952  *
6953  * Since zoneadm may call external applications which in turn invoke
6954  * zoneadm again, we introduce the notion of "lock inheritance".  Any
6955  * instance of zoneadm that has another instance in its ancestry is assumed
6956  * to be acting on behalf of the original zoneadm, and is thus allowed to
6957  * manipulate its zone.
6958  *
6959  * This inheritance is implemented via the _ZONEADM_LOCK_HELD environment
6960  * variable.  When zoneadm is granted a lock on its zone, this environment
6961  * variable is set to 1.  When it releases the lock, the variable is set to
6962  * 0.  Since a child process inherits its parent's environment, checking
6963  * the state of this variable indicates whether or not any ancestor owns
6964  * the lock.
6965  */
6966 void
6967 zonecfg_init_lock_file(const char *zone_name, char **lock_env)
6968 {
6969 	*lock_env = getenv(LOCK_ENV_VAR);
6970 	if (*lock_env == NULL) {
6971 		if (putenv(zoneadm_lock_not_held) != 0) {
6972 			zerror(zone_name, gettext("could not set env: %s"),
6973 			    strerror(errno));
6974 			exit(1);
6975 		}
6976 	} else {
6977 		if (atoi(*lock_env) == 1)
6978 			zone_lock_cnt = 1;
6979 	}
6980 }
6981 
6982 void
6983 zonecfg_release_lock_file(const char *zone_name, int lockfd)
6984 {
6985 	/*
6986 	 * If we are cleaning up from a failed attempt to lock the zone for
6987 	 * the first time, we might have a zone_lock_cnt of 0.  In that
6988 	 * error case, we don't want to do anything but close the lock
6989 	 * file.
6990 	 */
6991 	assert(zone_lock_cnt >= 0);
6992 	if (zone_lock_cnt > 0) {
6993 		assert(getenv(LOCK_ENV_VAR) != NULL);
6994 		assert(atoi(getenv(LOCK_ENV_VAR)) == 1);
6995 		if (--zone_lock_cnt > 0) {
6996 			assert(lockfd == -1);
6997 			return;
6998 		}
6999 		if (putenv(zoneadm_lock_not_held) != 0) {
7000 			zerror(zone_name, gettext("could not set env: %s"),
7001 			    strerror(errno));
7002 			exit(1);
7003 		}
7004 	}
7005 	assert(lockfd >= 0);
7006 	(void) close(lockfd);
7007 }
7008 
7009 int
7010 zonecfg_grab_lock_file(const char *zone_name, int *lockfd)
7011 {
7012 	char pathbuf[PATH_MAX];
7013 	struct flock flock;
7014 
7015 	/*
7016 	 * If we already have the lock, we can skip this expensive song
7017 	 * and dance.
7018 	 */
7019 	assert(zone_lock_cnt >= 0);
7020 	assert(getenv(LOCK_ENV_VAR) != NULL);
7021 	if (zone_lock_cnt > 0) {
7022 		assert(atoi(getenv(LOCK_ENV_VAR)) == 1);
7023 		zone_lock_cnt++;
7024 		*lockfd = -1;
7025 		return (Z_OK);
7026 	}
7027 	assert(getenv(LOCK_ENV_VAR) != NULL);
7028 	assert(atoi(getenv(LOCK_ENV_VAR)) == 0);
7029 
7030 	if (snprintf(pathbuf, sizeof (pathbuf), "%s%s", zonecfg_get_root(),
7031 	    ZONES_TMPDIR) >= sizeof (pathbuf)) {
7032 		zerror(zone_name, gettext("alternate root path is too long"));
7033 		return (-1);
7034 	}
7035 	if (mkdir(pathbuf, S_IRWXU) < 0 && errno != EEXIST) {
7036 		zerror(zone_name, gettext("could not mkdir %s: %s"), pathbuf,
7037 		    strerror(errno));
7038 		return (-1);
7039 	}
7040 	(void) chmod(pathbuf, S_IRWXU);
7041 
7042 	/*
7043 	 * One of these lock files is created for each zone (when needed).
7044 	 * The lock files are not cleaned up (except on system reboot),
7045 	 * but since there is only one per zone, there is no resource
7046 	 * starvation issue.
