libzfs_import.c revision ece3d9b3bacef51a5f34d993935eedbb7bb87059
1/*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21/*
22 * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23 * Use is subject to license terms.
24 */
25
26#pragma ident	"%Z%%M%	%I%	%E% SMI"
27
28/*
29 * Pool import support functions.
30 *
31 * To import a pool, we rely on reading the configuration information from the
32 * ZFS label of each device.  If we successfully read the label, then we
33 * organize the configuration information in the following hierarchy:
34 *
35 * 	pool guid -> toplevel vdev guid -> label txg
36 *
37 * Duplicate entries matching this same tuple will be discarded.  Once we have
38 * examined every device, we pick the best label txg config for each toplevel
39 * vdev.  We then arrange these toplevel vdevs into a complete pool config, and
40 * update any paths that have changed.  Finally, we attempt to import the pool
41 * using our derived config, and record the results.
42 */
43
44#include <devid.h>
45#include <dirent.h>
46#include <errno.h>
47#include <libintl.h>
48#include <stdlib.h>
49#include <string.h>
50#include <sys/stat.h>
51#include <unistd.h>
52#include <fcntl.h>
53
54#include <sys/vdev_impl.h>
55
56#include "libzfs.h"
57#include "libzfs_impl.h"
58
59/*
60 * Intermediate structures used to gather configuration information.
61 */
62typedef struct config_entry {
63	uint64_t		ce_txg;
64	nvlist_t		*ce_config;
65	struct config_entry	*ce_next;
66} config_entry_t;
67
68typedef struct vdev_entry {
69	uint64_t		ve_guid;
70	config_entry_t		*ve_configs;
71	struct vdev_entry	*ve_next;
72} vdev_entry_t;
73
74typedef struct pool_entry {
75	uint64_t		pe_guid;
76	vdev_entry_t		*pe_vdevs;
77	struct pool_entry	*pe_next;
78} pool_entry_t;
79
80typedef struct name_entry {
81	char			*ne_name;
82	uint64_t		ne_guid;
83	struct name_entry	*ne_next;
84} name_entry_t;
85
86typedef struct pool_list {
87	pool_entry_t		*pools;
88	name_entry_t		*names;
89} pool_list_t;
90
91static char *
92get_devid(const char *path)
93{
94	int fd;
95	ddi_devid_t devid;
96	char *minor, *ret;
97
98	if ((fd = open(path, O_RDONLY)) < 0)
99		return (NULL);
100
101	minor = NULL;
102	ret = NULL;
103	if (devid_get(fd, &devid) == 0) {
104		if (devid_get_minor_name(fd, &minor) == 0)
105			ret = devid_str_encode(devid, minor);
106		if (minor != NULL)
107			devid_str_free(minor);
108		devid_free(devid);
109	}
110	(void) close(fd);
111
112	return (ret);
113}
114
115
116/*
117 * Go through and fix up any path and/or devid information for the given vdev
118 * configuration.
119 */
120static int
121fix_paths(nvlist_t *nv, name_entry_t *names)
122{
123	nvlist_t **child;
124	uint_t c, children;
125	uint64_t guid;
126	name_entry_t *ne, *best;
127	char *path, *devid;
128	int matched;
129
130	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
131	    &child, &children) == 0) {
132		for (c = 0; c < children; c++)
133			if (fix_paths(child[c], names) != 0)
134				return (-1);
135		return (0);
136	}
137
138	/*
139	 * This is a leaf (file or disk) vdev.  In either case, go through
140	 * the name list and see if we find a matching guid.  If so, replace
141	 * the path and see if we can calculate a new devid.
142	 *
143	 * There may be multiple names associated with a particular guid, in
144	 * which case we have overlapping slices or multiple paths to the same
145	 * disk.  If this is the case, then we want to pick the path that is
146	 * the most similar to the original, where "most similar" is the number
147	 * of matching characters starting from the end of the path.  This will
148	 * preserve slice numbers even if the disks have been reorganized, and
149	 * will also catch preferred disk names if multiple paths exist.
