1fa9e4066Sahrens /*
2fa9e4066Sahrens  * CDDL HEADER START
3fa9e4066Sahrens  *
4fa9e4066Sahrens  * The contents of this file are subject to the terms of the
5ea8dc4b6Seschrock  * Common Development and Distribution License (the "License").
6ea8dc4b6Seschrock  * You may not use this file except in compliance with the License.
7fa9e4066Sahrens  *
8fa9e4066Sahrens  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9fa9e4066Sahrens  * or http://www.opensolaris.org/os/licensing.
10fa9e4066Sahrens  * See the License for the specific language governing permissions
11fa9e4066Sahrens  * and limitations under the License.
12fa9e4066Sahrens  *
13fa9e4066Sahrens  * When distributing Covered Code, include this CDDL HEADER in each
14fa9e4066Sahrens  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15fa9e4066Sahrens  * If applicable, add the following below this CDDL HEADER, with the
16fa9e4066Sahrens  * fields enclosed by brackets "[]" replaced with your own identifying
17fa9e4066Sahrens  * information: Portions Copyright [yyyy] [name of copyright owner]
18fa9e4066Sahrens  *
19fa9e4066Sahrens  * CDDL HEADER END
20fa9e4066Sahrens  */
21fa9e4066Sahrens /*
2239c23413Seschrock  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23fa9e4066Sahrens  * Use is subject to license terms.
24fa9e4066Sahrens  */
25fa9e4066Sahrens 
26fa9e4066Sahrens #pragma ident	"%Z%%M%	%I%	%E% SMI"
27fa9e4066Sahrens 
28fa9e4066Sahrens /*
29fa9e4066Sahrens  * Pool import support functions.
30fa9e4066Sahrens  *
31fa9e4066Sahrens  * To import a pool, we rely on reading the configuration information from the
32fa9e4066Sahrens  * ZFS label of each device.  If we successfully read the label, then we
33fa9e4066Sahrens  * organize the configuration information in the following hierarchy:
34fa9e4066Sahrens  *
35fa9e4066Sahrens  * 	pool guid -> toplevel vdev guid -> label txg
36fa9e4066Sahrens  *
37fa9e4066Sahrens  * Duplicate entries matching this same tuple will be discarded.  Once we have
38fa9e4066Sahrens  * examined every device, we pick the best label txg config for each toplevel
39fa9e4066Sahrens  * vdev.  We then arrange these toplevel vdevs into a complete pool config, and
40fa9e4066Sahrens  * update any paths that have changed.  Finally, we attempt to import the pool
41fa9e4066Sahrens  * using our derived config, and record the results.
42fa9e4066Sahrens  */
43fa9e4066Sahrens 
44fa9e4066Sahrens #include <devid.h>
45fa9e4066Sahrens #include <dirent.h>
46fa9e4066Sahrens #include <errno.h>
47fa9e4066Sahrens #include <libintl.h>
48fa9e4066Sahrens #include <stdlib.h>
49fa9e4066Sahrens #include <string.h>
50fa9e4066Sahrens #include <sys/stat.h>
51fa9e4066Sahrens #include <unistd.h>
52fa9e4066Sahrens #include <fcntl.h>
53fa9e4066Sahrens 
54fa9e4066Sahrens #include <sys/vdev_impl.h>
55fa9e4066Sahrens 
56fa9e4066Sahrens #include "libzfs.h"
57fa9e4066Sahrens #include "libzfs_impl.h"
58fa9e4066Sahrens 
59fa9e4066Sahrens /*
60fa9e4066Sahrens  * Intermediate structures used to gather configuration information.
61fa9e4066Sahrens  */
62fa9e4066Sahrens typedef struct config_entry {
63fa9e4066Sahrens 	uint64_t		ce_txg;
64fa9e4066Sahrens 	nvlist_t		*ce_config;
65fa9e4066Sahrens 	struct config_entry	*ce_next;
66fa9e4066Sahrens } config_entry_t;
67fa9e4066Sahrens 
68fa9e4066Sahrens typedef struct vdev_entry {
69fa9e4066Sahrens 	uint64_t		ve_guid;
70fa9e4066Sahrens 	config_entry_t		*ve_configs;
71fa9e4066Sahrens 	struct vdev_entry	*ve_next;
72fa9e4066Sahrens } vdev_entry_t;
73fa9e4066Sahrens 
74fa9e4066Sahrens typedef struct pool_entry {
75fa9e4066Sahrens 	uint64_t		pe_guid;
76fa9e4066Sahrens 	vdev_entry_t		*pe_vdevs;
77fa9e4066Sahrens 	struct pool_entry	*pe_next;
78fa9e4066Sahrens } pool_entry_t;
79fa9e4066Sahrens 
80fa9e4066Sahrens typedef struct name_entry {
8199653d4eSeschrock 	char			*ne_name;
82fa9e4066Sahrens 	uint64_t		ne_guid;
83fa9e4066Sahrens 	struct name_entry	*ne_next;
84fa9e4066Sahrens } name_entry_t;
85fa9e4066Sahrens 
86fa9e4066Sahrens typedef struct pool_list {
87fa9e4066Sahrens 	pool_entry_t		*pools;
88fa9e4066Sahrens 	name_entry_t		*names;
89fa9e4066Sahrens } pool_list_t;
90fa9e4066Sahrens 
91fa9e4066Sahrens static char *
92fa9e4066Sahrens get_devid(const char *path)
93fa9e4066Sahrens {
94fa9e4066Sahrens 	int fd;
95fa9e4066Sahrens 	ddi_devid_t devid;
96fa9e4066Sahrens 	char *minor, *ret;
97fa9e4066Sahrens 
98fa9e4066Sahrens 	if ((fd = open(path, O_RDONLY)) < 0)
99fa9e4066Sahrens 		return (NULL);
100fa9e4066Sahrens 
101fa9e4066Sahrens 	minor = NULL;
102fa9e4066Sahrens 	ret = NULL;
103fa9e4066Sahrens 	if (devid_get(fd, &devid) == 0) {
104fa9e4066Sahrens 		if (devid_get_minor_name(fd, &minor) == 0)
105fa9e4066Sahrens 			ret = devid_str_encode(devid, minor);
106fa9e4066Sahrens 		if (minor != NULL)
107fa9e4066Sahrens 			devid_str_free(minor);
108fa9e4066Sahrens 		devid_free(devid);
109fa9e4066Sahrens 	}
110c67d9675Seschrock 	(void) close(fd);
111fa9e4066Sahrens 
112fa9e4066Sahrens 	return (ret);
113fa9e4066Sahrens }
114fa9e4066Sahrens 
115fa9e4066Sahrens 
116fa9e4066Sahrens /*
117fa9e4066Sahrens  * Go through and fix up any path and/or devid information for the given vdev
118fa9e4066Sahrens  * configuration.
119fa9e4066Sahrens  */
12099653d4eSeschrock static int
121fa9e4066Sahrens fix_paths(nvlist_t *nv, name_entry_t *names)
122fa9e4066Sahrens {
123fa9e4066Sahrens 	nvlist_t **child;
124fa9e4066Sahrens 	uint_t c, children;
125fa9e4066Sahrens 	uint64_t guid;
126c67d9675Seschrock 	name_entry_t *ne, *best;
127c67d9675Seschrock 	char *path, *devid;
128c67d9675Seschrock 	int matched;
129fa9e4066Sahrens 
130fa9e4066Sahrens 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
131fa9e4066Sahrens 	    &child, &children) == 0) {
132fa9e4066Sahrens 		for (c = 0; c < children; c++)
13399653d4eSeschrock 			if (fix_paths(child[c], names) != 0)
13499653d4eSeschrock 				return (-1);
13599653d4eSeschrock 		return (0);
136fa9e4066Sahrens 	}
137fa9e4066Sahrens 
138fa9e4066Sahrens 	/*
139fa9e4066Sahrens 	 * This is a leaf (file or disk) vdev.  In either case, go through
140fa9e4066Sahrens 	 * the name list and see if we find a matching guid.  If so, replace
141fa9e4066Sahrens 	 * the path and see if we can calculate a new devid.
142c67d9675Seschrock 	 *
143c67d9675Seschrock 	 * There may be multiple names associated with a particular guid, in
144c67d9675Seschrock 	 * which case we have overlapping slices or multiple paths to the same
145c67d9675Seschrock 	 * disk.  If this is the case, then we want to pick the path that is
146c67d9675Seschrock 	 * the most similar to the original, where "most similar" is the number
147c67d9675Seschrock 	 * of matching characters starting from the end of the path.  This will
148c67d9675Seschrock 	 * preserve slice numbers even if the disks have been reorganized, and
149c67d9675Seschrock 	 * will also catch preferred disk names if multiple paths exist.