7047 	 */
7048 	if (snprintf(pathbuf, sizeof (pathbuf), "%s%s/%s.zoneadm.lock",
7049 	    zonecfg_get_root(), ZONES_TMPDIR, zone_name) >= sizeof (pathbuf)) {
7050 		zerror(zone_name, gettext("alternate root path is too long"));
7051 		return (-1);
7052 	}
7053 	if ((*lockfd = open(pathbuf, O_RDWR|O_CREAT, S_IRUSR|S_IWUSR)) < 0) {
7054 		zerror(zone_name, gettext("could not open %s: %s"), pathbuf,
7055 		    strerror(errno));
7056 		return (-1);
7057 	}
7058 	/*
7059 	 * Lock the file to synchronize with other zoneadmds
7060 	 */
7061 	flock.l_type = F_WRLCK;
7062 	flock.l_whence = SEEK_SET;
7063 	flock.l_start = (off_t)0;
7064 	flock.l_len = (off_t)0;
7065 	if ((fcntl(*lockfd, F_SETLKW, &flock) < 0) ||
7066 	    (putenv(zoneadm_lock_held) != 0)) {
7067 		zerror(zone_name, gettext("unable to lock %s: %s"), pathbuf,
7068 		    strerror(errno));
7069 		zonecfg_release_lock_file(zone_name, *lockfd);
7070 		return (-1);
7071 	}
7072 	zone_lock_cnt = 1;
7073 	return (Z_OK);
7074 }
7075 
7076 static boolean_t
7077 get_doorname(const char *zone_name, char *buffer)
7078 {
7079 	return (snprintf(buffer, PATH_MAX, "%s" ZONE_DOOR_PATH,
7080 	    zonecfg_get_root(), zone_name) < PATH_MAX);
7081 }
7082 
7083 /*
7084  * system daemons are not audited.  For the global zone, this occurs
7085  * "naturally" since init is started with the default audit
7086  * characteristics.  Since zoneadmd is a system daemon and it starts
7087  * init for a zone, it is necessary to clear out the audit
7088  * characteristics inherited from whomever started zoneadmd.  This is
7089  * indicated by the audit id, which is set from the ruid parameter of
7090  * adt_set_user(), below.
7091  */
7092 
7093 static void
7094 prepare_audit_context(const char *zone_name)
7095 {
7096 	adt_session_data_t	*ah;
7097 	char			*failure = gettext("audit failure: %s");
7098 
7099 	if (adt_start_session(&ah, NULL, 0)) {
7100 		zerror(zone_name, failure, strerror(errno));
7101 		return;
7102 	}
7103 	if (adt_set_user(ah, ADT_NO_AUDIT, ADT_NO_AUDIT,
7104 	    ADT_NO_AUDIT, ADT_NO_AUDIT, NULL, ADT_NEW)) {
7105 		zerror(zone_name, failure, strerror(errno));
7106 		(void) adt_end_session(ah);
7107 		return;
7108 	}
7109 	if (adt_set_proc(ah))
7110 		zerror(zone_name, failure, strerror(errno));
7111 
7112 	(void) adt_end_session(ah);
7113 }
7114 
7115 static int
7116 start_zoneadmd(const char *zone_name, boolean_t lock)
7117 {
7118 	char doorpath[PATH_MAX];
7119 	pid_t child_pid;
7120 	int error = -1;
7121 	int doorfd, lockfd;
7122 	struct door_info info;
7123 
7124 	if (!get_doorname(zone_name, doorpath))
7125 		return (-1);
7126 
7127 	if (lock)
7128 		if (zonecfg_grab_lock_file(zone_name, &lockfd) != Z_OK)
7129 			return (-1);
7130 
7131 	/*
7132 	 * Now that we have the lock, re-confirm that the daemon is
7133 	 * *not* up and working fine.  If it is still down, we have a green
7134 	 * light to start it.
7135 	 */
7136 	if ((doorfd = open(doorpath, O_RDONLY)) < 0) {
7137 		if (errno != ENOENT) {
7138 			zperror(doorpath);
7139 			goto out;
7140 		}
7141 	} else {
7142 		if (door_info(doorfd, &info) == 0 &&
7143 		    ((info.di_attributes & DOOR_REVOKED) == 0)) {
7144 			error = Z_OK;
7145 			(void) close(doorfd);
7146 			goto out;
7147 		}
7148 		(void) close(doorfd);
7149 	}
7150 
7151 	if ((child_pid = fork()) == -1) {
7152 		zperror(gettext("could not fork"));
7153 		goto out;
7154 	}
7155 
7156 	if (child_pid == 0) {
7157 		const char *argv[6], **ap;
7158 
7159 		/* child process */
7160 		prepare_audit_context(zone_name);
7161 
7162 		ap = argv;
7163 		*ap++ = "zoneadmd";
7164 		*ap++ = "-z";
7165 		*ap++ = zone_name;
7166 		if (zonecfg_in_alt_root()) {
7167 			*ap++ = "-R";
7168 			*ap++ = zonecfg_get_root();
7169 		}
7170 		*ap = NULL;
7171 
7172 		(void) execv("/usr/lib/zones/zoneadmd", (char * const *)argv);
7173 		/*
7174 		 * TRANSLATION_NOTE
7175 		 * zoneadmd is a literal that should not be translated.