150	 */
151	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
152	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
153		path = NULL;
154
155	matched = 0;
156	best = NULL;
157	for (ne = names; ne != NULL; ne = ne->ne_next) {
158		if (ne->ne_guid == guid) {
159			const char *src, *dst;
160			int count;
161
162			if (path == NULL) {
163				best = ne;
164				break;
165			}
166
167			src = ne->ne_name + strlen(ne->ne_name) - 1;
168			dst = path + strlen(path) - 1;
169			for (count = 0; src >= ne->ne_name && dst >= path;
170			    src--, dst--, count++)
171				if (*src != *dst)
172					break;
173
174			/*
175			 * At this point, 'count' is the number of characters
176			 * matched from the end.
177			 */
178			if (count > matched || best == NULL) {
179				best = ne;
180				matched = count;
181			}
182		}
183	}
184
185	if (best == NULL)
186		return (0);
187
188	if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
189		return (-1);
190
191	if ((devid = get_devid(best->ne_name)) == NULL) {
192		(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
193	} else {
194		if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
195			return (-1);
196		devid_str_free(devid);
197	}
198
199	return (0);
200}
201
202/*
203 * Add the given configuration to the list of known devices.
204 */
205static int
206add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
207    nvlist_t *config)
208{
209	uint64_t pool_guid, vdev_guid, top_guid, txg, state;
210	pool_entry_t *pe;
211	vdev_entry_t *ve;
212	config_entry_t *ce;
213	name_entry_t *ne;
214
215	/*
216	 * If this is a hot spare not currently in use, add it to the list of
217	 * names to translate, but don't do anything else.
218	 */
219	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
220	    &state) == 0 && state == POOL_STATE_SPARE &&
221	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
222		if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
223		    return (-1);
224
225		if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
226			free(ne);
227			return (-1);
228		}
229		ne->ne_guid = vdev_guid;
230		ne->ne_next = pl->names;
231		pl->names = ne;
232		return (0);
233	}
234
235	/*
236	 * If we have a valid config but cannot read any of these fields, then
237	 * it means we have a half-initialized label.  In vdev_label_init()
238	 * we write a label with txg == 0 so that we can identify the device
239	 * in case the user refers to the same disk later on.  If we fail to
240	 * create the pool, we'll be left with a label in this state
241	 * which should not be considered part of a valid pool.
242	 */
243	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
244	    &pool_guid) != 0 ||
245	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
246	    &vdev_guid) != 0 ||
247	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
248	    &top_guid) != 0 ||
249	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
250	    &txg) != 0 || txg == 0) {
251		nvlist_free(config);
252		return (0);
253	}
254
255	/*
256	 * First, see if we know about this pool.  If not, then add it to the
257	 * list of known pools.
258	 */
259	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
260		if (pe->pe_guid == pool_guid)
261			break;
262	}
263
264	if (pe == NULL) {
265		if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
266			nvlist_free(config);
267			return (-1);
268		}
269		pe->pe_guid = pool_guid;
270		pe->pe_next = pl->pools;
271		pl->pools = pe;
272	}
273
274	/*
275	 * Second, see if we know about this toplevel vdev.  Add it if its
276	 * missing.
277	 */
278	for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
279		if (ve->ve_guid == top_guid)
280			break;
281	}
282
283	if (ve == NULL) {
284		if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
285			nvlist_free(config);
286			return (-1);
287		}
288		ve->ve_guid = top_guid;
289		ve->ve_next = pe->pe_vdevs;
290		pe->pe_vdevs = ve;
291	}
292
293	/*
294	 * Third, see if we have a config with a matching transaction group.  If
295	 * so, then we do nothing.  Otherwise, add it to the list of known
296	 * configs.
