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 2017 Nexenta Systems, Inc.
24 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
25 * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
26 * Copyright 2015 RackTop Systems.
27 * Copyright (c) 2016, Intel Corporation.
28 */
29
30/*
31 * Pool import support functions.
32 *
33 * Used by zpool, ztest, zdb, and zhack to locate importable configs. Since
34 * these commands are expected to run in the global zone, we can assume
35 * that the devices are all readable when called.
36 *
37 * To import a pool, we rely on reading the configuration information from the
38 * ZFS label of each device.  If we successfully read the label, then we
39 * organize the configuration information in the following hierarchy:
40 *
41 *	pool guid -> toplevel vdev guid -> label txg
42 *
43 * Duplicate entries matching this same tuple will be discarded.  Once we have
44 * examined every device, we pick the best label txg config for each toplevel
45 * vdev.  We then arrange these toplevel vdevs into a complete pool config, and
46 * update any paths that have changed.  Finally, we attempt to import the pool
47 * using our derived config, and record the results.
48 */
49
50#include <stdio.h>
51#include <stdarg.h>
52#include <assert.h>
53#include <ctype.h>
54#include <devid.h>
55#include <dirent.h>
56#include <errno.h>
57#include <libintl.h>
58#include <stddef.h>
59#include <stdlib.h>
60#include <string.h>
61#include <sys/stat.h>
62#include <unistd.h>
63#include <fcntl.h>
64#include <sys/vtoc.h>
65#include <sys/dktp/fdisk.h>
66#include <sys/efi_partition.h>
67#include <sys/vdev_impl.h>
68#include <sys/fs/zfs.h>
69
70#include <thread_pool.h>
71#include <libzutil.h>
72#include <libnvpair.h>
73
74#include "zutil_import.h"
75
76#ifdef NDEBUG
77#define	verify(EX)	((void)(EX))
78#else
79#define	verify(EX)	assert(EX)
80#endif
81
82/*PRINTFLIKE2*/
83static void
84zutil_error_aux(libpc_handle_t *hdl, const char *fmt, ...)
85{
86	va_list ap;
87
88	va_start(ap, fmt);
89
90	(void) vsnprintf(hdl->lpc_desc, sizeof (hdl->lpc_desc), fmt, ap);
91	hdl->lpc_desc_active = B_TRUE;
92
93	va_end(ap);
94}
95
96static void
97zutil_verror(libpc_handle_t *hdl, const char *error, const char *fmt,
98    va_list ap)
99{
100	char action[1024];
101
102	(void) vsnprintf(action, sizeof (action), fmt, ap);
103
104	if (hdl->lpc_desc_active)
105		hdl->lpc_desc_active = B_FALSE;
106	else
107		hdl->lpc_desc[0] = '\0';
108
109	if (hdl->lpc_printerr) {
110		if (hdl->lpc_desc[0] != '\0')
111			error = hdl->lpc_desc;
112
113		(void) fprintf(stderr, "%s: %s\n", action, error);
114	}
115}
116
117/*PRINTFLIKE3*/
118static int
119zutil_error_fmt(libpc_handle_t *hdl, const char *error, const char *fmt, ...)
120{
121	va_list ap;
122
123	va_start(ap, fmt);
124
125	zutil_verror(hdl, error, fmt, ap);
126
127	va_end(ap);
128
129	return (-1);
130}
131
132static int
133zutil_error(libpc_handle_t *hdl, const char *error, const char *msg)
134{
135	return (zutil_error_fmt(hdl, error, "%s", msg));
136}
137
138static int
139zutil_no_memory(libpc_handle_t *hdl)
140{
141	(void) zutil_error(hdl, EZFS_NOMEM, "internal error");
142	exit(1);
143}
144
145void *
146zutil_alloc(libpc_handle_t *hdl, size_t size)
147{
148	void *data;
149
150	if ((data = calloc(1, size)) == NULL)
151		(void) zutil_no_memory(hdl);
152
153	return (data);
154}
155
156char *
157zutil_strdup(libpc_handle_t *hdl, const char *str)
158{
159	char *ret;
160
161	if ((ret = strdup(str)) == NULL)
162		(void) zutil_no_memory(hdl);
163
164	return (ret);
165}
166
167/*
168 * Intermediate structures used to gather configuration information.
169 */
170typedef struct config_entry {
171	uint64_t		ce_txg;
172	nvlist_t		*ce_config;
173	struct config_entry	*ce_next;
174} config_entry_t;
175
176typedef struct vdev_entry {
177	uint64_t		ve_guid;
178	config_entry_t		*ve_configs;
179	struct vdev_entry	*ve_next;
180} vdev_entry_t;
181
182typedef struct pool_entry {
183	uint64_t		pe_guid;
184	vdev_entry_t		*pe_vdevs;
185	struct pool_entry	*pe_next;
186} pool_entry_t;
187
188typedef struct name_entry {
189	char			*ne_name;
190	uint64_t		ne_guid;
191	struct name_entry	*ne_next;
192} name_entry_t;
193
194typedef struct pool_list {
195	pool_entry_t		*pools;
196	name_entry_t		*names;
197} pool_list_t;
198
199/*
200 * Go through and fix up any path and/or devid information for the given vdev
201 * configuration.
202 */
203static int
204fix_paths(nvlist_t *nv, name_entry_t *names)
205{
206	nvlist_t **child;
207	uint_t c, children;
208	uint64_t guid;
209	name_entry_t *ne, *best;
210	char *path, *devid;
211	int matched;
212
213	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
214	    &child, &children) == 0) {
215		for (c = 0; c < children; c++)
216			if (fix_paths(child[c], names) != 0)
217				return (-1);
218		return (0);
219	}
220
221	/*
222	 * This is a leaf (file or disk) vdev.  In either case, go through
223	 * the name list and see if we find a matching guid.  If so, replace
224	 * the path and see if we can calculate a new devid.
