libzfs_import.c revision 3bb79bece53191f2cf27aa61a72ea1784a7ce700
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	char *packed;
386	size_t len;
387	int err;
388	uint_t children = 0;
389	nvlist_t **child = NULL;
390	uint_t c;
391	boolean_t isactive;
392
393	if (nvlist_alloc(&ret, 0, 0) != 0)
394		goto nomem;
395
396	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
397		uint64_t id;
398
399		if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
400			goto nomem;
401		config_seen = B_FALSE;
402
403		/*
404		 * Iterate over all toplevel vdevs.  Grab the pool configuration
405		 * from the first one we find, and then go through the rest and
406		 * add them as necessary to the 'vdevs' member of the config.
407		 */
408		for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
409
410			/*
411			 * Determine the best configuration for this vdev by
412			 * selecting the config with the latest transaction
413			 * group.
414			 */
415			best_txg = 0;
416			for (ce = ve->ve_configs; ce != NULL;
417			    ce = ce->ce_next) {
418
419				if (ce->ce_txg > best_txg) {
420					tmp = ce->ce_config;
421					best_txg = ce->ce_txg;
422				}
423			}
424
425			if (!config_seen) {
426				/*
427				 * Copy the relevant pieces of data to the pool
428				 * configuration:
429				 *
430				 *	version
431				 * 	pool guid
432				 * 	name
433				 * 	pool state
434				 */
435				uint64_t state;
436
437				verify(nvlist_lookup_uint64(tmp,
438				    ZPOOL_CONFIG_VERSION, &version) == 0);
439				if (nvlist_add_uint64(config,
440				    ZPOOL_CONFIG_VERSION, version) != 0)
441					goto nomem;
442				verify(nvlist_lookup_uint64(tmp,
443				    ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
444				if (nvlist_add_uint64(config,
445				    ZPOOL_CONFIG_POOL_GUID, guid) != 0)
446					goto nomem;
447				verify(nvlist_lookup_string(tmp,
448				    ZPOOL_CONFIG_POOL_NAME, &name) == 0);
449				if (nvlist_add_string(config,
450				    ZPOOL_CONFIG_POOL_NAME, name) != 0)
451					goto nomem;
452				verify(nvlist_lookup_uint64(tmp,
453				    ZPOOL_CONFIG_POOL_STATE, &state) == 0);
454				if (nvlist_add_uint64(config,
455				    ZPOOL_CONFIG_POOL_STATE, state) != 0)
456					goto nomem;
457
458				config_seen = B_TRUE;
459			}
460
461			/*
462			 * Add this top-level vdev to the child array.
463			 */
464			verify(nvlist_lookup_nvlist(tmp,
465			    ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
466			verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
467			    &id) == 0);
468			if (id >= children) {
469				nvlist_t **newchild;
470
471				newchild = zfs_alloc(hdl, (id + 1) *
472				    sizeof (nvlist_t *));
473				if (newchild == NULL)
474					goto nomem;
475
476				for (c = 0; c < children; c++)
477					newchild[c] = child[c];
478
479				free(child);
480				child = newchild;
481				children = id + 1;
482			}
483			if (nvlist_dup(nvtop, &child[id], 0) != 0)
484				goto nomem;
485
486		}
487
488		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
489		    &guid) == 0);
490
491		/*
492		 * Look for any missing top-level vdevs.  If this is the case,
493		 * create a faked up 'missing' vdev as a placeholder.  We cannot
494		 * simply compress the child array, because the kernel performs
495		 * certain checks to make sure the vdev IDs match their location
496		 * in the configuration.
497		 */
498		for (c = 0; c < children; c++)
499			if (child[c] == NULL) {
500				nvlist_t *missing;
501				if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
502				    0) != 0)
503					goto nomem;
504				if (nvlist_add_string(missing,
505				    ZPOOL_CONFIG_TYPE,
506				    VDEV_TYPE_MISSING) != 0 ||
507				    nvlist_add_uint64(missing,
508				    ZPOOL_CONFIG_ID, c) != 0 ||
509				    nvlist_add_uint64(missing,
510				    ZPOOL_CONFIG_GUID, 0ULL) != 0) {
511					nvlist_free(missing);
512					goto nomem;
513				}
514				child[c] = missing;
515			}
516
517		/*
518		 * Put all of this pool's top-level vdevs into a root vdev.