150fa9e4066Sahrens 	 */
151fa9e4066Sahrens 	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0);
152c67d9675Seschrock 	if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0)
153c67d9675Seschrock 		path = NULL;
154c67d9675Seschrock 
155c67d9675Seschrock 	matched = 0;
156c67d9675Seschrock 	best = NULL;
157c67d9675Seschrock 	for (ne = names; ne != NULL; ne = ne->ne_next) {
158c67d9675Seschrock 		if (ne->ne_guid == guid) {
159c67d9675Seschrock 			const char *src, *dst;
160c67d9675Seschrock 			int count;
161c67d9675Seschrock 
162c67d9675Seschrock 			if (path == NULL) {
163c67d9675Seschrock 				best = ne;
164c67d9675Seschrock 				break;
165c67d9675Seschrock 			}
166c67d9675Seschrock 
167c67d9675Seschrock 			src = ne->ne_name + strlen(ne->ne_name) - 1;
168c67d9675Seschrock 			dst = path + strlen(path) - 1;
169c67d9675Seschrock 			for (count = 0; src >= ne->ne_name && dst >= path;
170c67d9675Seschrock 			    src--, dst--, count++)
171c67d9675Seschrock 				if (*src != *dst)
172c67d9675Seschrock 					break;
173c67d9675Seschrock 
174c67d9675Seschrock 			/*
175c67d9675Seschrock 			 * At this point, 'count' is the number of characters
176c67d9675Seschrock 			 * matched from the end.
177c67d9675Seschrock 			 */
178c67d9675Seschrock 			if (count > matched || best == NULL) {
179c67d9675Seschrock 				best = ne;
180c67d9675Seschrock 				matched = count;
181c67d9675Seschrock 			}
182c67d9675Seschrock 		}
183c67d9675Seschrock 	}
184fa9e4066Sahrens 
185c67d9675Seschrock 	if (best == NULL)
18699653d4eSeschrock 		return (0);
187fa9e4066Sahrens 
18899653d4eSeschrock 	if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0)
18999653d4eSeschrock 		return (-1);
190fa9e4066Sahrens 
191c67d9675Seschrock 	if ((devid = get_devid(best->ne_name)) == NULL) {
192fa9e4066Sahrens 		(void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID);
193fa9e4066Sahrens 	} else {
19499653d4eSeschrock 		if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0)
19599653d4eSeschrock 			return (-1);
196fa9e4066Sahrens 		devid_str_free(devid);
197fa9e4066Sahrens 	}
19899653d4eSeschrock 
19999653d4eSeschrock 	return (0);
200fa9e4066Sahrens }
201fa9e4066Sahrens 
202fa9e4066Sahrens /*
203fa9e4066Sahrens  * Add the given configuration to the list of known devices.
204fa9e4066Sahrens  */
20599653d4eSeschrock static int
20699653d4eSeschrock add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path,
20799653d4eSeschrock     nvlist_t *config)
208fa9e4066Sahrens {
20999653d4eSeschrock 	uint64_t pool_guid, vdev_guid, top_guid, txg, state;
210fa9e4066Sahrens 	pool_entry_t *pe;
211fa9e4066Sahrens 	vdev_entry_t *ve;
212fa9e4066Sahrens 	config_entry_t *ce;
213fa9e4066Sahrens 	name_entry_t *ne;
214fa9e4066Sahrens 
21599653d4eSeschrock 	/*
216fa94a07fSbrendan 	 * If this is a hot spare not currently in use or level 2 cache
217fa94a07fSbrendan 	 * device, add it to the list of names to translate, but don't do
218fa94a07fSbrendan 	 * anything else.
21999653d4eSeschrock 	 */
22099653d4eSeschrock 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
221fa94a07fSbrendan 	    &state) == 0 &&
222fa94a07fSbrendan 	    (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) &&
22399653d4eSeschrock 	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) {
22499653d4eSeschrock 		if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
225ccae0b50Seschrock 			return (-1);
22699653d4eSeschrock 
22799653d4eSeschrock 		if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
22899653d4eSeschrock 			free(ne);
22999653d4eSeschrock 			return (-1);
23099653d4eSeschrock 		}
23199653d4eSeschrock 		ne->ne_guid = vdev_guid;
23299653d4eSeschrock 		ne->ne_next = pl->names;
23399653d4eSeschrock 		pl->names = ne;
23499653d4eSeschrock 		return (0);
23599653d4eSeschrock 	}
23699653d4eSeschrock 
237fa9e4066Sahrens 	/*
238fa9e4066Sahrens 	 * If we have a valid config but cannot read any of these fields, then
239fa9e4066Sahrens 	 * it means we have a half-initialized label.  In vdev_label_init()
240fa9e4066Sahrens 	 * we write a label with txg == 0 so that we can identify the device
241fa9e4066Sahrens 	 * in case the user refers to the same disk later on.  If we fail to
242fa9e4066Sahrens 	 * create the pool, we'll be left with a label in this state
243fa9e4066Sahrens 	 * which should not be considered part of a valid pool.
244fa9e4066Sahrens 	 */
245fa9e4066Sahrens 	if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
246fa9e4066Sahrens 	    &pool_guid) != 0 ||
247fa9e4066Sahrens 	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
248fa9e4066Sahrens 	    &vdev_guid) != 0 ||
249fa9e4066Sahrens 	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID,
250fa9e4066Sahrens 	    &top_guid) != 0 ||
251fa9e4066Sahrens 	    nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG,
252fa9e4066Sahrens 	    &txg) != 0 || txg == 0) {
253fa9e4066Sahrens 		nvlist_free(config);
25499653d4eSeschrock 		return (0);
255fa9e4066Sahrens 	}
256fa9e4066Sahrens 
257fa9e4066Sahrens 	/*
258fa9e4066Sahrens 	 * First, see if we know about this pool.  If not, then add it to the
259fa9e4066Sahrens 	 * list of known pools.
260fa9e4066Sahrens 	 */
261fa9e4066Sahrens 	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
262fa9e4066Sahrens 		if (pe->pe_guid == pool_guid)
263fa9e4066Sahrens 			break;
264fa9e4066Sahrens 	}
265fa9e4066Sahrens 
266fa9e4066Sahrens 	if (pe == NULL) {
26799653d4eSeschrock 		if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) {
26899653d4eSeschrock 			nvlist_free(config);
26999653d4eSeschrock 			return (-1);
27099653d4eSeschrock 		}
271fa9e4066Sahrens 		pe->pe_guid = pool_guid;
272fa9e4066Sahrens 		pe->pe_next = pl->pools;
273fa9e4066Sahrens 		pl->pools = pe;
274fa9e4066Sahrens 	}
275fa9e4066Sahrens 
276fa9e4066Sahrens 	/*
277fa9e4066Sahrens 	 * Second, see if we know about this toplevel vdev.  Add it if its
278fa9e4066Sahrens 	 * missing.
279fa9e4066Sahrens 	 */
280fa9e4066Sahrens 	for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
281fa9e4066Sahrens 		if (ve->ve_guid == top_guid)
282fa9e4066Sahrens 			break;
283fa9e4066Sahrens 	}
284fa9e4066Sahrens 
285fa9e4066Sahrens 	if (ve == NULL) {
28699653d4eSeschrock 		if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) {
28799653d4eSeschrock 			nvlist_free(config);
28899653d4eSeschrock 			return (-1);
28999653d4eSeschrock 		}
290fa9e4066Sahrens 		ve->ve_guid = top_guid;
291fa9e4066Sahrens 		ve->ve_next = pe->pe_vdevs;
292fa9e4066Sahrens 		pe->pe_vdevs = ve;
293fa9e4066Sahrens 	}
294fa9e4066Sahrens 
295fa9e4066Sahrens 	/*
296fa9e4066Sahrens 	 * Third, see if we have a config with a matching transaction group.  If
297fa9e4066Sahrens 	 * so, then we do nothing.  Otherwise, add it to the list of known
298fa9e4066Sahrens 	 * configs.