297	 */
298	for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
299		if (ce->ce_txg == txg)
300			break;
301	}
302
303	if (ce == NULL) {
304		if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
305			nvlist_free(config);
306			return (-1);
307		}
308		ce->ce_txg = txg;
309		ce->ce_config = config;
310		ce->ce_next = ve->ve_configs;
311		ve->ve_configs = ce;
312	} else {
313		nvlist_free(config);
314	}
315
316	/*
317	 * At this point we've successfully added our config to the list of
318	 * known configs.  The last thing to do is add the vdev guid -> path
319	 * mappings so that we can fix up the configuration as necessary before
320	 * doing the import.
321	 */
322	if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
323		return (-1);
324
325	if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
326		free(ne);
327		return (-1);
328	}
329
330	ne->ne_guid = vdev_guid;
331	ne->ne_next = pl->names;
332	pl->names = ne;
333
334	return (0);
335}
336
337/*
338 * Returns true if the named pool matches the given GUID.
339 */
340static int
341pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
342    boolean_t *isactive)
343{
344	zpool_handle_t *zhp;
345	uint64_t theguid;
346
347	if (zpool_open_silent(hdl, name, &zhp) != 0)
348		return (-1);
349
350	if (zhp == NULL) {
351		*isactive = B_FALSE;
352		return (0);
353	}
354
355	verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
356	    &theguid) == 0);
357
358	zpool_close(zhp);
359
360	*isactive = (theguid == guid);
361	return (0);
362}
363
364/*
365 * Convert our list of pools into the definitive set of configurations.  We
366 * start by picking the best config for each toplevel vdev.  Once that's done,
367 * we assemble the toplevel vdevs into a full config for the pool.  We make a
368 * pass to fix up any incorrect paths, and then add it to the main list to
369 * return to the user.
370 */
371static nvlist_t *
372get_configs(libzfs_handle_t *hdl, pool_list_t *pl)
373{
374	pool_entry_t *pe;
375	vdev_entry_t *ve;
376	config_entry_t *ce;
377	nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
378	nvlist_t **spares;
379	uint_t i, nspares;
380	boolean_t config_seen;
381	uint64_t best_txg;
382	char *name;
383	zfs_cmd_t zc = { 0 };
384	uint64_t version, guid;
385	size_t len;
386	int err;
387	uint_t children = 0;
388	nvlist_t **child = NULL;
389	uint_t c;
390	boolean_t isactive;
391
392	if (nvlist_alloc(&ret, 0, 0) != 0)
393		goto nomem;
394
395	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
396		uint64_t id;
397
398		if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
399			goto nomem;
400		config_seen = B_FALSE;
401
402		/*
403		 * Iterate over all toplevel vdevs.  Grab the pool configuration
404		 * from the first one we find, and then go through the rest and
405		 * add them as necessary to the 'vdevs' member of the config.
406		 */
407		for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
408
409			/*
410			 * Determine the best configuration for this vdev by
411			 * selecting the config with the latest transaction
412			 * group.
413			 */
414			best_txg = 0;
415			for (ce = ve->ve_configs; ce != NULL;
416			    ce = ce->ce_next) {
417
418				if (ce->ce_txg > best_txg) {
419					tmp = ce->ce_config;
420					best_txg = ce->ce_txg;
421				}
422			}
423
424			if (!config_seen) {
425				/*
426				 * Copy the relevant pieces of data to the pool
427				 * configuration:
428				 *
429				 *	version
430				 * 	pool guid
431				 * 	name
432				 * 	pool state
433				 */
434				uint64_t state;
435
436				verify(nvlist_lookup_uint64(tmp,
437				    ZPOOL_CONFIG_VERSION, &version) == 0);
438				if (nvlist_add_uint64(config,
439				    ZPOOL_CONFIG_VERSION, version) != 0)
440					goto nomem;
441				verify(nvlist_lookup_uint64(tmp,
442				    ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
443				if (nvlist_add_uint64(config,
444				    ZPOOL_CONFIG_POOL_GUID, guid) != 0)
445					goto nomem;
446				verify(nvlist_lookup_string(tmp,
447				    ZPOOL_CONFIG_POOL_NAME, &name) == 0);
448				if (nvlist_add_string(config,
449				    ZPOOL_CONFIG_POOL_NAME, name) != 0)
450					goto nomem;
451				verify(nvlist_lookup_uint64(tmp,
452				    ZPOOL_CONFIG_POOL_STATE, &state) == 0);
453				if (nvlist_add_uint64(config,
454				    ZPOOL_CONFIG_POOL_STATE, state) != 0)
455					goto nomem;
456
457				config_seen = B_TRUE;
458			}
459
460			/*
461			 * Add this top-level vdev to the child array.