225	 *
226	 * There may be multiple names associated with a particular guid, in
227	 * which case we have overlapping slices or multiple paths to the same
228	 * disk.  If this is the case, then we want to pick the path that is
229	 * the most similar to the original, where "most similar" is the number
230	 * of matching characters starting from the end of the path.  This will
231	 * preserve slice numbers even if the disks have been reorganized, and
232	 * will also catch preferred disk names if multiple paths exist.
233	 */
234	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
235	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
236		path = NULL;
237
238	matched = 0;
239	best = NULL;
240	for (ne = names; ne != NULL; ne = ne->ne_next) {
241		if (ne->ne_guid == guid) {
242			const char *src, *dst;
243			int count;
244
245			if (path == NULL) {
246				best = ne;
247				break;
248			}
249
250			src = ne->ne_name + strlen(ne->ne_name) - 1;
251			dst = path + strlen(path) - 1;
252			for (count = 0; src >= ne->ne_name && dst >= path;
253			    src--, dst--, count++)
254				if (*src != *dst)
255					break;
256
257			/*
258			 * At this point, 'count' is the number of characters
259			 * matched from the end.
260			 */
261			if (count > matched || best == NULL) {
262				best = ne;
263				matched = count;
264			}
265		}
266	}
267
268	if (best == NULL)
269		return (0);
270
271	if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
272		return (-1);
273
274	if ((devid = devid_str_from_path(best->ne_name)) == NULL) {
275		(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
276	} else {
277		if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) {
278			devid_str_free(devid);
279			return (-1);
280		}
281		devid_str_free(devid);
282	}
283
284	return (0);
285}
286
287/*
288 * Add the given configuration to the list of known devices.
289 */
290static int
291add_config(libpc_handle_t *hdl, pool_list_t *pl, const char *path,
292    int order, int num_labels, nvlist_t *config)
293{
294	uint64_t pool_guid, vdev_guid, top_guid, txg, state;
295	pool_entry_t *pe;
296	vdev_entry_t *ve;
297	config_entry_t *ce;
298	name_entry_t *ne;
299
300	/*
301	 * If this is a hot spare not currently in use or level 2 cache
302	 * device, add it to the list of names to translate, but don't do
303	 * anything else.
304	 */
305	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
306	    &state) == 0 &&
307	    (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
308	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
309		if ((ne = zutil_alloc(hdl, sizeof (name_entry_t))) == NULL)
310			return (-1);
311
312		if ((ne->ne_name = zutil_strdup(hdl, path)) == NULL) {
313			free(ne);
314			return (-1);
315		}
316
317		ne->ne_guid = vdev_guid;
318		ne->ne_next = pl->names;
319		pl->names = ne;
320
321		return (0);
322	}
323
324	/*
325	 * If we have a valid config but cannot read any of these fields, then
326	 * it means we have a half-initialized label.  In vdev_label_init()
327	 * we write a label with txg == 0 so that we can identify the device
328	 * in case the user refers to the same disk later on.  If we fail to
329	 * create the pool, we'll be left with a label in this state
330	 * which should not be considered part of a valid pool.
331	 */
332	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
333	    &pool_guid) != 0 ||
334	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
335	    &vdev_guid) != 0 ||
336	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
337	    &top_guid) != 0 ||
338	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
339	    &txg) != 0 || txg == 0) {
340		return (0);
341	}
342
343	/*
344	 * First, see if we know about this pool.  If not, then add it to the
345	 * list of known pools.
346	 */
347	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
348		if (pe->pe_guid == pool_guid)
349			break;
350	}
351
352	if (pe == NULL) {
353		if ((pe = zutil_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
354			return (-1);
355		}
356		pe->pe_guid = pool_guid;
357		pe->pe_next = pl->pools;
358		pl->pools = pe;
359	}
360
361	/*
362	 * Second, see if we know about this toplevel vdev.  Add it if its
363	 * missing.
364	 */
365	for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
366		if (ve->ve_guid == top_guid)
367			break;
368	}
369
370	if (ve == NULL) {
371		if ((ve = zutil_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
372			return (-1);
373		}
374		ve->ve_guid = top_guid;
375		ve->ve_next = pe->pe_vdevs;
376		pe->pe_vdevs = ve;
377	}
378
379	/*
380	 * Third, see if we have a config with a matching transaction group.  If
381	 * so, then we do nothing.  Otherwise, add it to the list of known
382	 * configs.
383	 */
384	for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
385		if (ce->ce_txg == txg)
386			break;
387	}
388
389	if (ce == NULL) {
390		if ((ce = zutil_alloc(hdl, sizeof (config_entry_t))) == NULL) {
391			return (-1);
392		}
393		ce->ce_txg = txg;
394		ce->ce_config = fnvlist_dup(config);
395		ce->ce_next = ve->ve_configs;
396		ve->ve_configs = ce;
397	}
398
399	/*
400	 * At this point we've successfully added our config to the list of
401	 * known configs.  The last thing to do is add the vdev guid -> path
402	 * mappings so that we can fix up the configuration as necessary before
403	 * doing the import.
404	 */
405	if ((ne = zutil_alloc(hdl, sizeof (name_entry_t))) == NULL)
406		return (-1);
407
408	if ((ne->ne_name = zutil_strdup(hdl, path)) == NULL) {
409		free(ne);
410		return (-1);
411	}
412
413	ne->ne_guid = vdev_guid;
414	ne->ne_next = pl->names;
415	pl->names = ne;
416
417	return (0);
418}
419
420/*
421 * Returns true if the named pool matches the given GUID.
422 */
423static int
424zutil_pool_active(libpc_handle_t *hdl, const char *name, uint64_t guid,
425    boolean_t *isactive)
426{
427	ASSERT(hdl->lpc_ops->pco_pool_active != NULL);
428
429	int error = hdl->lpc_ops->pco_pool_active(hdl->lpc_lib_handle, name,
430	    guid, isactive);
431
432	return (error);
433}
434
435static nvlist_t *
436zutil_refresh_config(libpc_handle_t *hdl, nvlist_t *tryconfig)
437{
438	ASSERT(hdl->lpc_ops->pco_refresh_config != NULL);
439
440	return (hdl->lpc_ops->pco_refresh_config(hdl->lpc_lib_handle,
441	    tryconfig));
442}
443
444/*
445 * Determine if the vdev id is a hole in the namespace.