299fa9e4066Sahrens 	 */
300fa9e4066Sahrens 	for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) {
301fa9e4066Sahrens 		if (ce->ce_txg == txg)
302fa9e4066Sahrens 			break;
303fa9e4066Sahrens 	}
304fa9e4066Sahrens 
305fa9e4066Sahrens 	if (ce == NULL) {
30699653d4eSeschrock 		if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) {
30799653d4eSeschrock 			nvlist_free(config);
30899653d4eSeschrock 			return (-1);
30999653d4eSeschrock 		}
310fa9e4066Sahrens 		ce->ce_txg = txg;
311fa9e4066Sahrens 		ce->ce_config = config;
312fa9e4066Sahrens 		ce->ce_next = ve->ve_configs;
313fa9e4066Sahrens 		ve->ve_configs = ce;
314fa9e4066Sahrens 	} else {
315fa9e4066Sahrens 		nvlist_free(config);
316fa9e4066Sahrens 	}
317fa9e4066Sahrens 
318fa9e4066Sahrens 	/*
319fa9e4066Sahrens 	 * At this point we've successfully added our config to the list of
320fa9e4066Sahrens 	 * known configs.  The last thing to do is add the vdev guid -> path
321fa9e4066Sahrens 	 * mappings so that we can fix up the configuration as necessary before
322fa9e4066Sahrens 	 * doing the import.
323fa9e4066Sahrens 	 */
32499653d4eSeschrock 	if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL)
32599653d4eSeschrock 		return (-1);
32699653d4eSeschrock 
32799653d4eSeschrock 	if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) {
32899653d4eSeschrock 		free(ne);
32999653d4eSeschrock 		return (-1);
33099653d4eSeschrock 	}
331fa9e4066Sahrens 
332fa9e4066Sahrens 	ne->ne_guid = vdev_guid;
333fa9e4066Sahrens 	ne->ne_next = pl->names;
334fa9e4066Sahrens 	pl->names = ne;
33599653d4eSeschrock 
33699653d4eSeschrock 	return (0);
337fa9e4066Sahrens }
338fa9e4066Sahrens 
339eaca9bbdSeschrock /*
340eaca9bbdSeschrock  * Returns true if the named pool matches the given GUID.
341eaca9bbdSeschrock  */
34294de1d4cSeschrock static int
34394de1d4cSeschrock pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid,
34494de1d4cSeschrock     boolean_t *isactive)
345eaca9bbdSeschrock {
346eaca9bbdSeschrock 	zpool_handle_t *zhp;
347eaca9bbdSeschrock 	uint64_t theguid;
348eaca9bbdSeschrock 
34994de1d4cSeschrock 	if (zpool_open_silent(hdl, name, &zhp) != 0)
35094de1d4cSeschrock 		return (-1);
35194de1d4cSeschrock 
35294de1d4cSeschrock 	if (zhp == NULL) {
35394de1d4cSeschrock 		*isactive = B_FALSE;
35494de1d4cSeschrock 		return (0);
35594de1d4cSeschrock 	}
356eaca9bbdSeschrock 
357eaca9bbdSeschrock 	verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID,
358eaca9bbdSeschrock 	    &theguid) == 0);
359eaca9bbdSeschrock 
360eaca9bbdSeschrock 	zpool_close(zhp);
361eaca9bbdSeschrock 
36294de1d4cSeschrock 	*isactive = (theguid == guid);
36394de1d4cSeschrock 	return (0);
364eaca9bbdSeschrock }
365eaca9bbdSeschrock 
3662f8aaab3Seschrock static nvlist_t *
3672f8aaab3Seschrock refresh_config(libzfs_handle_t *hdl, nvlist_t *config)
3682f8aaab3Seschrock {
3692f8aaab3Seschrock 	nvlist_t *nvl;
3702f8aaab3Seschrock 	zfs_cmd_t zc = { 0 };
3712f8aaab3Seschrock 	int err;
3722f8aaab3Seschrock 
3732f8aaab3Seschrock 	if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0)
3742f8aaab3Seschrock 		return (NULL);
3752f8aaab3Seschrock 
3762f8aaab3Seschrock 	if (zcmd_alloc_dst_nvlist(hdl, &zc,
3772f8aaab3Seschrock 	    zc.zc_nvlist_conf_size * 2) != 0) {
3782f8aaab3Seschrock 		zcmd_free_nvlists(&zc);
3792f8aaab3Seschrock 		return (NULL);
3802f8aaab3Seschrock 	}
3812f8aaab3Seschrock 
3822f8aaab3Seschrock 	while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT,
3832f8aaab3Seschrock 	    &zc)) != 0 && errno == ENOMEM) {
3842f8aaab3Seschrock 		if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3852f8aaab3Seschrock 			zcmd_free_nvlists(&zc);
3862f8aaab3Seschrock 			return (NULL);
3872f8aaab3Seschrock 		}
3882f8aaab3Seschrock 	}
3892f8aaab3Seschrock 
3902f8aaab3Seschrock 	if (err) {
3912f8aaab3Seschrock 		(void) zpool_standard_error(hdl, errno,
3922f8aaab3Seschrock 		    dgettext(TEXT_DOMAIN, "cannot discover pools"));
3932f8aaab3Seschrock 		zcmd_free_nvlists(&zc);
3942f8aaab3Seschrock 		return (NULL);
3952f8aaab3Seschrock 	}
3962f8aaab3Seschrock 
3972f8aaab3Seschrock 	if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) {
3982f8aaab3Seschrock 		zcmd_free_nvlists(&zc);
3992f8aaab3Seschrock 		return (NULL);
4002f8aaab3Seschrock 	}
4012f8aaab3Seschrock 
4022f8aaab3Seschrock 	zcmd_free_nvlists(&zc);
4032f8aaab3Seschrock 	return (nvl);
4042f8aaab3Seschrock }
4052f8aaab3Seschrock 
406fa9e4066Sahrens /*
407fa9e4066Sahrens  * Convert our list of pools into the definitive set of configurations.  We
408fa9e4066Sahrens  * start by picking the best config for each toplevel vdev.  Once that's done,
409fa9e4066Sahrens  * we assemble the toplevel vdevs into a full config for the pool.  We make a
410fa9e4066Sahrens  * pass to fix up any incorrect paths, and then add it to the main list to
411fa9e4066Sahrens  * return to the user.
412fa9e4066Sahrens  */
413fa9e4066Sahrens static nvlist_t *
41499653d4eSeschrock get_configs(libzfs_handle_t *hdl, pool_list_t *pl)
415fa9e4066Sahrens {
41699653d4eSeschrock 	pool_entry_t *pe;
41799653d4eSeschrock 	vdev_entry_t *ve;
41899653d4eSeschrock 	config_entry_t *ce;
41999653d4eSeschrock 	nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot;
420fa94a07fSbrendan 	nvlist_t **spares, **l2cache;
421fa94a07fSbrendan 	uint_t i, nspares, nl2cache;
42299653d4eSeschrock 	boolean_t config_seen;
423fa9e4066Sahrens 	uint64_t best_txg;
42495173954Sek 	char *name, *hostname;
42599653d4eSeschrock 	uint64_t version, guid;
42699653d4eSeschrock 	uint_t children = 0;
42799653d4eSeschrock 	nvlist_t **child = NULL;
42899653d4eSeschrock 	uint_t c;
42994de1d4cSeschrock 	boolean_t isactive;
43095173954Sek 	uint64_t hostid;
4312f8aaab3Seschrock 	nvlist_t *nvl;
432fa9e4066Sahrens 
43399653d4eSeschrock 	if (nvlist_alloc(&ret, 0, 0) != 0)
43499653d4eSeschrock 		goto nomem;
435fa9e4066Sahrens 
43699653d4eSeschrock 	for (pe = pl->pools; pe != NULL; pe = pe->pe_next) {
437fa9e4066Sahrens 		uint64_t id;
438fa9e4066Sahrens 
43999653d4eSeschrock 		if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0)
44099653d4eSeschrock 			goto nomem;
44199653d4eSeschrock 		config_seen = B_FALSE;
442fa9e4066Sahrens 
443fa9e4066Sahrens 		/*
444fa9e4066Sahrens 		 * Iterate over all toplevel vdevs.  Grab the pool configuration
445fa9e4066Sahrens 		 * from the first one we find, and then go through the rest and
446fa9e4066Sahrens 		 * add them as necessary to the 'vdevs' member of the config.
447fa9e4066Sahrens 		 */
44899653d4eSeschrock 		for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) {
449fa9e4066Sahrens 
450fa9e4066Sahrens 			/*
451fa9e4066Sahrens 			 * Determine the best configuration for this vdev by
452fa9e4066Sahrens 			 * selecting the config with the latest transaction
453fa9e4066Sahrens 			 * group.