519		 */
520		if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
521			goto nomem;
522		if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
523		    VDEV_TYPE_ROOT) != 0 ||
524		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
525		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
526		    nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
527		    child, children) != 0) {
528			nvlist_free(nvroot);
529			goto nomem;
530		}
531
532		for (c = 0; c < children; c++)
533			nvlist_free(child[c]);
534		free(child);
535		children = 0;
536		child = NULL;
537
538		/*
539		 * Go through and fix up any paths and/or devids based on our
540		 * known list of vdev GUID -> path mappings.
541		 */
542		if (fix_paths(nvroot, pl->names) != 0) {
543			nvlist_free(nvroot);
544			goto nomem;
545		}
546
547		/*
548		 * Add the root vdev to this pool's configuration.
549		 */
550		if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
551		    nvroot) != 0) {
552			nvlist_free(nvroot);
553			goto nomem;
554		}
555		nvlist_free(nvroot);
556
557		/*
558		 * Determine if this pool is currently active, in which case we
559		 * can't actually import it.
560		 */
561		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
562		    &name) == 0);
563		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
564		    &guid) == 0);
565
566		if (pool_active(hdl, name, guid, &isactive) != 0)
567			goto error;
568
569		if (isactive) {
570			nvlist_free(config);
571			config = NULL;
572			continue;
573		}
574
575		/*
576		 * Try to do the import in order to get vdev state.
577		 */
578		if ((err = nvlist_size(config, &len, NV_ENCODE_NATIVE)) != 0)
579			goto nomem;
580
581		if ((packed = zfs_alloc(hdl, len)) == NULL)
582			goto nomem;
583
584		if ((err = nvlist_pack(config, &packed, &len,
585		    NV_ENCODE_NATIVE, 0)) != 0)
586			goto nomem;
587
588		nvlist_free(config);
589		config = NULL;
590
591		zc.zc_config_src_size = len;
592		zc.zc_config_src = (uint64_t)(uintptr_t)packed;
593
594		zc.zc_config_dst_size = 2 * len;
595		if ((zc.zc_config_dst = (uint64_t)(uintptr_t)
596		    zfs_alloc(hdl, zc.zc_config_dst_size)) == NULL)
597			goto nomem;
598
599		while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
600		    &zc)) != 0 && errno == ENOMEM) {
601			free((void *)(uintptr_t)zc.zc_config_dst);
602			if ((zc.zc_config_dst = (uint64_t)(uintptr_t)
603			    zfs_alloc(hdl, zc.zc_config_dst_size)) == NULL)
604				goto nomem;
605		}
606
607		free(packed);
608
609		if (err) {
610			(void) zpool_standard_error(hdl, errno,
611			    dgettext(TEXT_DOMAIN, "cannot discover pools"));
612			free((void *)(uintptr_t)zc.zc_config_dst);
613			goto error;
614		}
615
616		if (nvlist_unpack((void *)(uintptr_t)zc.zc_config_dst,
617		    zc.zc_config_dst_size, &config, 0) != 0) {
618			free((void *)(uintptr_t)zc.zc_config_dst);
619			goto nomem;
620		}
621		free((void *)(uintptr_t)zc.zc_config_dst);
622
623		/*
624		 * Go through and update the paths for spares, now that we have
625		 * them.
626		 */
627		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
628		    &nvroot) == 0);
629		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
630		    &spares, &nspares) == 0) {
631			for (i = 0; i < nspares; i++) {
632				if (fix_paths(spares[i], pl->names) != 0)
633					goto nomem;
634			}
635		}
636
637		if (set_pool_health(config) != 0)
638			goto nomem;
639
640		/*
641		 * Add this pool to the list of configs.
642		 */
643		if (nvlist_add_nvlist(ret, name, config) != 0)
644			goto nomem;
645
646		nvlist_free(config);
647		config = NULL;
648	}
649
650	return (ret);
651
652nomem:
653	(void) no_memory(hdl);
654error:
655	nvlist_free(config);
656	nvlist_free(ret);
657	for (c = 0; c < children; c++)
658		nvlist_free(child[c]);
659	free(child);
660
661	return (NULL);
662}
663
664/*
665 * Return the offset of the given label.
666 */
667static uint64_t
668label_offset(size_t size, int l)
669{
670	return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
671	    0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
672}
673
674/*
675 * Given a file descriptor, read the label information and return an nvlist
676 * describing the configuration, if there is one.