462			 */
463			verify(nvlist_lookup_nvlist(tmp,
464			    ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
465			verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
466			    &id) == 0);
467			if (id >= children) {
468				nvlist_t **newchild;
469
470				newchild = zfs_alloc(hdl, (id + 1) *
471				    sizeof (nvlist_t *));
472				if (newchild == NULL)
473					goto nomem;
474
475				for (c = 0; c < children; c++)
476					newchild[c] = child[c];
477
478				free(child);
479				child = newchild;
480				children = id + 1;
481			}
482			if (nvlist_dup(nvtop, &child[id], 0) != 0)
483				goto nomem;
484
485		}
486
487		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
488		    &guid) == 0);
489
490		/*
491		 * Look for any missing top-level vdevs.  If this is the case,
492		 * create a faked up 'missing' vdev as a placeholder.  We cannot
493		 * simply compress the child array, because the kernel performs
494		 * certain checks to make sure the vdev IDs match their location
495		 * in the configuration.
496		 */
497		for (c = 0; c < children; c++)
498			if (child[c] == NULL) {
499				nvlist_t *missing;
500				if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
501				    0) != 0)
502					goto nomem;
503				if (nvlist_add_string(missing,
504				    ZPOOL_CONFIG_TYPE,
505				    VDEV_TYPE_MISSING) != 0 ||
506				    nvlist_add_uint64(missing,
507				    ZPOOL_CONFIG_ID, c) != 0 ||
508				    nvlist_add_uint64(missing,
509				    ZPOOL_CONFIG_GUID, 0ULL) != 0) {
510					nvlist_free(missing);
511					goto nomem;
512				}
513				child[c] = missing;
514			}
515
516		/*
517		 * Put all of this pool's top-level vdevs into a root vdev.
518		 */
519		if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
520			goto nomem;
521		if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
522		    VDEV_TYPE_ROOT) != 0 ||
523		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
524		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
525		    nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
526		    child, children) != 0) {
527			nvlist_free(nvroot);
528			goto nomem;
529		}
530
531		for (c = 0; c < children; c++)
532			nvlist_free(child[c]);
533		free(child);
534		children = 0;
535		child = NULL;
536
537		/*
538		 * Go through and fix up any paths and/or devids based on our
539		 * known list of vdev GUID -> path mappings.
540		 */
541		if (fix_paths(nvroot, pl->names) != 0) {
542			nvlist_free(nvroot);
543			goto nomem;
544		}
545
546		/*
547		 * Add the root vdev to this pool's configuration.
548		 */
549		if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
550		    nvroot) != 0) {
551			nvlist_free(nvroot);
552			goto nomem;
553		}
554		nvlist_free(nvroot);
555
556		/*
557		 * Determine if this pool is currently active, in which case we
558		 * can't actually import it.
559		 */
560		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
561		    &name) == 0);
562		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
563		    &guid) == 0);
564
565		if (pool_active(hdl, name, guid, &isactive) != 0)
566			goto error;
567
568		if (isactive) {
569			nvlist_free(config);
570			config = NULL;
571			continue;
572		}
573
574		/*
575		 * Try to do the import in order to get vdev state.