454fa9e4066Sahrens 			 */
455fa9e4066Sahrens 			best_txg = 0;
456fa9e4066Sahrens 			for (ce = ve->ve_configs; ce != NULL;
457fa9e4066Sahrens 			    ce = ce->ce_next) {
458fa9e4066Sahrens 
45999653d4eSeschrock 				if (ce->ce_txg > best_txg) {
460fa9e4066Sahrens 					tmp = ce->ce_config;
46199653d4eSeschrock 					best_txg = ce->ce_txg;
46299653d4eSeschrock 				}
463fa9e4066Sahrens 			}
464fa9e4066Sahrens 
465fa9e4066Sahrens 			if (!config_seen) {
466fa9e4066Sahrens 				/*
467fa9e4066Sahrens 				 * Copy the relevant pieces of data to the pool
468fa9e4066Sahrens 				 * configuration:
469fa9e4066Sahrens 				 *
47099653d4eSeschrock 				 *	version
471fa9e4066Sahrens 				 * 	pool guid
472fa9e4066Sahrens 				 * 	name
473fa9e4066Sahrens 				 * 	pool state
47495173954Sek 				 *	hostid (if available)
47595173954Sek 				 *	hostname (if available)
476fa9e4066Sahrens 				 */
477fa9e4066Sahrens 				uint64_t state;
478fa9e4066Sahrens 
47999653d4eSeschrock 				verify(nvlist_lookup_uint64(tmp,
48099653d4eSeschrock 				    ZPOOL_CONFIG_VERSION, &version) == 0);
48199653d4eSeschrock 				if (nvlist_add_uint64(config,
48299653d4eSeschrock 				    ZPOOL_CONFIG_VERSION, version) != 0)
48399653d4eSeschrock 					goto nomem;
484fa9e4066Sahrens 				verify(nvlist_lookup_uint64(tmp,
485fa9e4066Sahrens 				    ZPOOL_CONFIG_POOL_GUID, &guid) == 0);
48699653d4eSeschrock 				if (nvlist_add_uint64(config,
48799653d4eSeschrock 				    ZPOOL_CONFIG_POOL_GUID, guid) != 0)
48899653d4eSeschrock 					goto nomem;
489fa9e4066Sahrens 				verify(nvlist_lookup_string(tmp,
490fa9e4066Sahrens 				    ZPOOL_CONFIG_POOL_NAME, &name) == 0);
49199653d4eSeschrock 				if (nvlist_add_string(config,
49299653d4eSeschrock 				    ZPOOL_CONFIG_POOL_NAME, name) != 0)
49399653d4eSeschrock 					goto nomem;
494fa9e4066Sahrens 				verify(nvlist_lookup_uint64(tmp,
495fa9e4066Sahrens 				    ZPOOL_CONFIG_POOL_STATE, &state) == 0);
49699653d4eSeschrock 				if (nvlist_add_uint64(config,
49799653d4eSeschrock 				    ZPOOL_CONFIG_POOL_STATE, state) != 0)
49899653d4eSeschrock 					goto nomem;
49995173954Sek 				hostid = 0;
50095173954Sek 				if (nvlist_lookup_uint64(tmp,
50195173954Sek 				    ZPOOL_CONFIG_HOSTID, &hostid) == 0) {
50295173954Sek 					if (nvlist_add_uint64(config,
50395173954Sek 					    ZPOOL_CONFIG_HOSTID, hostid) != 0)
50495173954Sek 						goto nomem;
50595173954Sek 					verify(nvlist_lookup_string(tmp,
50695173954Sek 					    ZPOOL_CONFIG_HOSTNAME,
50795173954Sek 					    &hostname) == 0);
50895173954Sek 					if (nvlist_add_string(config,
50995173954Sek 					    ZPOOL_CONFIG_HOSTNAME,
51095173954Sek 					    hostname) != 0)
51195173954Sek 						goto nomem;
51295173954Sek 				}
513fa9e4066Sahrens 
51499653d4eSeschrock 				config_seen = B_TRUE;
515fa9e4066Sahrens 			}
516fa9e4066Sahrens 
517fa9e4066Sahrens 			/*
518fa9e4066Sahrens 			 * Add this top-level vdev to the child array.
519fa9e4066Sahrens 			 */
520fa9e4066Sahrens 			verify(nvlist_lookup_nvlist(tmp,
521fa9e4066Sahrens 			    ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0);
522fa9e4066Sahrens 			verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID,
523fa9e4066Sahrens 			    &id) == 0);
524fa9e4066Sahrens 			if (id >= children) {
525fa9e4066Sahrens 				nvlist_t **newchild;
526fa9e4066Sahrens 
52799653d4eSeschrock 				newchild = zfs_alloc(hdl, (id + 1) *
528fa9e4066Sahrens 				    sizeof (nvlist_t *));
52999653d4eSeschrock 				if (newchild == NULL)
53099653d4eSeschrock 					goto nomem;
531fa9e4066Sahrens 
532fa9e4066Sahrens 				for (c = 0; c < children; c++)
533fa9e4066Sahrens 					newchild[c] = child[c];
534fa9e4066Sahrens 
535fa9e4066Sahrens 				free(child);
536fa9e4066Sahrens 				child = newchild;
537fa9e4066Sahrens 				children = id + 1;
538fa9e4066Sahrens 			}
53999653d4eSeschrock 			if (nvlist_dup(nvtop, &child[id], 0) != 0)
54099653d4eSeschrock 				goto nomem;
541fa9e4066Sahrens 
542fa9e4066Sahrens 		}
543fa9e4066Sahrens 
544fa9e4066Sahrens 		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
545fa9e4066Sahrens 		    &guid) == 0);
546fa9e4066Sahrens 
547fa9e4066Sahrens 		/*
548fa9e4066Sahrens 		 * Look for any missing top-level vdevs.  If this is the case,
549fa9e4066Sahrens 		 * create a faked up 'missing' vdev as a placeholder.  We cannot
550fa9e4066Sahrens 		 * simply compress the child array, because the kernel performs
551fa9e4066Sahrens 		 * certain checks to make sure the vdev IDs match their location
552fa9e4066Sahrens 		 * in the configuration.
553fa9e4066Sahrens 		 */
554fa9e4066Sahrens 		for (c = 0; c < children; c++)
555fa9e4066Sahrens 			if (child[c] == NULL) {
556fa9e4066Sahrens 				nvlist_t *missing;
55799653d4eSeschrock 				if (nvlist_alloc(&missing, NV_UNIQUE_NAME,
55899653d4eSeschrock 				    0) != 0)
55999653d4eSeschrock 					goto nomem;
56099653d4eSeschrock 				if (nvlist_add_string(missing,
56199653d4eSeschrock 				    ZPOOL_CONFIG_TYPE,
56299653d4eSeschrock 				    VDEV_TYPE_MISSING) != 0 ||
56399653d4eSeschrock 				    nvlist_add_uint64(missing,
56499653d4eSeschrock 				    ZPOOL_CONFIG_ID, c) != 0 ||
56599653d4eSeschrock 				    nvlist_add_uint64(missing,
56699653d4eSeschrock 				    ZPOOL_CONFIG_GUID, 0ULL) != 0) {
56799653d4eSeschrock 					nvlist_free(missing);
56899653d4eSeschrock 					goto nomem;
56999653d4eSeschrock 				}
570fa9e4066Sahrens 				child[c] = missing;
571fa9e4066Sahrens 			}
572fa9e4066Sahrens 
573fa9e4066Sahrens 		/*
574fa9e4066Sahrens 		 * Put all of this pool's top-level vdevs into a root vdev.
575fa9e4066Sahrens 		 */
57699653d4eSeschrock 		if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0)
57799653d4eSeschrock 			goto nomem;
57899653d4eSeschrock 		if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
57999653d4eSeschrock 		    VDEV_TYPE_ROOT) != 0 ||
58099653d4eSeschrock 		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 ||
58199653d4eSeschrock 		    nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 ||
58299653d4eSeschrock 		    nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
58399653d4eSeschrock 		    child, children) != 0) {
58499653d4eSeschrock 			nvlist_free(nvroot);
58599653d4eSeschrock 			goto nomem;
58699653d4eSeschrock 		}
587fa9e4066Sahrens 
588fa9e4066Sahrens 		for (c = 0; c < children; c++)
589fa9e4066Sahrens 			nvlist_free(child[c]);
590fa9e4066Sahrens 		free(child);
59199653d4eSeschrock 		children = 0;
59299653d4eSeschrock 		child = NULL;
593fa9e4066Sahrens 
594fa9e4066Sahrens 		/*
595fa9e4066Sahrens 		 * Go through and fix up any paths and/or devids based on our
596fa9e4066Sahrens 		 * known list of vdev GUID -> path mappings.
597fa9e4066Sahrens 		 */
59899653d4eSeschrock 		if (fix_paths(nvroot, pl->names) != 0) {
59999653d4eSeschrock 			nvlist_free(nvroot);
60099653d4eSeschrock 			goto nomem;
60199653d4eSeschrock 		}
602fa9e4066Sahrens 
603fa9e4066Sahrens 		/*
604fa9e4066Sahrens 		 * Add the root vdev to this pool's configuration.