446 */
447static boolean_t
448vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id)
449{
450	for (int c = 0; c < holes; c++) {
451
452		/* Top-level is a hole */
453		if (hole_array[c] == id)
454			return (B_TRUE);
455	}
456	return (B_FALSE);
457}
458
459/*
460 * Convert our list of pools into the definitive set of configurations.  We
461 * start by picking the best config for each toplevel vdev.  Once that's done,
462 * we assemble the toplevel vdevs into a full config for the pool.  We make a
463 * pass to fix up any incorrect paths, and then add it to the main list to
464 * return to the user.
465 */
466static nvlist_t *
467get_configs(libpc_handle_t *hdl, pool_list_t *pl, boolean_t active_ok,
468    nvlist_t *policy)
469{
470	pool_entry_t *pe;
471	vdev_entry_t *ve;
472	config_entry_t *ce;
473	nvlist_t *ret = NULL, *config = NULL, *tmp = NULL, *nvtop, *nvroot;
474	nvlist_t **spares, **l2cache;
475	uint_t i, nspares, nl2cache;
476	boolean_t config_seen;
477	uint64_t best_txg;
478	char *name, *hostname = NULL;
479	uint64_t guid;
480	uint_t children = 0;
481	nvlist_t **child = NULL;
482	uint_t holes;
483	uint64_t *hole_array, max_id;
484	uint_t c;
485	boolean_t isactive;
486	uint64_t hostid;
487	nvlist_t *nvl;
488	boolean_t found_one = B_FALSE;
489	boolean_t valid_top_config = B_FALSE;
490
491	if (nvlist_alloc(&ret, 0, 0) != 0)
492		goto nomem;
493
494	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
495		uint64_t id, max_txg = 0;
496
497		if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
498			goto nomem;
499		config_seen = B_FALSE;
500
501		/*
502		 * Iterate over all toplevel vdevs.  Grab the pool configuration
503		 * from the first one we find, and then go through the rest and
504		 * add them as necessary to the 'vdevs' member of the config.
505		 */
506		for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
507
508			/*
509			 * Determine the best configuration for this vdev by
510			 * selecting the config with the latest transaction
511			 * group.
512			 */
513			best_txg = 0;
514			for (ce = ve->ve_configs; ce != NULL;
515			    ce = ce->ce_next) {
516
517				if (ce->ce_txg > best_txg) {
518					tmp = ce->ce_config;
519					best_txg = ce->ce_txg;
520				}
521			}
522
523			/*
524			 * We rely on the fact that the max txg for the
525			 * pool will contain the most up-to-date information
526			 * about the valid top-levels in the vdev namespace.
527			 */
528			if (best_txg > max_txg) {
529				(void) nvlist_remove(config,
530				    ZPOOL_CONFIG_VDEV_CHILDREN,
531				    DATA_TYPE_UINT64);
532				(void) nvlist_remove(config,
533				    ZPOOL_CONFIG_HOLE_ARRAY,
534				    DATA_TYPE_UINT64_ARRAY);
535
536				max_txg = best_txg;
537				hole_array = NULL;
538				holes = 0;
539				max_id = 0;
540				valid_top_config = B_FALSE;
541
542				if (nvlist_lookup_uint64(tmp,
543				    ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) {
544					verify(nvlist_add_uint64(config,
545					    ZPOOL_CONFIG_VDEV_CHILDREN,
546					    max_id) == 0);
547					valid_top_config = B_TRUE;
548				}
549
550				if (nvlist_lookup_uint64_array(tmp,
551				    ZPOOL_CONFIG_HOLE_ARRAY, &hole_array,
552				    &holes) == 0) {
553					verify(nvlist_add_uint64_array(config,
554					    ZPOOL_CONFIG_HOLE_ARRAY,
555					    hole_array, holes) == 0);
556				}
557			}
558
559			if (!config_seen) {
560				/*
561				 * Copy the relevant pieces of data to the pool
562				 * configuration:
563				 *
564				 *	version
565				 *	pool guid
566				 *	name
567				 *	comment (if available)
568				 *	pool state
569				 *	hostid (if available)
570				 *	hostname (if available)
571				 */
572				uint64_t state, version;
573				char *comment = NULL;
574
575				version = fnvlist_lookup_uint64(tmp,
576				    ZPOOL_CONFIG_VERSION);
577				fnvlist_add_uint64(config,
578				    ZPOOL_CONFIG_VERSION, version);
579				guid = fnvlist_lookup_uint64(tmp,
580				    ZPOOL_CONFIG_POOL_GUID);
581				fnvlist_add_uint64(config,
582				    ZPOOL_CONFIG_POOL_GUID, guid);
583				name = fnvlist_lookup_string(tmp,
584				    ZPOOL_CONFIG_POOL_NAME);
585				fnvlist_add_string(config,
586				    ZPOOL_CONFIG_POOL_NAME, name);
587
588				if (nvlist_lookup_string(tmp,
589				    ZPOOL_CONFIG_COMMENT, &comment) == 0)
590					fnvlist_add_string(config,
591					    ZPOOL_CONFIG_COMMENT, comment);
592
593				state = fnvlist_lookup_uint64(tmp,
594				    ZPOOL_CONFIG_POOL_STATE);
595				fnvlist_add_uint64(config,
596				    ZPOOL_CONFIG_POOL_STATE, state);
597
598				hostid = 0;
599				if (nvlist_lookup_uint64(tmp,
600				    ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
601					fnvlist_add_uint64(config,
602					    ZPOOL_CONFIG_HOSTID, hostid);
603					hostname = fnvlist_lookup_string(tmp,
604					    ZPOOL_CONFIG_HOSTNAME);
605					fnvlist_add_string(config,
606					    ZPOOL_CONFIG_HOSTNAME, hostname);
607				}
608
609				config_seen = B_TRUE;
610			}
611
612			/*
613			 * Add this top-level vdev to the child array.