576		 */
577		if (zcmd_write_src_nvlist(hdl, &zc, config, &len) != 0)
578			goto error;
579
580		nvlist_free(config);
581		config = NULL;
582
583		if (zcmd_alloc_dst_nvlist(hdl, &zc, len * 2) != 0) {
584			zcmd_free_nvlists(&zc);
585			goto error;
586		}
587
588		while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
589		    &zc)) != 0 && errno == ENOMEM) {
590			if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
591				zcmd_free_nvlists(&zc);
592				goto error;
593			}
594		}
595
596		if (err) {
597			(void) zpool_standard_error(hdl, errno,
598			    dgettext(TEXT_DOMAIN, "cannot discover pools"));
599			zcmd_free_nvlists(&zc);
600			goto error;
601		}
602
603		if (zcmd_read_dst_nvlist(hdl, &zc, &config) != 0) {
604			zcmd_free_nvlists(&zc);
605			goto error;
606		}
607
608		zcmd_free_nvlists(&zc);
609
610		/*
611		 * Go through and update the paths for spares, now that we have
612		 * them.
613		 */
614		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
615		    &nvroot) == 0);
616		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
617		    &spares, &nspares) == 0) {
618			for (i = 0; i < nspares; i++) {
619				if (fix_paths(spares[i], pl->names) != 0)
620					goto nomem;
621			}
622		}
623
624		if (set_pool_health(config) != 0)
625			goto nomem;
626
627		/*
628		 * Add this pool to the list of configs.
629		 */
630		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
631		    &name) == 0);
632		if (nvlist_add_nvlist(ret, name, config) != 0)
633			goto nomem;
634
635		nvlist_free(config);
636		config = NULL;
637	}
638
639	return (ret);
640
641nomem:
642	(void) no_memory(hdl);
643error:
644	nvlist_free(config);
645	nvlist_free(ret);
646	for (c = 0; c < children; c++)
647		nvlist_free(child[c]);
648	free(child);
649
650	return (NULL);
651}
652
653/*
654 * Return the offset of the given label.
655 */
656static uint64_t
657label_offset(size_t size, int l)
658{
659	return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
660	    0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
661}
662
663/*
664 * Given a file descriptor, read the label information and return an nvlist
665 * describing the configuration, if there is one.
666 */
667int
668zpool_read_label(int fd, nvlist_t **config)
669{
670	struct stat64 statbuf;
671	int l;
672	vdev_label_t *label;
673	uint64_t state, txg;
674
675	*config = NULL;
676
677	if (fstat64(fd, &statbuf) == -1)
678		return (0);
679
680	if ((label = malloc(sizeof (vdev_label_t))) == NULL)
681		return (-1);
682
683	for (l = 0; l < VDEV_LABELS; l++) {
684		if (pread(fd, label, sizeof (vdev_label_t),
685		    label_offset(statbuf.st_size, l)) != sizeof (vdev_label_t))
686			continue;
687
688		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
689		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
690			continue;
691
692		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
693		    &state) != 0 || state > POOL_STATE_SPARE) {
694			nvlist_free(*config);
695			continue;
696		}
697
698		if (state != POOL_STATE_SPARE &&
699		    (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
700		    &txg) != 0 || txg == 0)) {
701			nvlist_free(*config);
702			continue;
703		}
704
705		free(label);
706		return (0);
707	}
708
709	free(label);
710	*config = NULL;
711	return (0);
712}
713
714/*
715 * Given a list of directories to search, find all pools stored on disk.  This
716 * includes partial pools which are not available to import.  If no args are
717 * given (argc is 0), then the default directory (/dev/dsk) is searched.
718 */
719nvlist_t *
720zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
721{
722	int i;
723	DIR *dirp;
724	struct dirent64 *dp;
725	char path[MAXPATHLEN];
726	struct stat64 statbuf;
727	nvlist_t *ret = NULL, *config;
728	static char *default_dir = "/dev/dsk";
729	int fd;
730	pool_list_t pools = { 0 };
731	pool_entry_t *pe, *penext;
732	vdev_entry_t *ve, *venext;
733	config_entry_t *ce, *cenext;
734	name_entry_t *ne, *nenext;
735
736
737	if (argc == 0) {
738		argc = 1;
739		argv = &default_dir;
740	}
741
742	/*
743	 * Go through and read the label configuration information from every
744	 * possible device, organizing the information according to pool GUID
745	 * and toplevel GUID.