605fa9e4066Sahrens 		 */
60699653d4eSeschrock 		if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
60799653d4eSeschrock 		    nvroot) != 0) {
60899653d4eSeschrock 			nvlist_free(nvroot);
60999653d4eSeschrock 			goto nomem;
61099653d4eSeschrock 		}
611fa9e4066Sahrens 		nvlist_free(nvroot);
612fa9e4066Sahrens 
613fa9e4066Sahrens 		/*
614fa9e4066Sahrens 		 * Determine if this pool is currently active, in which case we
615fa9e4066Sahrens 		 * can't actually import it.
616fa9e4066Sahrens 		 */
617fa9e4066Sahrens 		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
618fa9e4066Sahrens 		    &name) == 0);
619fa9e4066Sahrens 		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
620fa9e4066Sahrens 		    &guid) == 0);
621fa9e4066Sahrens 
62294de1d4cSeschrock 		if (pool_active(hdl, name, guid, &isactive) != 0)
62394de1d4cSeschrock 			goto error;
62494de1d4cSeschrock 
6250192a278Seschrock 		if (isactive) {
626fa9e4066Sahrens 			nvlist_free(config);
62799653d4eSeschrock 			config = NULL;
628fa9e4066Sahrens 			continue;
629fa9e4066Sahrens 		}
630fa9e4066Sahrens 
6312f8aaab3Seschrock 		if ((nvl = refresh_config(hdl, config)) == NULL)
632e9dbad6fSeschrock 			goto error;
633fa9e4066Sahrens 
634fa9e4066Sahrens 		nvlist_free(config);
6352f8aaab3Seschrock 		config = nvl;
636fa9e4066Sahrens 
63799653d4eSeschrock 		/*
63899653d4eSeschrock 		 * Go through and update the paths for spares, now that we have
63999653d4eSeschrock 		 * them.
64099653d4eSeschrock 		 */
64199653d4eSeschrock 		verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
64299653d4eSeschrock 		    &nvroot) == 0);
64399653d4eSeschrock 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
64499653d4eSeschrock 		    &spares, &nspares) == 0) {
64599653d4eSeschrock 			for (i = 0; i < nspares; i++) {
64699653d4eSeschrock 				if (fix_paths(spares[i], pl->names) != 0)
64799653d4eSeschrock 					goto nomem;
64899653d4eSeschrock 			}
64999653d4eSeschrock 		}
65099653d4eSeschrock 
651fa94a07fSbrendan 		/*
652fa94a07fSbrendan 		 * Update the paths for l2cache devices.
653fa94a07fSbrendan 		 */
654fa94a07fSbrendan 		if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
655fa94a07fSbrendan 		    &l2cache, &nl2cache) == 0) {
656fa94a07fSbrendan 			for (i = 0; i < nl2cache; i++) {
657fa94a07fSbrendan 				if (fix_paths(l2cache[i], pl->names) != 0)
658fa94a07fSbrendan 					goto nomem;
659fa94a07fSbrendan 			}
660fa94a07fSbrendan 		}
661fa94a07fSbrendan 
66295173954Sek 		/*
66395173954Sek 		 * Restore the original information read from the actual label.
66495173954Sek 		 */
66595173954Sek 		(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID,
66695173954Sek 		    DATA_TYPE_UINT64);
66795173954Sek 		(void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME,
66895173954Sek 		    DATA_TYPE_STRING);
66995173954Sek 		if (hostid != 0) {
67095173954Sek 			verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID,
67195173954Sek 			    hostid) == 0);
67295173954Sek 			verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME,
67395173954Sek 			    hostname) == 0);
67495173954Sek 		}
67595173954Sek 
676fa9e4066Sahrens 		/*
677fa9e4066Sahrens 		 * Add this pool to the list of configs.
678fa9e4066Sahrens 		 */
679e9dbad6fSeschrock 		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
680e9dbad6fSeschrock 		    &name) == 0);
68199653d4eSeschrock 		if (nvlist_add_nvlist(ret, name, config) != 0)
68299653d4eSeschrock 			goto nomem;
683fa9e4066Sahrens 
684fa9e4066Sahrens 		nvlist_free(config);
68599653d4eSeschrock 		config = NULL;
686fa9e4066Sahrens 	}
687fa9e4066Sahrens 
688fa9e4066Sahrens 	return (ret);
68999653d4eSeschrock 
69099653d4eSeschrock nomem:
69199653d4eSeschrock 	(void) no_memory(hdl);
69299653d4eSeschrock error:
69394de1d4cSeschrock 	nvlist_free(config);
69494de1d4cSeschrock 	nvlist_free(ret);
69599653d4eSeschrock 	for (c = 0; c < children; c++)
69699653d4eSeschrock 		nvlist_free(child[c]);
69794de1d4cSeschrock 	free(child);
69899653d4eSeschrock 
69999653d4eSeschrock 	return (NULL);
700fa9e4066Sahrens }
701fa9e4066Sahrens 
702fa9e4066Sahrens /*
703fa9e4066Sahrens  * Return the offset of the given label.
704fa9e4066Sahrens  */
705fa9e4066Sahrens static uint64_t
706e7437265Sahrens label_offset(uint64_t size, int l)
707fa9e4066Sahrens {
708e7437265Sahrens 	ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0);
709fa9e4066Sahrens 	return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ?
710fa9e4066Sahrens 	    0 : size - VDEV_LABELS * sizeof (vdev_label_t)));
711fa9e4066Sahrens }
712fa9e4066Sahrens 
713fa9e4066Sahrens /*
714fa9e4066Sahrens  * Given a file descriptor, read the label information and return an nvlist
715fa9e4066Sahrens  * describing the configuration, if there is one.
716fa9e4066Sahrens  */
71799653d4eSeschrock int
71899653d4eSeschrock zpool_read_label(int fd, nvlist_t **config)
719fa9e4066Sahrens {
720fa9e4066Sahrens 	struct stat64 statbuf;
721fa9e4066Sahrens 	int l;
722fa9e4066Sahrens 	vdev_label_t *label;
723e7437265Sahrens 	uint64_t state, txg, size;
724fa9e4066Sahrens 
72599653d4eSeschrock 	*config = NULL;
72699653d4eSeschrock 
727fa9e4066Sahrens 	if (fstat64(fd, &statbuf) == -1)
72899653d4eSeschrock 		return (0);
729e7437265Sahrens 	size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t);
730fa9e4066Sahrens 
73199653d4eSeschrock 	if ((label = malloc(sizeof (vdev_label_t))) == NULL)
73299653d4eSeschrock 		return (-1);
733fa9e4066Sahrens 
734fa9e4066Sahrens 	for (l = 0; l < VDEV_LABELS; l++) {
735fa9e4066Sahrens 		if (pread(fd, label, sizeof (vdev_label_t),
736e7437265Sahrens 		    label_offset(size, l)) != sizeof (vdev_label_t))
737fa9e4066Sahrens 			continue;
738fa9e4066Sahrens 
739fa9e4066Sahrens 		if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
74099653d4eSeschrock 		    sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0)
741fa9e4066Sahrens 			continue;
742fa9e4066Sahrens 
74399653d4eSeschrock 		if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE,
744fa94a07fSbrendan 		    &state) != 0 || state > POOL_STATE_L2CACHE) {
74599653d4eSeschrock 			nvlist_free(*config);
746fa9e4066Sahrens 			continue;
747fa9e4066Sahrens 		}
748fa9e4066Sahrens 
749fa94a07fSbrendan 		if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE &&
75099653d4eSeschrock 		    (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG,
75199653d4eSeschrock 		    &txg) != 0 || txg == 0)) {
75299653d4eSeschrock 			nvlist_free(*config);
753fa9e4066Sahrens 			continue;
754fa9e4066Sahrens 		}
755fa9e4066Sahrens 
756fa9e4066Sahrens 		free(label);
75799653d4eSeschrock 		return (0);
758fa9e4066Sahrens 	}
759fa9e4066Sahrens 
760fa9e4066Sahrens 	free(label);
76199653d4eSeschrock 	*config = NULL;
76299653d4eSeschrock 	return (0);
763fa9e4066Sahrens }
764fa9e4066Sahrens 
765fa9e4066Sahrens /*
766fa9e4066Sahrens  * Given a list of directories to search, find all pools stored on disk.  This
767fa9e4066Sahrens  * includes partial pools which are not available to import.  If no args are
768fa9e4066Sahrens  * given (argc is 0), then the default directory (/dev/dsk) is searched.