614			 */
615			verify(nvlist_lookup_nvlist(tmp,
616			    ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
617			verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
618			    &id) == 0);
619
620			if (id >= children) {
621				nvlist_t **newchild;
622
623				newchild = zutil_alloc(hdl, (id + 1) *
624				    sizeof (nvlist_t *));
625				if (newchild == NULL)
626					goto nomem;
627
628				for (c = 0; c < children; c++)
629					newchild[c] = child[c];
630
631				free(child);
632				child = newchild;
633				children = id + 1;
634			}
635			if (nvlist_dup(nvtop, &child[id], 0) != 0)
636				goto nomem;
637
638		}
639
640		/*
641		 * If we have information about all the top-levels then
642		 * clean up the nvlist which we've constructed. This
643		 * means removing any extraneous devices that are
644		 * beyond the valid range or adding devices to the end
645		 * of our array which appear to be missing.
646		 */
647		if (valid_top_config) {
648			if (max_id < children) {
649				for (c = max_id; c < children; c++)
650					nvlist_free(child[c]);
651				children = max_id;
652			} else if (max_id > children) {
653				nvlist_t **newchild;
654
655				newchild = zutil_alloc(hdl, (max_id) *
656				    sizeof (nvlist_t *));
657				if (newchild == NULL)
658					goto nomem;
659
660				for (c = 0; c < children; c++)
661					newchild[c] = child[c];
662
663				free(child);
664				child = newchild;
665				children = max_id;
666			}
667		}
668
669		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
670		    &guid) == 0);
671
672		/*
673		 * The vdev namespace may contain holes as a result of
674		 * device removal. We must add them back into the vdev
675		 * tree before we process any missing devices.
676		 */
677		if (holes > 0) {
678			ASSERT(valid_top_config);
679
680			for (c = 0; c < children; c++) {
681				nvlist_t *holey;
682
683				if (child[c] != NULL ||
684				    !vdev_is_hole(hole_array, holes, c))
685					continue;
686
687				if (nvlist_alloc(&holey, NV_UNIQUE_NAME,
688				    0) != 0)
689					goto nomem;
690
691				/*
692				 * Holes in the namespace are treated as
693				 * "hole" top-level vdevs and have a
694				 * special flag set on them.
695				 */
696				if (nvlist_add_string(holey,
697				    ZPOOL_CONFIG_TYPE,
698				    VDEV_TYPE_HOLE) != 0 ||
699				    nvlist_add_uint64(holey,
700				    ZPOOL_CONFIG_ID, c) != 0 ||
701				    nvlist_add_uint64(holey,
702				    ZPOOL_CONFIG_GUID, 0ULL) != 0) {
703					nvlist_free(holey);
704					goto nomem;
705				}
706				child[c] = holey;
707			}
708		}
709
710		/*
711		 * Look for any missing top-level vdevs.  If this is the case,
712		 * create a faked up 'missing' vdev as a placeholder.  We cannot
713		 * simply compress the child array, because the kernel performs
714		 * certain checks to make sure the vdev IDs match their location
715		 * in the configuration.
716		 */
717		for (c = 0; c < children; c++) {
718			if (child[c] == NULL) {
719				nvlist_t *missing;
720				if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
721				    0) != 0)
722					goto nomem;
723				if (nvlist_add_string(missing,
724				    ZPOOL_CONFIG_TYPE,
725				    VDEV_TYPE_MISSING) != 0 ||
726				    nvlist_add_uint64(missing,
727				    ZPOOL_CONFIG_ID, c) != 0 ||
728				    nvlist_add_uint64(missing,
729				    ZPOOL_CONFIG_GUID, 0ULL) != 0) {
730					nvlist_free(missing);
731					goto nomem;
732				}
733				child[c] = missing;
734			}
735		}
736
737		/*
738		 * Put all of this pool's top-level vdevs into a root vdev.
739		 */
740		if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
741			goto nomem;
742		if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
743		    VDEV_TYPE_ROOT) != 0 ||
744		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
745		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
746		    nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
747		    child, children) != 0) {
748			nvlist_free(nvroot);
749			goto nomem;
750		}
751
752		for (c = 0; c < children; c++)
753			nvlist_free(child[c]);
754		free(child);
755		children = 0;
756		child = NULL;
757
758		/*
759		 * Go through and fix up any paths and/or devids based on our
760		 * known list of vdev GUID -> path mappings.
761		 */
762		if (fix_paths(nvroot, pl->names) != 0) {
763			nvlist_free(nvroot);
764			goto nomem;
765		}
766
767		/*
768		 * Add the root vdev to this pool's configuration.
769		 */
770		if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
771		    nvroot) != 0) {
772			nvlist_free(nvroot);
773			goto nomem;
774		}
775		nvlist_free(nvroot);
776
777		/*
778		 * zdb uses this path to report on active pools that were
779		 * imported or created using -R.
780		 */
781		if (active_ok)
782			goto add_pool;
783
784		/*
785		 * Determine if this pool is currently active, in which case we
786		 * can't actually import it.
787		 */
788		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
789		    &name) == 0);
790		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
791		    &guid) == 0);
792
793		if (zutil_pool_active(hdl, name, guid, &isactive) != 0)
794			goto error;
795
796		if (isactive) {
797			nvlist_free(config);
798			config = NULL;
799			continue;
800		}
801
802		if (policy != NULL) {
803			if (nvlist_add_nvlist(config, ZPOOL_LOAD_POLICY,
804			    policy) != 0)
805				goto nomem;
806		}
807
808		if ((nvl = zutil_refresh_config(hdl, config)) == NULL) {
809			nvlist_free(config);
810			config = NULL;
811			continue;
812		}
813
814		nvlist_free(config);
815		config = nvl;
816
817		/*
818		 * Go through and update the paths for spares, now that we have
819		 * them.