746	 */
747	for (i = 0; i < argc; i++) {
748		if (argv[i][0] != '/') {
749			(void) zfs_error_fmt(hdl, EZFS_BADPATH,
750			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
751			    argv[i]);
752			goto error;
753		}
754
755		if ((dirp = opendir(argv[i])) == NULL) {
756			zfs_error_aux(hdl, strerror(errno));
757			(void) zfs_error_fmt(hdl, EZFS_BADPATH,
758			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
759			    argv[i]);
760			goto error;
761		}
762
763		/*
764		 * This is not MT-safe, but we have no MT consumers of libzfs
765		 */
766		while ((dp = readdir64(dirp)) != NULL) {
767
768			(void) snprintf(path, sizeof (path), "%s/%s",
769			    argv[i], dp->d_name);
770
771			if (stat64(path, &statbuf) != 0)
772				continue;
773
774			/*
775			 * Ignore directories (which includes "." and "..").
776			 */
777			if (S_ISDIR(statbuf.st_mode))
778				continue;
779
780			/*
781			 * Ignore special (non-character or non-block) files.
782			 */
783			if (!S_ISREG(statbuf.st_mode) &&
784			    !S_ISBLK(statbuf.st_mode))
785				continue;
786
787			if ((fd = open64(path, O_RDONLY)) < 0)
788				continue;
789
790			if ((zpool_read_label(fd, &config)) != 0) {
791				(void) no_memory(hdl);
792				goto error;
793			}
794
795			(void) close(fd);
796
797			if (config != NULL)
798				if (add_config(hdl, &pools, path, config) != 0)
799					goto error;
800		}
801	}
802
803	ret = get_configs(hdl, &pools);
804
805error:
806	for (pe = pools.pools; pe != NULL; pe = penext) {
807		penext = pe->pe_next;
808		for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
809			venext = ve->ve_next;
810			for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
811				cenext = ce->ce_next;
812				if (ce->ce_config)
813					nvlist_free(ce->ce_config);
814				free(ce);
815			}
816			free(ve);
817		}
818		free(pe);
819	}
820
821	for (ne = pools.names; ne != NULL; ne = nenext) {
822		nenext = ne->ne_next;
823		if (ne->ne_name)
824			free(ne->ne_name);
825		free(ne);
826	}
827
828
829	return (ret);
830}
831
832boolean_t
833find_guid(nvlist_t *nv, uint64_t guid)
834{
835	uint64_t tmp;
836	nvlist_t **child;
837	uint_t c, children;
838
839	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
840	if (tmp == guid)
841		return (B_TRUE);
842
843	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
844	    &child, &children) == 0) {
845		for (c = 0; c < children; c++)
846			if (find_guid(child[c], guid))
847				return (B_TRUE);
848	}
849
850	return (B_FALSE);
851}
852
853typedef struct spare_cbdata {
854	uint64_t	cb_guid;
855	zpool_handle_t	*cb_zhp;
856} spare_cbdata_t;
857
858static int
859find_spare(zpool_handle_t *zhp, void *data)
860{
861	spare_cbdata_t *cbp = data;
862	nvlist_t **spares;
863	uint_t i, nspares;
864	uint64_t guid;
865	nvlist_t *nvroot;
866
867	verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
868	    &nvroot) == 0);
869
870	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
871	    &spares, &nspares) == 0) {
872		for (i = 0; i < nspares; i++) {
873			verify(nvlist_lookup_uint64(spares[i],
874			    ZPOOL_CONFIG_GUID, &guid) == 0);
875			if (guid == cbp->cb_guid) {
876				cbp->cb_zhp = zhp;
877				return (1);
878			}
879		}
880	}
881
882	zpool_close(zhp);
883	return (0);
884}
885
886/*
887 * Determines if the pool is in use.  If so, it returns true and the state of
888 * the pool as well as the name of the pool.  Both strings are allocated and
889 * must be freed by the caller.