769fa9e4066Sahrens  */
770fa9e4066Sahrens nvlist_t *
77199653d4eSeschrock zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv)
772fa9e4066Sahrens {
773fa9e4066Sahrens 	int i;
774ccae0b50Seschrock 	DIR *dirp = NULL;
775fa9e4066Sahrens 	struct dirent64 *dp;
776fa9e4066Sahrens 	char path[MAXPATHLEN];
777fa9e4066Sahrens 	struct stat64 statbuf;
77899653d4eSeschrock 	nvlist_t *ret = NULL, *config;
779fa9e4066Sahrens 	static char *default_dir = "/dev/dsk";
780fa9e4066Sahrens 	int fd;
781fa9e4066Sahrens 	pool_list_t pools = { 0 };
78299653d4eSeschrock 	pool_entry_t *pe, *penext;
78399653d4eSeschrock 	vdev_entry_t *ve, *venext;
78499653d4eSeschrock 	config_entry_t *ce, *cenext;
78599653d4eSeschrock 	name_entry_t *ne, *nenext;
78699653d4eSeschrock 
787fa9e4066Sahrens 
788fa9e4066Sahrens 	if (argc == 0) {
789fa9e4066Sahrens 		argc = 1;
790fa9e4066Sahrens 		argv = &default_dir;
791fa9e4066Sahrens 	}
792fa9e4066Sahrens 
793fa9e4066Sahrens 	/*
794fa9e4066Sahrens 	 * Go through and read the label configuration information from every
795fa9e4066Sahrens 	 * possible device, organizing the information according to pool GUID
796fa9e4066Sahrens 	 * and toplevel GUID.
797fa9e4066Sahrens 	 */
798fa9e4066Sahrens 	for (i = 0; i < argc; i++) {
799fa9e4066Sahrens 		if (argv[i][0] != '/') {
800ece3d9b3Slling 			(void) zfs_error_fmt(hdl, EZFS_BADPATH,
80199653d4eSeschrock 			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
802fa9e4066Sahrens 			    argv[i]);
80399653d4eSeschrock 			goto error;
804fa9e4066Sahrens 		}
805fa9e4066Sahrens 
806fa9e4066Sahrens 		if ((dirp = opendir(argv[i])) == NULL) {
80799653d4eSeschrock 			zfs_error_aux(hdl, strerror(errno));
808ece3d9b3Slling 			(void) zfs_error_fmt(hdl, EZFS_BADPATH,
80999653d4eSeschrock 			    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
81099653d4eSeschrock 			    argv[i]);
81199653d4eSeschrock 			goto error;
812fa9e4066Sahrens 		}
813fa9e4066Sahrens 
814fa9e4066Sahrens 		/*
815fa9e4066Sahrens 		 * This is not MT-safe, but we have no MT consumers of libzfs
816fa9e4066Sahrens 		 */
817fa9e4066Sahrens 		while ((dp = readdir64(dirp)) != NULL) {
818fa9e4066Sahrens 
819fa9e4066Sahrens 			(void) snprintf(path, sizeof (path), "%s/%s",
820fa9e4066Sahrens 			    argv[i], dp->d_name);
821fa9e4066Sahrens 
822fa9e4066Sahrens 			if (stat64(path, &statbuf) != 0)
823fa9e4066Sahrens 				continue;
824fa9e4066Sahrens 
825fa9e4066Sahrens 			/*
826fa9e4066Sahrens 			 * Ignore directories (which includes "." and "..").
827fa9e4066Sahrens 			 */
828fa9e4066Sahrens 			if (S_ISDIR(statbuf.st_mode))
829fa9e4066Sahrens 				continue;
8303bb79becSeschrock 
8313bb79becSeschrock 			/*
8323bb79becSeschrock 			 * Ignore special (non-character or non-block) files.
8333bb79becSeschrock 			 */
8343bb79becSeschrock 			if (!S_ISREG(statbuf.st_mode) &&
8353bb79becSeschrock 			    !S_ISBLK(statbuf.st_mode))
8363bb79becSeschrock 				continue;
837fa9e4066Sahrens 
838fa9e4066Sahrens 			if ((fd = open64(path, O_RDONLY)) < 0)
839fa9e4066Sahrens 				continue;
840fa9e4066Sahrens 
84199653d4eSeschrock 			if ((zpool_read_label(fd, &config)) != 0) {
842ccae0b50Seschrock 				(void) close(fd);
84399653d4eSeschrock 				(void) no_memory(hdl);
84499653d4eSeschrock 				goto error;
84599653d4eSeschrock 			}
846fa9e4066Sahrens 
847fa9e4066Sahrens 			(void) close(fd);
848fa9e4066Sahrens 
849fa9e4066Sahrens 			if (config != NULL)
85099653d4eSeschrock 				if (add_config(hdl, &pools, path, config) != 0)
85199653d4eSeschrock 					goto error;
852fa9e4066Sahrens 		}
853ccae0b50Seschrock 
854ccae0b50Seschrock 		(void) closedir(dirp);
855ccae0b50Seschrock 		dirp = NULL;
856fa9e4066Sahrens 	}
857fa9e4066Sahrens 
85899653d4eSeschrock 	ret = get_configs(hdl, &pools);
85999653d4eSeschrock 
86099653d4eSeschrock error:
86199653d4eSeschrock 	for (pe = pools.pools; pe != NULL; pe = penext) {
86299653d4eSeschrock 		penext = pe->pe_next;
86399653d4eSeschrock 		for (ve = pe->pe_vdevs; ve != NULL; ve = venext) {
86499653d4eSeschrock 			venext = ve->ve_next;
86599653d4eSeschrock 			for (ce = ve->ve_configs; ce != NULL; ce = cenext) {
86699653d4eSeschrock 				cenext = ce->ce_next;
86799653d4eSeschrock 				if (ce->ce_config)
86899653d4eSeschrock 					nvlist_free(ce->ce_config);
86999653d4eSeschrock 				free(ce);
87099653d4eSeschrock 			}
87199653d4eSeschrock 			free(ve);
87299653d4eSeschrock 		}
87399653d4eSeschrock 		free(pe);
87499653d4eSeschrock 	}
87599653d4eSeschrock 
87699653d4eSeschrock 	for (ne = pools.names; ne != NULL; ne = nenext) {
87799653d4eSeschrock 		nenext = ne->ne_next;
87899653d4eSeschrock 		if (ne->ne_name)
87999653d4eSeschrock 			free(ne->ne_name);
88099653d4eSeschrock 		free(ne);
88199653d4eSeschrock 	}
88299653d4eSeschrock 
883ccae0b50Seschrock 	if (dirp)
884ccae0b50Seschrock 		(void) closedir(dirp);
885fa9e4066Sahrens 
886fa9e4066Sahrens 	return (ret);
887fa9e4066Sahrens }
888fa9e4066Sahrens 
8892f8aaab3Seschrock /*
8902f8aaab3Seschrock  * Given a cache file, return the contents as a list of importable pools.