820		 */
821		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
822		    &nvroot) == 0);
823		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
824		    &spares, &nspares) == 0) {
825			for (i = 0; i < nspares; i++) {
826				if (fix_paths(spares[i], pl->names) != 0)
827					goto nomem;
828			}
829		}
830
831		/*
832		 * Update the paths for l2cache devices.
833		 */
834		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
835		    &l2cache, &nl2cache) == 0) {
836			for (i = 0; i < nl2cache; i++) {
837				if (fix_paths(l2cache[i], pl->names) != 0)
838					goto nomem;
839			}
840		}
841
842		/*
843		 * Restore the original information read from the actual label.
844		 */
845		(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
846		    DATA_TYPE_UINT64);
847		(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
848		    DATA_TYPE_STRING);
849		if (hostid != 0) {
850			verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
851			    hostid) == 0);
852			verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
853			    hostname) == 0);
854		}
855
856add_pool:
857		/*
858		 * Add this pool to the list of configs.
859		 */
860		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
861		    &name) == 0);
862		if (nvlist_add_nvlist(ret, name, config) != 0)
863			goto nomem;
864
865		found_one = B_TRUE;
866		nvlist_free(config);
867		config = NULL;
868	}
869
870	if (!found_one) {
871		nvlist_free(ret);
872		ret = NULL;
873	}
874
875	return (ret);
876
877nomem:
878	(void) zutil_no_memory(hdl);
879error:
880	nvlist_free(config);
881	nvlist_free(ret);
882	for (c = 0; c < children; c++)
883		nvlist_free(child[c]);
884	free(child);
885
886	return (NULL);
887}
888
889/*
890 * Return the offset of the given label.
891 */
892static uint64_t
893label_offset(uint64_t size, int l)
894{
895	ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
896	return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
897	    0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
898}
899
900/*
901 * Given a file descriptor, read the label information and return an nvlist
902 * describing the configuration, if there is one. The number of valid
903 * labels found will be returned in num_labels when non-NULL.
904 */
905int
906zpool_read_label(int fd, nvlist_t **config, int *num_labels)
907{
908	struct stat64 statbuf;
909	int l, count = 0;
910	vdev_label_t *label;
911	nvlist_t *expected_config = NULL;
912	uint64_t expected_guid = 0, size;
913
914	*config = NULL;
915
916	if (fstat64(fd, &statbuf) == -1)
917		return (-1);
918	size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
919
920	if ((label = malloc(sizeof (vdev_label_t))) == NULL)
921		return (-1);
922
923	for (l = 0; l < VDEV_LABELS; l++) {
924		uint64_t state, guid, txg;
925
926		if (pread64(fd, label, sizeof (vdev_label_t),
927		    label_offset(size, l)) != sizeof (vdev_label_t))
928			continue;
929
930		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
931		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
932			continue;
933
934		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_GUID,
935		    &guid) != 0 || guid == 0) {
936			nvlist_free(*config);
937			continue;
938		}
939
940		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
941		    &state) != 0 || state > POOL_STATE_L2CACHE) {
942			nvlist_free(*config);
943			continue;
944		}
945
946		if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
947		    (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
948		    &txg) != 0 || txg == 0)) {
949			nvlist_free(*config);
950			continue;
951		}
952
953		if (expected_guid) {
954			if (expected_guid == guid)
955				count++;
956
957			nvlist_free(*config);
958		} else {
959			expected_config = *config;
960			expected_guid = guid;
961			count++;
962		}
963	}
964
965	if (num_labels != NULL)
966		*num_labels = count;
967
968	free(label);
969	*config = expected_config;
970
971	if (count == 0) {
972		errno = ENOENT;
973		return (-1);
974	}
975
976	return (0);
977}
978
979static int
980slice_cache_compare(const void *arg1, const void *arg2)
981{
982	const char  *nm1 = ((rdsk_node_t *)arg1)->rn_name;
983	const char  *nm2 = ((rdsk_node_t *)arg2)->rn_name;
984	char *nm1slice, *nm2slice;
985	int rv;
986
987	/*
988	 * slices zero and two are the most likely to provide results,
989	 * so put those first
990	 */
991	nm1slice = strstr(nm1, "s0");
992	nm2slice = strstr(nm2, "s0");
993	if (nm1slice && !nm2slice) {
994		return (-1);
995	}
996	if (!nm1slice && nm2slice) {
997		return (1);
998	}
999	nm1slice = strstr(nm1, "s2");
1000	nm2slice = strstr(nm2, "s2");
1001	if (nm1slice && !nm2slice) {
1002		return (-1);
1003	}
1004	if (!nm1slice && nm2slice) {
1005		return (1);
1006	}
1007
1008	rv = strcmp(nm1, nm2);
1009	if (rv == 0)
1010		return (0);
1011	return (rv > 0 ? 1 : -1);
1012}
1013
1014static void
1015check_one_slice(avl_tree_t *r, char *diskname, uint_t partno,
1016    diskaddr_t size, uint_t blksz)
1017{
1018	rdsk_node_t tmpnode;
1019	rdsk_node_t *node;
1020	char sname[MAXNAMELEN];
1021
1022	tmpnode.rn_name = &sname[0];
1023	(void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u",
1024	    diskname, partno);
1025	/*
1026	 * protect against division by zero for disk labels that
1027	 * contain a bogus sector size