890 */
891int
892zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
893    boolean_t *inuse)
894{
895	nvlist_t *config;
896	char *name;
897	boolean_t ret;
898	uint64_t guid, vdev_guid;
899	zpool_handle_t *zhp;
900	nvlist_t *pool_config;
901	uint64_t stateval;
902	spare_cbdata_t cb = { 0 };
903	boolean_t isactive;
904
905	*inuse = B_FALSE;
906
907	if (zpool_read_label(fd, &config) != 0) {
908		(void) no_memory(hdl);
909		return (-1);
910	}
911
912	if (config == NULL)
913		return (0);
914
915	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
916	    &stateval) == 0);
917	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
918	    &vdev_guid) == 0);
919
920	if (stateval != POOL_STATE_SPARE) {
921		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
922		    &name) == 0);
923		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
924		    &guid) == 0);
925	}
926
927	switch (stateval) {
928	case POOL_STATE_EXPORTED:
929		ret = B_TRUE;
930		break;
931
932	case POOL_STATE_ACTIVE:
933		/*
934		 * For an active pool, we have to determine if it's really part
935		 * of a currently active pool (in which case the pool will exist
936		 * and the guid will be the same), or whether it's part of an
937		 * active pool that was disconnected without being explicitly
938		 * exported.
939		 */
940		if (pool_active(hdl, name, guid, &isactive) != 0) {
941			nvlist_free(config);
942			return (-1);
943		}
944
945		if (isactive) {
946			/*
947			 * Because the device may have been removed while
948			 * offlined, we only report it as active if the vdev is
949			 * still present in the config.  Otherwise, pretend like
950			 * it's not in use.
951			 */
952			if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
953			    (pool_config = zpool_get_config(zhp, NULL))
954			    != NULL) {
955				nvlist_t *nvroot;
956
957				verify(nvlist_lookup_nvlist(pool_config,
958				    ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
959				ret = find_guid(nvroot, vdev_guid);
960			} else {
961				ret = B_FALSE;
962			}
963
964			if (zhp != NULL)
965				zpool_close(zhp);
966		} else {
967			stateval = POOL_STATE_POTENTIALLY_ACTIVE;
968			ret = B_TRUE;
969		}
970		break;
971
972	case POOL_STATE_SPARE:
973		/*
974		 * For a hot spare, it can be either definitively in use, or
975		 * potentially active.  To determine if it's in use, we iterate
976		 * over all pools in the system and search for one with a spare
977		 * with a matching guid.
978		 *
979		 * Due to the shared nature of spares, we don't actually report
980		 * the potentially active case as in use.  This means the user
981		 * can freely create pools on the hot spares of exported pools,
982		 * but to do otherwise makes the resulting code complicated, and
983		 * we end up having to deal with this case anyway.
984		 */
985		cb.cb_zhp = NULL;
986		cb.cb_guid = vdev_guid;
987		if (zpool_iter(hdl, find_spare, &cb) == 1) {
988			name = (char *)zpool_get_name(cb.cb_zhp);
989			ret = TRUE;
990		} else {
991			ret = FALSE;
992		}
993		break;
994
995	default:
996		ret = B_FALSE;
997	}
998
999
1000	if (ret) {
1001		if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1002			nvlist_free(config);
1003			return (-1);
1004		}
1005		*state = (pool_state_t)stateval;
1006	}
1007
1008	if (cb.cb_zhp)
1009		zpool_close(cb.cb_zhp);
1010
1011	nvlist_free(config);
1012	*inuse = ret;
1013	return (0);
1014}
1015