8912f8aaab3Seschrock  */
8922f8aaab3Seschrock nvlist_t *
8932f8aaab3Seschrock zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile)
8942f8aaab3Seschrock {
8952f8aaab3Seschrock 	char *buf;
8962f8aaab3Seschrock 	int fd;
8972f8aaab3Seschrock 	struct stat64 statbuf;
8982f8aaab3Seschrock 	nvlist_t *raw, *src, *dst;
8992f8aaab3Seschrock 	nvlist_t *pools;
9002f8aaab3Seschrock 	nvpair_t *elem;
9012f8aaab3Seschrock 	char *name;
9022f8aaab3Seschrock 	uint64_t guid;
9032f8aaab3Seschrock 	boolean_t active;
9042f8aaab3Seschrock 
9052f8aaab3Seschrock 	if ((fd = open(cachefile, O_RDONLY)) < 0) {
9062f8aaab3Seschrock 		zfs_error_aux(hdl, "%s", strerror(errno));
9072f8aaab3Seschrock 		(void) zfs_error(hdl, EZFS_BADCACHE,
9082f8aaab3Seschrock 		    dgettext(TEXT_DOMAIN, "failed to open cache file"));
9092f8aaab3Seschrock 		return (NULL);
9102f8aaab3Seschrock 	}
9112f8aaab3Seschrock 
9122f8aaab3Seschrock 	if (fstat64(fd, &statbuf) != 0) {
9132f8aaab3Seschrock 		zfs_error_aux(hdl, "%s", strerror(errno));
9142f8aaab3Seschrock 		(void) close(fd);
9152f8aaab3Seschrock 		(void) zfs_error(hdl, EZFS_BADCACHE,
9162f8aaab3Seschrock 		    dgettext(TEXT_DOMAIN, "failed to get size of cache file"));
9172f8aaab3Seschrock 		return (NULL);
9182f8aaab3Seschrock 	}
9192f8aaab3Seschrock 
9202f8aaab3Seschrock 	if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) {
9212f8aaab3Seschrock 		(void) close(fd);
9222f8aaab3Seschrock 		return (NULL);
9232f8aaab3Seschrock 	}
9242f8aaab3Seschrock 
9252f8aaab3Seschrock 	if (read(fd, buf, statbuf.st_size) != statbuf.st_size) {
9262f8aaab3Seschrock 		(void) close(fd);
9272f8aaab3Seschrock 		free(buf);
9282f8aaab3Seschrock 		(void) zfs_error(hdl, EZFS_BADCACHE,
9292f8aaab3Seschrock 		    dgettext(TEXT_DOMAIN,
9302f8aaab3Seschrock 		    "failed to read cache file contents"));
9312f8aaab3Seschrock 		return (NULL);
9322f8aaab3Seschrock 	}
9332f8aaab3Seschrock 
9342f8aaab3Seschrock 	(void) close(fd);
9352f8aaab3Seschrock 
9362f8aaab3Seschrock 	if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) {
9372f8aaab3Seschrock 		free(buf);
9382f8aaab3Seschrock 		(void) zfs_error(hdl, EZFS_BADCACHE,
9392f8aaab3Seschrock 		    dgettext(TEXT_DOMAIN,
9402f8aaab3Seschrock 		    "invalid or corrupt cache file contents"));
9412f8aaab3Seschrock 		return (NULL);
9422f8aaab3Seschrock 	}
9432f8aaab3Seschrock 
9442f8aaab3Seschrock 	free(buf);
9452f8aaab3Seschrock 
9462f8aaab3Seschrock 	/*
9472f8aaab3Seschrock 	 * Go through and get the current state of the pools and refresh their
9482f8aaab3Seschrock 	 * state.
9492f8aaab3Seschrock 	 */
9502f8aaab3Seschrock 	if (nvlist_alloc(&pools, 0, 0) != 0) {
9512f8aaab3Seschrock 		(void) no_memory(hdl);
9522f8aaab3Seschrock 		nvlist_free(raw);
9532f8aaab3Seschrock 		return (NULL);
9542f8aaab3Seschrock 	}
9552f8aaab3Seschrock 
9562f8aaab3Seschrock 	elem = NULL;
9572f8aaab3Seschrock 	while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) {
9582f8aaab3Seschrock 		verify(nvpair_value_nvlist(elem, &src) == 0);
9592f8aaab3Seschrock 
9602f8aaab3Seschrock 		verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME,
9612f8aaab3Seschrock 		    &name) == 0);
9622f8aaab3Seschrock 		verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID,
9632f8aaab3Seschrock 		    &guid) == 0);
9642f8aaab3Seschrock 
9652f8aaab3Seschrock 		if (pool_active(hdl, name, guid, &active) != 0) {
9662f8aaab3Seschrock 			nvlist_free(raw);
9672f8aaab3Seschrock 			nvlist_free(pools);
9682f8aaab3Seschrock 			return (NULL);
9692f8aaab3Seschrock 		}
9702f8aaab3Seschrock 
9712f8aaab3Seschrock 		if (active)
9722f8aaab3Seschrock 			continue;
9732f8aaab3Seschrock 
9742f8aaab3Seschrock 		if ((dst = refresh_config(hdl, src)) == NULL) {
9752f8aaab3Seschrock 			nvlist_free(raw);
9762f8aaab3Seschrock 			nvlist_free(pools);
9772f8aaab3Seschrock 			return (NULL);
9782f8aaab3Seschrock 		}
9792f8aaab3Seschrock 
9802f8aaab3Seschrock 		if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) {
9812f8aaab3Seschrock 			(void) no_memory(hdl);
9822f8aaab3Seschrock 			nvlist_free(dst);
9832f8aaab3Seschrock 			nvlist_free(raw);
9842f8aaab3Seschrock 			nvlist_free(pools);
9852f8aaab3Seschrock 			return (NULL);
9862f8aaab3Seschrock 		}
9872f8aaab3Seschrock 
9882f8aaab3Seschrock 		nvlist_free(dst);
9892f8aaab3Seschrock 	}
9902f8aaab3Seschrock 
9912f8aaab3Seschrock 	nvlist_free(raw);
9922f8aaab3Seschrock 	return (pools);
9932f8aaab3Seschrock }
9942f8aaab3Seschrock 
9952f8aaab3Seschrock 
99699653d4eSeschrock boolean_t
997fa9e4066Sahrens find_guid(nvlist_t *nv, uint64_t guid)
998fa9e4066Sahrens {
999fa9e4066Sahrens 	uint64_t tmp;
1000fa9e4066Sahrens 	nvlist_t **child;
1001fa9e4066Sahrens 	uint_t c, children;
1002fa9e4066Sahrens 
1003fa9e4066Sahrens 	verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0);
1004fa9e4066Sahrens 	if (tmp == guid)
100599653d4eSeschrock 		return (B_TRUE);
1006fa9e4066Sahrens 
1007fa9e4066Sahrens 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1008fa9e4066Sahrens 	    &child, &children) == 0) {
1009fa9e4066Sahrens 		for (c = 0; c < children; c++)
1010fa9e4066Sahrens 			if (find_guid(child[c], guid))
101199653d4eSeschrock 				return (B_TRUE);
101299653d4eSeschrock 	}
101399653d4eSeschrock 
101499653d4eSeschrock 	return (B_FALSE);
101599653d4eSeschrock }
101699653d4eSeschrock 
1017fa94a07fSbrendan typedef struct aux_cbdata {
1018fa94a07fSbrendan 	const char	*cb_type;
101999653d4eSeschrock 	uint64_t	cb_guid;
102099653d4eSeschrock 	zpool_handle_t	*cb_zhp;
1021fa94a07fSbrendan } aux_cbdata_t;
102299653d4eSeschrock 
102399653d4eSeschrock static int
1024fa94a07fSbrendan find_aux(zpool_handle_t *zhp, void *data)
102599653d4eSeschrock {
1026fa94a07fSbrendan 	aux_cbdata_t *cbp = data;
1027fa94a07fSbrendan 	nvlist_t **list;
1028fa94a07fSbrendan 	uint_t i, count;
102999653d4eSeschrock 	uint64_t guid;
103099653d4eSeschrock 	nvlist_t *nvroot;
103199653d4eSeschrock 
103299653d4eSeschrock 	verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
103399653d4eSeschrock 	    &nvroot) == 0);
103499653d4eSeschrock 
1035fa94a07fSbrendan 	if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type,
1036fa94a07fSbrendan 	    &list, &count) == 0) {
1037fa94a07fSbrendan 		for (i = 0; i < count; i++) {
1038fa94a07fSbrendan 			verify(nvlist_lookup_uint64(list[i],
103999653d4eSeschrock 			    ZPOOL_CONFIG_GUID, &guid) == 0);
104099653d4eSeschrock 			if (guid == cbp->cb_guid) {
104199653d4eSeschrock 				cbp->cb_zhp = zhp;
104299653d4eSeschrock 				return (1);
104399653d4eSeschrock 			}
104499653d4eSeschrock 		}
1045fa9e4066Sahrens 	}
1046fa9e4066Sahrens 
104799653d4eSeschrock 	zpool_close(zhp);
104899653d4eSeschrock 	return (0);
1049fa9e4066Sahrens }
1050fa9e4066Sahrens 
1051fa9e4066Sahrens /*
105299653d4eSeschrock  * Determines if the pool is in use.  If so, it returns true and the state of
1053fa9e4066Sahrens  * the pool as well as the name of the pool.  Both strings are allocated and
1054fa9e4066Sahrens  * must be freed by the caller.