1028	 */
1029	if (blksz == 0)
1030		blksz = DEV_BSIZE;
1031	/* too small to contain a zpool? */
1032	if ((size < (SPA_MINDEVSIZE / blksz)) &&
1033	    (node = avl_find(r, &tmpnode, NULL)))
1034		node->rn_nozpool = B_TRUE;
1035}
1036
1037static void
1038nozpool_all_slices(avl_tree_t *r, const char *sname)
1039{
1040	char diskname[MAXNAMELEN];
1041	char *ptr;
1042	int i;
1043
1044	(void) strncpy(diskname, sname, MAXNAMELEN);
1045	if (((ptr = strrchr(diskname, 's')) == NULL) &&
1046	    ((ptr = strrchr(diskname, 'p')) == NULL))
1047		return;
1048	ptr[0] = 's';
1049	ptr[1] = '\0';
1050	for (i = 0; i < NDKMAP; i++)
1051		check_one_slice(r, diskname, i, 0, 1);
1052	ptr[0] = 'p';
1053	for (i = 0; i <= FD_NUMPART; i++)
1054		check_one_slice(r, diskname, i, 0, 1);
1055}
1056
1057static void
1058check_slices(avl_tree_t *r, int fd, const char *sname)
1059{
1060	struct extvtoc vtoc;
1061	struct dk_gpt *gpt;
1062	char diskname[MAXNAMELEN];
1063	char *ptr;
1064	int i;
1065
1066	(void) strncpy(diskname, sname, MAXNAMELEN);
1067	if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1]))
1068		return;
1069	ptr[1] = '\0';
1070
1071	if (read_extvtoc(fd, &vtoc) >= 0) {
1072		for (i = 0; i < NDKMAP; i++)
1073			check_one_slice(r, diskname, i,
1074			    vtoc.v_part[i].p_size, vtoc.v_sectorsz);
1075	} else if (efi_alloc_and_read(fd, &gpt) >= 0) {
1076		/*
1077		 * on x86 we'll still have leftover links that point
1078		 * to slices s[9-15], so use NDKMAP instead
1079		 */
1080		for (i = 0; i < NDKMAP; i++)
1081			check_one_slice(r, diskname, i,
1082			    gpt->efi_parts[i].p_size, gpt->efi_lbasize);
1083		/* nodes p[1-4] are never used with EFI labels */
1084		ptr[0] = 'p';
1085		for (i = 1; i <= FD_NUMPART; i++)
1086			check_one_slice(r, diskname, i, 0, 1);
1087		efi_free(gpt);
1088	}
1089}
1090
1091void
1092zpool_open_func(void *arg)
1093{
1094	rdsk_node_t *rn = arg;
1095	struct stat64 statbuf;
1096	nvlist_t *config;
1097	int error;
1098	int num_labels = 0;
1099	int fd;
1100
1101	if (rn->rn_nozpool)
1102		return;
1103	if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) {
1104		/* symlink to a device that's no longer there */
1105		if (errno == ENOENT)
1106			nozpool_all_slices(rn->rn_avl, rn->rn_name);
1107		return;
1108	}
1109	/*
1110	 * Ignore failed stats.  We only want regular
1111	 * files, character devs and block devs.
1112	 */
1113	if (fstat64(fd, &statbuf) != 0 ||
1114	    (!S_ISREG(statbuf.st_mode) &&
1115	    !S_ISCHR(statbuf.st_mode) &&
1116	    !S_ISBLK(statbuf.st_mode))) {
1117		(void) close(fd);
1118		return;
1119	}
1120	/* this file is too small to hold a zpool */
1121	if (S_ISREG(statbuf.st_mode) &&
1122	    statbuf.st_size < SPA_MINDEVSIZE) {
1123		(void) close(fd);
1124		return;
1125	} else if (!S_ISREG(statbuf.st_mode)) {
1126		/*
1127		 * Try to read the disk label first so we don't have to
1128		 * open a bunch of minor nodes that can't have a zpool.
1129		 */
1130		check_slices(rn->rn_avl, fd, rn->rn_name);
1131	}
1132
1133	error = zpool_read_label(fd, &config, &num_labels);
1134	if (error != 0) {
1135		(void) close(fd);
1136		return;
1137	}
1138
1139	if (num_labels == 0) {
1140		(void) close(fd);
1141		nvlist_free(config);
1142		return;
1143	}
1144
1145	(void) close(fd);
1146
1147	rn->rn_config = config;
1148	rn->rn_num_labels = num_labels;
1149}
1150
1151/*
1152 * Given a list of directories to search, find all pools stored on disk.  This
1153 * includes partial pools which are not available to import.  If no args are
1154 * given (argc is 0), then the default directory (/dev/dsk) is searched.
1155 * poolname or guid (but not both) are provided by the caller when trying
1156 * to import a specific pool.
1157 */
1158static nvlist_t *
1159zpool_find_import_impl(libpc_handle_t *hdl, importargs_t *iarg)
1160{
1161	int i, dirs = iarg->paths;
1162	struct dirent64 *dp;
1163	char path[MAXPATHLEN];
1164	char *end, **dir = iarg->path;
1165	size_t pathleft;
1166	nvlist_t *ret = NULL;
1167	static char *default_dir = ZFS_DISK_ROOT;
1168	pool_list_t pools = { 0 };
1169	pool_entry_t *pe, *penext;
1170	vdev_entry_t *ve, *venext;
1171	config_entry_t *ce, *cenext;
1172	name_entry_t *ne, *nenext;
1173	avl_tree_t slice_cache;
1174	rdsk_node_t *slice;
1175	void *cookie;
1176
1177	if (dirs == 0) {
1178		dirs = 1;
1179		dir = &default_dir;
1180	}
1181
1182	/*
1183	 * Go through and read the label configuration information from every
1184	 * possible device, organizing the information according to pool GUID
1185	 * and toplevel GUID.
1186	 */
1187	for (i = 0; i < dirs; i++) {
1188		tpool_t *t;
1189		char rdsk[MAXPATHLEN];
1190		int dfd;
1191		boolean_t config_failed = B_FALSE;
1192		DIR *dirp;
1193
1194		/* use realpath to normalize the path */
1195		if (realpath(dir[i], path) == 0) {
1196			(void) zutil_error_fmt(hdl, EZFS_BADPATH,
1197			    dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]);
1198			goto error;
1199		}
1200		end = &path[strlen(path)];
1201		*end++ = '/';
1202		*end = 0;
1203		pathleft = &path[sizeof (path)] - end;
1204
1205		/*
1206		 * Using raw devices instead of block devices when we're
1207		 * reading the labels skips a bunch of slow operations during
1208		 * close(2) processing, so we replace /dev/dsk with /dev/rdsk.