677 */
678int
679zpool_read_label(int fd, nvlist_t **config)
680{
681	struct stat64 statbuf;
682	int l;
683	vdev_label_t *label;
684	uint64_t state, txg;
685
686	*config = NULL;
687
688	if (fstat64(fd, &statbuf) == -1)
689		return (0);
690
691	if ((label = malloc(sizeof (vdev_label_t))) == NULL)
692		return (-1);
693
694	for (l = 0; l < VDEV_LABELS; l++) {
695		if (pread(fd, label, sizeof (vdev_label_t),
696		    label_offset(statbuf.st_size, l)) != sizeof (vdev_label_t))
697			continue;
698
699		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
700		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
701			continue;
702
703		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
704		    &state) != 0 || state > POOL_STATE_SPARE) {
705			nvlist_free(*config);
706			continue;
707		}
708
709		if (state != POOL_STATE_SPARE &&
710		    (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
711		    &txg) != 0 || txg == 0)) {
712			nvlist_free(*config);
713			continue;
714		}
715
716		free(label);
717		return (0);
718	}
719
720	free(label);
721	*config = NULL;
722	return (0);
723}
724
725/*
726 * Given a list of directories to search, find all pools stored on disk.  This
727 * includes partial pools which are not available to import.  If no args are
728 * given (argc is 0), then the default directory (/dev/dsk) is searched.
729 */
730nvlist_t *
731zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
732{
733	int i;
734	DIR *dirp;
735	struct dirent64 *dp;
736	char path[MAXPATHLEN];
737	struct stat64 statbuf;
738	nvlist_t *ret = NULL, *config;
739	static char *default_dir = "/dev/dsk";
740	int fd;
741	pool_list_t pools = { 0 };
742	pool_entry_t *pe, *penext;
743	vdev_entry_t *ve, *venext;
744	config_entry_t *ce, *cenext;
745	name_entry_t *ne, *nenext;
746
747
748	if (argc == 0) {
749		argc = 1;
750		argv = &default_dir;
751	}
752
753	/*
754	 * Go through and read the label configuration information from every
755	 * possible device, organizing the information according to pool GUID
756	 * and toplevel GUID.
757	 */
758	for (i = 0; i < argc; i++) {
759		if (argv[i][0] != '/') {
760			(void) zfs_error(hdl, EZFS_BADPATH,
761			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
762			    argv[i]);
763			goto error;
764		}
765
766		if ((dirp = opendir(argv[i])) == NULL) {
767			zfs_error_aux(hdl, strerror(errno));
768			(void) zfs_error(hdl, EZFS_BADPATH,
769			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
770			    argv[i]);
771			goto error;
772		}
773
774		/*
775		 * This is not MT-safe, but we have no MT consumers of libzfs
776		 */
777		while ((dp = readdir64(dirp)) != NULL) {
778
779			(void) snprintf(path, sizeof (path), "%s/%s",
780			    argv[i], dp->d_name);
781
782			if (stat64(path, &statbuf) != 0)
783				continue;
784
785			/*
786			 * Ignore directories (which includes "." and "..").
787			 */
788			if (S_ISDIR(statbuf.st_mode))
789				continue;
790
791			/*
792			 * Ignore special (non-character or non-block) files.
793			 */
794			if (!S_ISREG(statbuf.st_mode) &&
795			    !S_ISBLK(statbuf.st_mode))
796				continue;
797
798			if ((fd = open64(path, O_RDONLY)) < 0)
799				continue;
800
801			if ((zpool_read_label(fd, &config)) != 0) {
802				(void) no_memory(hdl);
803				goto error;
804			}
805
806			(void) close(fd);
807
808			if (config != NULL)
809				if (add_config(hdl, &pools, path, config) != 0)
810					goto error;
811		}
812	}
813
814	ret = get_configs(hdl, &pools);
815
816error:
817	for (pe = pools.pools; pe != NULL; pe = penext) {
818		penext = pe->pe_next;
819		for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
820			venext = ve->ve_next;
821			for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
822				cenext = ce->ce_next;
823				if (ce->ce_config)
824					nvlist_free(ce->ce_config);
825				free(ce);
826			}
827			free(ve);
828		}
829		free(pe);
830	}
831
832	for (ne = pools.names; ne != NULL; ne = nenext) {
833		nenext = ne->ne_next;
834		if (ne->ne_name)
835			free(ne->ne_name);
836		free(ne);
837	}
838
839
840	return (ret);
841}
842
843boolean_t
844find_guid(nvlist_t *nv, uint64_t guid)
845{
846	uint64_t tmp;
847	nvlist_t **child;
848	uint_t c, children;
849
850	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
851	if (tmp == guid)
852		return (B_TRUE);
853
854	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
855	    &child, &children) == 0) {
856		for (c = 0; c < children; c++)
857			if (find_guid(child[c], guid))
858				return (B_TRUE);
859	}
860
861	return (B_FALSE);
862}
863
864typedef struct spare_cbdata {
865	uint64_t	cb_guid;
866	zpool_handle_t	*cb_zhp;
867} spare_cbdata_t;
868
869static int
870find_spare(zpool_handle_t *zhp, void *data)
871{
872	spare_cbdata_t *cbp = data;
873	nvlist_t **spares;
874	uint_t i, nspares;
875	uint64_t guid;
876	nvlist_t *nvroot;
877
878	verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
879	    &nvroot) == 0);
880
881	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
882	    &spares, &nspares) == 0) {
883		for (i = 0; i < nspares; i++) {
884			verify(nvlist_lookup_uint64(spares[i],
885			    ZPOOL_CONFIG_GUID, &guid) == 0);
886			if (guid == cbp->cb_guid) {
887				cbp->cb_zhp = zhp;
888				return (1);
889			}
890		}
891	}
892
893	zpool_close(zhp);
894	return (0);
895}
896
897/*
898 * Determines if the pool is in use.  If so, it returns true and the state of
899 * the pool as well as the name of the pool.  Both strings are allocated and
900 * must be freed by the caller.