1055fa9e4066Sahrens  */
1056fa9e4066Sahrens int
105799653d4eSeschrock zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr,
105899653d4eSeschrock     boolean_t *inuse)
1059fa9e4066Sahrens {
1060fa9e4066Sahrens 	nvlist_t *config;
1061fa9e4066Sahrens 	char *name;
106299653d4eSeschrock 	boolean_t ret;
1063fa9e4066Sahrens 	uint64_t guid, vdev_guid;
1064fa9e4066Sahrens 	zpool_handle_t *zhp;
1065fa9e4066Sahrens 	nvlist_t *pool_config;
106639c23413Seschrock 	uint64_t stateval, isspare;
1067fa94a07fSbrendan 	aux_cbdata_t cb = { 0 };
106894de1d4cSeschrock 	boolean_t isactive;
106999653d4eSeschrock 
107099653d4eSeschrock 	*inuse = B_FALSE;
1071fa9e4066Sahrens 
107299653d4eSeschrock 	if (zpool_read_label(fd, &config) != 0) {
107399653d4eSeschrock 		(void) no_memory(hdl);
107499653d4eSeschrock 		return (-1);
107599653d4eSeschrock 	}
107699653d4eSeschrock 
107799653d4eSeschrock 	if (config == NULL)
107899653d4eSeschrock 		return (0);
1079fa9e4066Sahrens 
1080fa9e4066Sahrens 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE,
108146a2abf2Seschrock 	    &stateval) == 0);
1082fa9e4066Sahrens 	verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID,
1083fa9e4066Sahrens 	    &vdev_guid) == 0);
1084fa9e4066Sahrens 
1085fa94a07fSbrendan 	if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) {
108699653d4eSeschrock 		verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
108799653d4eSeschrock 		    &name) == 0);
108899653d4eSeschrock 		verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
108999653d4eSeschrock 		    &guid) == 0);
109099653d4eSeschrock 	}
109199653d4eSeschrock 
109246a2abf2Seschrock 	switch (stateval) {
1093fa9e4066Sahrens 	case POOL_STATE_EXPORTED:
109499653d4eSeschrock 		ret = B_TRUE;
1095fa9e4066Sahrens 		break;
1096fa9e4066Sahrens 
1097fa9e4066Sahrens 	case POOL_STATE_ACTIVE:
1098fa9e4066Sahrens 		/*
1099fa9e4066Sahrens 		 * For an active pool, we have to determine if it's really part
1100eaca9bbdSeschrock 		 * of a currently active pool (in which case the pool will exist
1101eaca9bbdSeschrock 		 * and the guid will be the same), or whether it's part of an
1102eaca9bbdSeschrock 		 * active pool that was disconnected without being explicitly
1103eaca9bbdSeschrock 		 * exported.
1104fa9e4066Sahrens 		 */
110594de1d4cSeschrock 		if (pool_active(hdl, name, guid, &isactive) != 0) {
110694de1d4cSeschrock 			nvlist_free(config);
110794de1d4cSeschrock 			return (-1);
110894de1d4cSeschrock 		}
110994de1d4cSeschrock 
111094de1d4cSeschrock 		if (isactive) {
1111fa9e4066Sahrens 			/*
1112fa9e4066Sahrens 			 * Because the device may have been removed while
1113fa9e4066Sahrens 			 * offlined, we only report it as active if the vdev is
1114fa9e4066Sahrens 			 * still present in the config.  Otherwise, pretend like
1115fa9e4066Sahrens 			 * it's not in use.
1116fa9e4066Sahrens 			 */
111799653d4eSeschrock 			if ((zhp = zpool_open_canfail(hdl, name)) != NULL &&
1118088e9d47Seschrock 			    (pool_config = zpool_get_config(zhp, NULL))
1119088e9d47Seschrock 			    != NULL) {
1120fa9e4066Sahrens 				nvlist_t *nvroot;
1121fa9e4066Sahrens 
1122fa9e4066Sahrens 				verify(nvlist_lookup_nvlist(pool_config,
1123fa9e4066Sahrens 				    ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
112446a2abf2Seschrock 				ret = find_guid(nvroot, vdev_guid);
1125fa9e4066Sahrens 			} else {
112699653d4eSeschrock 				ret = B_FALSE;
1127fa9e4066Sahrens 			}
112899653d4eSeschrock 
112939c23413Seschrock 			/*
113039c23413Seschrock 			 * If this is an active spare within another pool, we
113139c23413Seschrock 			 * treat it like an unused hot spare.  This allows the
113239c23413Seschrock 			 * user to create a pool with a hot spare that currently
113339c23413Seschrock 			 * in use within another pool.  Since we return B_TRUE,
113439c23413Seschrock 			 * libdiskmgt will continue to prevent generic consumers
113539c23413Seschrock 			 * from using the device.
113639c23413Seschrock 			 */
113739c23413Seschrock 			if (ret && nvlist_lookup_uint64(config,
113839c23413Seschrock 			    ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare)
113939c23413Seschrock 				stateval = POOL_STATE_SPARE;
114039c23413Seschrock 
114199653d4eSeschrock 			if (zhp != NULL)
114299653d4eSeschrock 				zpool_close(zhp);
1143fa9e4066Sahrens 		} else {
114446a2abf2Seschrock 			stateval = POOL_STATE_POTENTIALLY_ACTIVE;
114599653d4eSeschrock 			ret = B_TRUE;
114699653d4eSeschrock 		}
114799653d4eSeschrock 		break;
114899653d4eSeschrock 
114999653d4eSeschrock 	case POOL_STATE_SPARE:
115099653d4eSeschrock 		/*
115199653d4eSeschrock 		 * For a hot spare, it can be either definitively in use, or
115299653d4eSeschrock 		 * potentially active.  To determine if it's in use, we iterate
115399653d4eSeschrock 		 * over all pools in the system and search for one with a spare
115499653d4eSeschrock 		 * with a matching guid.
115599653d4eSeschrock 		 *
115699653d4eSeschrock 		 * Due to the shared nature of spares, we don't actually report
115799653d4eSeschrock 		 * the potentially active case as in use.  This means the user
115899653d4eSeschrock 		 * can freely create pools on the hot spares of exported pools,
115999653d4eSeschrock 		 * but to do otherwise makes the resulting code complicated, and
116099653d4eSeschrock 		 * we end up having to deal with this case anyway.
116199653d4eSeschrock 		 */
116299653d4eSeschrock 		cb.cb_zhp = NULL;
116399653d4eSeschrock 		cb.cb_guid = vdev_guid;
1164fa94a07fSbrendan 		cb.cb_type = ZPOOL_CONFIG_SPARES;
1165fa94a07fSbrendan 		if (zpool_iter(hdl, find_aux, &cb) == 1) {
1166fa94a07fSbrendan 			name = (char *)zpool_get_name(cb.cb_zhp);
1167fa94a07fSbrendan 			ret = TRUE;
1168fa94a07fSbrendan 		} else {
1169fa94a07fSbrendan 			ret = FALSE;
1170fa94a07fSbrendan 		}
1171fa94a07fSbrendan 		break;
1172fa94a07fSbrendan 
1173fa94a07fSbrendan 	case POOL_STATE_L2CACHE:
1174fa94a07fSbrendan 
1175fa94a07fSbrendan 		/*
1176fa94a07fSbrendan 		 * Check if any pool is currently using this l2cache device.
1177fa94a07fSbrendan 		 */
1178fa94a07fSbrendan 		cb.cb_zhp = NULL;
1179fa94a07fSbrendan 		cb.cb_guid = vdev_guid;
1180fa94a07fSbrendan 		cb.cb_type = ZPOOL_CONFIG_L2CACHE;
1181fa94a07fSbrendan 		if (zpool_iter(hdl, find_aux, &cb) == 1) {
118299653d4eSeschrock 			name = (char *)zpool_get_name(cb.cb_zhp);
1183fa9e4066Sahrens 			ret = TRUE;
118499653d4eSeschrock 		} else {
118599653d4eSeschrock 			ret = FALSE;
1186fa9e4066Sahrens 		}
1187fa9e4066Sahrens 		break;
1188fa9e4066Sahrens 
1189fa9e4066Sahrens 	default:
119099653d4eSeschrock 		ret = B_FALSE;
1191fa9e4066Sahrens 	}
1192fa9e4066Sahrens 
119346a2abf2Seschrock 
119446a2abf2Seschrock 	if (ret) {
119599653d4eSeschrock 		if ((*namestr = zfs_strdup(hdl, name)) == NULL) {
1196*c6765aabSeschrock 			if (cb.cb_zhp)
1197*c6765aabSeschrock 				zpool_close(cb.cb_zhp);
119899653d4eSeschrock 			nvlist_free(config);
119999653d4eSeschrock 			return (-1);
120099653d4eSeschrock 		}
120146a2abf2Seschrock 		*state = (pool_state_t)stateval;
120246a2abf2Seschrock 	}
120346a2abf2Seschrock 
120499653d4eSeschrock 	if (cb.cb_zhp)
120599653d4eSeschrock 		zpool_close(cb.cb_zhp);
120699653d4eSeschrock 
1207fa9e4066Sahrens 	nvlist_free(config);
120899653d4eSeschrock 	*inuse = ret;
120999653d4eSeschrock 	return (0);
1210fa9e4066Sahrens }
1211