1209		 */
1210		if (strcmp(path, ZFS_DISK_ROOTD) == 0)
1211			(void) strlcpy(rdsk, ZFS_RDISK_ROOTD, sizeof (rdsk));
1212		else
1213			(void) strlcpy(rdsk, path, sizeof (rdsk));
1214
1215		if ((dfd = open64(rdsk, O_RDONLY)) < 0 ||
1216		    (dirp = fdopendir(dfd)) == NULL) {
1217			if (dfd >= 0)
1218				(void) close(dfd);
1219			zutil_error_aux(hdl, strerror(errno));
1220			(void) zutil_error_fmt(hdl, EZFS_BADPATH,
1221			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1222			    rdsk);
1223			goto error;
1224		}
1225
1226		avl_create(&slice_cache, slice_cache_compare,
1227		    sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node));
1228		/*
1229		 * This is not MT-safe, but we have no MT consumers of libzutil
1230		 */
1231		while ((dp = readdir64(dirp)) != NULL) {
1232			const char *name = dp->d_name;
1233			if (name[0] == '.' &&
1234			    (name[1] == 0 || (name[1] == '.' && name[2] == 0)))
1235				continue;
1236
1237			slice = zutil_alloc(hdl, sizeof (rdsk_node_t));
1238			slice->rn_name = zutil_strdup(hdl, name);
1239			slice->rn_avl = &slice_cache;
1240			slice->rn_dfd = dfd;
1241			slice->rn_hdl = hdl;
1242			slice->rn_nozpool = B_FALSE;
1243			avl_add(&slice_cache, slice);
1244		}
1245		/*
1246		 * create a thread pool to do all of this in parallel;
1247		 * rn_nozpool is not protected, so this is racy in that
1248		 * multiple tasks could decide that the same slice can
1249		 * not hold a zpool, which is benign.  Also choose
1250		 * double the number of processors; we hold a lot of
1251		 * locks in the kernel, so going beyond this doesn't
1252		 * buy us much.
1253		 */
1254		t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN),
1255		    0, NULL);
1256		for (slice = avl_first(&slice_cache); slice;
1257		    (slice = avl_walk(&slice_cache, slice,
1258		    AVL_AFTER)))
1259			(void) tpool_dispatch(t, zpool_open_func, slice);
1260		tpool_wait(t);
1261		tpool_destroy(t);
1262
1263		cookie = NULL;
1264		while ((slice = avl_destroy_nodes(&slice_cache,
1265		    &cookie)) != NULL) {
1266			if (slice->rn_config != NULL && !config_failed) {
1267				nvlist_t *config = slice->rn_config;
1268				boolean_t matched = B_TRUE;
1269
1270				if (iarg->poolname != NULL) {
1271					char *pname;
1272
1273					matched = nvlist_lookup_string(config,
1274					    ZPOOL_CONFIG_POOL_NAME,
1275					    &pname) == 0 &&
1276					    strcmp(iarg->poolname, pname) == 0;
1277				} else if (iarg->guid != 0) {
1278					uint64_t this_guid;
1279
1280					matched = nvlist_lookup_uint64(config,
1281					    ZPOOL_CONFIG_POOL_GUID,
1282					    &this_guid) == 0 &&
1283					    iarg->guid == this_guid;
1284				}
1285				if (matched) {
1286					/*
1287					 * use the non-raw path for the config
1288					 */
1289					(void) strlcpy(end, slice->rn_name,
1290					    pathleft);
1291					(void) add_config(hdl, &pools,
1292					    path, slice->rn_order,
1293					    slice->rn_num_labels, config);
1294				}
1295				nvlist_free(config);
1296			}
1297			free(slice->rn_name);
1298			free(slice);
1299		}
1300		avl_destroy(&slice_cache);
1301
1302		(void) closedir(dirp);
1303
1304		if (config_failed)
1305			goto error;
1306	}
1307
1308	ret = get_configs(hdl, &pools, iarg->can_be_active, iarg->policy);
1309
1310error:
1311	for (pe = pools.pools; pe != NULL; pe = penext) {
1312		penext = pe->pe_next;
1313		for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
1314			venext = ve->ve_next;
1315			for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
1316				cenext = ce->ce_next;
1317				nvlist_free(ce->ce_config);
1318				free(ce);
1319			}
1320			free(ve);
1321		}
1322		free(pe);
1323	}
1324
1325	for (ne = pools.names; ne != NULL; ne = nenext) {
1326		nenext = ne->ne_next;
1327		free(ne->ne_name);
1328		free(ne);
1329	}
1330
1331	return (ret);
1332}
1333
1334/*
1335 * Given a cache file, return the contents as a list of importable pools.
1336 * poolname or guid (but not both) are provided by the caller when trying
1337 * to import a specific pool.