901 */
902int
903zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
904    boolean_t *inuse)
905{
906	nvlist_t *config;
907	char *name;
908	boolean_t ret;
909	uint64_t guid, vdev_guid;
910	zpool_handle_t *zhp;
911	nvlist_t *pool_config;
912	uint64_t stateval;
913	spare_cbdata_t cb = { 0 };
914	boolean_t isactive;
915
916	*inuse = B_FALSE;
917
918	if (zpool_read_label(fd, &config) != 0) {
919		(void) no_memory(hdl);
920		return (-1);
921	}
922
923	if (config == NULL)
924		return (0);
925
926	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
927	    &stateval) == 0);
928	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
929	    &vdev_guid) == 0);
930
931	if (stateval != POOL_STATE_SPARE) {
932		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
933		    &name) == 0);
934		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
935		    &guid) == 0);
936	}
937
938	switch (stateval) {
939	case POOL_STATE_EXPORTED:
940		ret = B_TRUE;
941		break;
942
943	case POOL_STATE_ACTIVE:
944		/*
945		 * For an active pool, we have to determine if it's really part
946		 * of a currently active pool (in which case the pool will exist
947		 * and the guid will be the same), or whether it's part of an
948		 * active pool that was disconnected without being explicitly
949		 * exported.
950		 */
951		if (pool_active(hdl, name, guid, &isactive) != 0) {
952			nvlist_free(config);
953			return (-1);
954		}
955
956		if (isactive) {
957			/*
958			 * Because the device may have been removed while
959			 * offlined, we only report it as active if the vdev is
960			 * still present in the config.  Otherwise, pretend like
961			 * it's not in use.
962			 */
963			if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
964			    (pool_config = zpool_get_config(zhp, NULL))
965			    != NULL) {
966				nvlist_t *nvroot;
967
968				verify(nvlist_lookup_nvlist(pool_config,
969				    ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
970				ret = find_guid(nvroot, vdev_guid);
971			} else {
972				ret = B_FALSE;
973			}
974
975			if (zhp != NULL)
976				zpool_close(zhp);
977		} else {
978			stateval = POOL_STATE_POTENTIALLY_ACTIVE;
979			ret = B_TRUE;
980		}
981		break;
982
983	case POOL_STATE_SPARE:
984		/*
985		 * For a hot spare, it can be either definitively in use, or
986		 * potentially active.  To determine if it's in use, we iterate
987		 * over all pools in the system and search for one with a spare
988		 * with a matching guid.
989		 *
990		 * Due to the shared nature of spares, we don't actually report
991		 * the potentially active case as in use.  This means the user
992		 * can freely create pools on the hot spares of exported pools,
993		 * but to do otherwise makes the resulting code complicated, and
994		 * we end up having to deal with this case anyway.
995		 */
996		cb.cb_zhp = NULL;
997		cb.cb_guid = vdev_guid;
998		if (zpool_iter(hdl, find_spare, &cb) == 1) {
999			name = (char *)zpool_get_name(cb.cb_zhp);
1000			ret = TRUE;
1001		} else {
1002			ret = FALSE;
1003		}
1004		break;
1005
1006	default:
1007		ret = B_FALSE;
1008	}
1009
1010
1011	if (ret) {
1012		if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1013			nvlist_free(config);
1014			return (-1);
1015		}
1016		*state = (pool_state_t)stateval;
1017	}
1018
1019	if (cb.cb_zhp)
1020		zpool_close(cb.cb_zhp);
1021
1022	nvlist_free(config);
1023	*inuse = ret;
1024	return (0);
1025}
1026