7176 		 */
7177 		zperror(gettext("could not exec zoneadmd"));
7178 		_exit(1);
7179 	} else {
7180 		/* parent process */
7181 		pid_t retval;
7182 		int pstatus = 0;
7183 
7184 		do {
7185 			retval = waitpid(child_pid, &pstatus, 0);
7186 		} while (retval != child_pid);
7187 		if (WIFSIGNALED(pstatus) || (WIFEXITED(pstatus) &&
7188 		    WEXITSTATUS(pstatus) != 0)) {
7189 			zerror(zone_name, gettext("could not start %s"),
7190 			    "zoneadmd");
7191 			goto out;
7192 		}
7193 	}
7194 	error = Z_OK;
7195 out:
7196 	if (lock)
7197 		zonecfg_release_lock_file(zone_name, lockfd);
7198 	return (error);
7199 }
7200 
7201 int
7202 zonecfg_ping_zoneadmd(const char *zone_name)
7203 {
7204 	char doorpath[PATH_MAX];
7205 	int doorfd;
7206 	struct door_info info;
7207 
7208 	if (!get_doorname(zone_name, doorpath))
7209 		return (-1);
7210 
7211 	if ((doorfd = open(doorpath, O_RDONLY)) < 0) {
7212 		return (-1);
7213 	}
7214 	if (door_info(doorfd, &info) == 0 &&
7215 	    ((info.di_attributes & DOOR_REVOKED) == 0)) {
7216 		(void) close(doorfd);
7217 		return (Z_OK);
7218 	}
7219 	(void) close(doorfd);
7220 	return (-1);
7221 }
7222 
7223 int
7224 zonecfg_call_zoneadmd(const char *zone_name, zone_cmd_arg_t *arg, char *locale,
7225     boolean_t lock)
7226 {
7227 	char doorpath[PATH_MAX];
7228 	int doorfd, result;
7229 	door_arg_t darg;
7230 
7231 	zoneid_t zoneid;
7232 	uint64_t uniqid = 0;
7233 
7234 	zone_cmd_rval_t *rvalp;
7235 	size_t rlen;
7236 	char *cp, *errbuf;
7237 
7238 	rlen = getpagesize();
7239 	if ((rvalp = malloc(rlen)) == NULL) {
7240 		zerror(zone_name, gettext("failed to allocate %lu bytes: %s"),
7241 		    rlen, strerror(errno));
7242 		return (-1);
7243 	}
7244 
7245 	if ((zoneid = getzoneidbyname(zone_name)) != ZONE_ID_UNDEFINED) {
7246 		(void) zone_getattr(zoneid, ZONE_ATTR_UNIQID, &uniqid,
7247 		    sizeof (uniqid));
7248 	}
7249 	arg->uniqid = uniqid;
7250 	(void) strlcpy(arg->locale, locale, sizeof (arg->locale));
7251 	if (!get_doorname(zone_name, doorpath)) {
7252 		zerror(zone_name, gettext("alternate root path is too long"));
7253 		free(rvalp);
7254 		return (-1);
7255 	}
7256 
7257 	/*
7258 	 * Loop trying to start zoneadmd; if something goes seriously
7259 	 * wrong we break out and fail.
7260 	 */
7261 	for (;;) {
7262 		if (start_zoneadmd(zone_name, lock) != Z_OK)
7263 			break;
7264 
7265 		if ((doorfd = open(doorpath, O_RDONLY)) < 0) {
7266 			zperror(gettext("failed to open zone door"));
7267 			break;
7268 		}
7269 
7270 		darg.data_ptr = (char *)arg;
7271 		darg.data_size = sizeof (*arg);
7272 		darg.desc_ptr = NULL;
7273 		darg.desc_num = 0;
7274 		darg.rbuf = (char *)rvalp;
7275 		darg.rsize = rlen;
7276 		if (door_call(doorfd, &darg) != 0) {
7277 			(void) close(doorfd);
7278 			/*
7279 			 * We'll get EBADF if the door has been revoked.
7280 			 */
7281 			if (errno != EBADF) {
7282 				zperror(gettext("door_call failed"));
7283 				break;
7284 			}
7285 			continue;	/* take another lap */
7286 		}
7287 		(void) close(doorfd);
7288 
7289 		if (darg.data_size == 0) {
7290 			/* Door server is going away; kick it again. */
7291 			continue;
7292 		}
7293 
7294 		errbuf = rvalp->errbuf;
7295 		while (*errbuf != '\0') {
7296 			/*
7297 			 * Remove any newlines since zerror()
7298 			 * will append one automatically.
7299 			 */
7300 			cp = strchr(errbuf, '\n');
7301 			if (cp != NULL)
7302 				*cp = '\0';
7303 			zerror(zone_name, "%s", errbuf);
7304 			if (cp == NULL)
7305 				break;
7306 			errbuf = cp + 1;
7307 		}
7308 		result = rvalp->rval == 0 ? 0 : -1;
7309 		free(rvalp);
7310 		return (result);
7311 	}
7312 
7313 	free(rvalp);
7314 	return (-1);
7315 }
7316