1338 */
1339static nvlist_t *
1340zpool_find_import_cached(libpc_handle_t *hdl, const char *cachefile,
1341    const char *poolname, uint64_t guid)
1342{
1343	char *buf;
1344	int fd;
1345	struct stat64 statbuf;
1346	nvlist_t *raw, *src, *dst;
1347	nvlist_t *pools;
1348	nvpair_t *elem;
1349	char *name;
1350	uint64_t this_guid;
1351	boolean_t active;
1352
1353	verify(poolname == NULL || guid == 0);
1354
1355	if ((fd = open(cachefile, O_RDONLY)) < 0) {
1356		zutil_error_aux(hdl, "%s", strerror(errno));
1357		(void) zutil_error(hdl, EZFS_BADCACHE,
1358		    dgettext(TEXT_DOMAIN, "failed to open cache file"));
1359		return (NULL);
1360	}
1361
1362	if (fstat64(fd, &statbuf) != 0) {
1363		zutil_error_aux(hdl, "%s", strerror(errno));
1364		(void) close(fd);
1365		(void) zutil_error(hdl, EZFS_BADCACHE,
1366		    dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
1367		return (NULL);
1368	}
1369
1370	if ((buf = zutil_alloc(hdl, statbuf.st_size)) == NULL) {
1371		(void) close(fd);
1372		return (NULL);
1373	}
1374
1375	if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
1376		(void) close(fd);
1377		free(buf);
1378		(void) zutil_error(hdl, EZFS_BADCACHE,
1379		    dgettext(TEXT_DOMAIN,
1380		    "failed to read cache file contents"));
1381		return (NULL);
1382	}
1383
1384	(void) close(fd);
1385
1386	if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
1387		free(buf);
1388		(void) zutil_error(hdl, EZFS_BADCACHE,
1389		    dgettext(TEXT_DOMAIN,
1390		    "invalid or corrupt cache file contents"));
1391		return (NULL);
1392	}
1393
1394	free(buf);
1395
1396	/*
1397	 * Go through and get the current state of the pools and refresh their
1398	 * state.
1399	 */
1400	if (nvlist_alloc(&pools, 0, 0) != 0) {
1401		(void) zutil_no_memory(hdl);
1402		nvlist_free(raw);
1403		return (NULL);
1404	}
1405
1406	elem = NULL;
1407	while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
1408		src = fnvpair_value_nvlist(elem);
1409
1410		name = fnvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME);
1411		if (poolname != NULL && strcmp(poolname, name) != 0)
1412			continue;
1413
1414		this_guid = fnvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID);
1415		if (guid != 0 && guid != this_guid)
1416			continue;
1417
1418		if (zutil_pool_active(hdl, name, this_guid, &active) != 0) {
1419			nvlist_free(raw);
1420			nvlist_free(pools);
1421			return (NULL);
1422		}
1423
1424		if (active)
1425			continue;
1426
1427		if (nvlist_add_string(src, ZPOOL_CONFIG_CACHEFILE,
1428		    cachefile) != 0) {
1429			(void) zutil_no_memory(hdl);
1430			nvlist_free(raw);
1431			nvlist_free(pools);
1432			return (NULL);
1433		}
1434
1435		if ((dst = zutil_refresh_config(hdl, src)) == NULL) {
1436			nvlist_free(raw);
1437			nvlist_free(pools);
1438			return (NULL);
1439		}
1440
1441		if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
1442			(void) zutil_no_memory(hdl);
1443			nvlist_free(dst);
1444			nvlist_free(raw);
1445			nvlist_free(pools);
1446			return (NULL);
1447		}
1448		nvlist_free(dst);
1449	}
1450
1451	nvlist_free(raw);
1452	return (pools);
1453}
1454
1455nvlist_t *
1456zpool_search_import(void *hdl, importargs_t *import,
1457    const pool_config_ops_t *pco)
1458{
1459	libpc_handle_t handle = { 0 };
1460	nvlist_t *pools = NULL;
1461
1462	handle.lpc_lib_handle = hdl;
1463	handle.lpc_ops = pco;
1464	handle.lpc_printerr = B_TRUE;
1465
1466	verify(import->poolname == NULL || import->guid == 0);
1467
1468	if (import->cachefile != NULL)
1469		pools = zpool_find_import_cached(&handle, import->cachefile,
1470		    import->poolname, import->guid);
1471	else
1472		pools = zpool_find_import_impl(&handle, import);
1473
1474	if ((pools == NULL || nvlist_empty(pools)) &&
1475	    handle.lpc_open_access_error && geteuid() != 0) {
1476		(void) zutil_error(&handle, EZFS_EACESS, dgettext(TEXT_DOMAIN,
1477		    "no pools found"));
1478	}
1479
1480	return (pools);
1481}
1482
1483static boolean_t
1484pool_match(nvlist_t *cfg, char *tgt)
1485{
1486	uint64_t v, guid = strtoull(tgt, NULL, 0);
1487	char *s;
1488
1489	if (guid != 0) {
1490		if (nvlist_lookup_uint64(cfg, ZPOOL_CONFIG_POOL_GUID, &v) == 0)
1491			return (v == guid);
1492	} else {
1493		if (nvlist_lookup_string(cfg, ZPOOL_CONFIG_POOL_NAME, &s) == 0)
1494			return (strcmp(s, tgt) == 0);
1495	}
1496	return (B_FALSE);
1497}
1498
1499int
1500zpool_find_config(void *hdl, const char *target, nvlist_t **configp,
1501    importargs_t *args, const pool_config_ops_t *pco)
1502{
1503	nvlist_t *pools;
1504	nvlist_t *match = NULL;
1505	nvlist_t *config = NULL;
1506	char *sepp = NULL;
1507	int count = 0;
1508	char *targetdup = strdup(target);
1509
1510	*configp = NULL;
1511
1512	if ((sepp = strpbrk(targetdup, "/@")) != NULL) {
1513		*sepp = '\0';
1514	}
1515
1516	pools = zpool_search_import(hdl, args, pco);
1517
1518	if (pools != NULL) {
1519		nvpair_t *elem = NULL;
1520		while ((elem = nvlist_next_nvpair(pools, elem)) != NULL) {
1521			VERIFY0(nvpair_value_nvlist(elem, &config));
1522			if (pool_match(config, targetdup)) {
1523				count++;
1524				if (match != NULL) {
1525					/* multiple matches found */
1526					continue;
1527				} else {
1528					match = config;
1529				}
1530			}
1531		}
1532	}
1533
1534	if (count == 0) {
1535		free(targetdup);
1536		return (ENOENT);
1537	}
1538
1539	if (count > 1) {
1540		free(targetdup);
1541		return (EINVAL);
1542	}
1543
1544	*configp = match;
1545	free(targetdup);
1546
1547	return (0);
1548}
1549