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 /* 2288ecc943SGeorge Wilson * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens /* 27fa9e4066Sahrens * Pool import support functions. 28fa9e4066Sahrens * 29fa9e4066Sahrens * To import a pool, we rely on reading the configuration information from the 30fa9e4066Sahrens * ZFS label of each device. If we successfully read the label, then we 31fa9e4066Sahrens * organize the configuration information in the following hierarchy: 32fa9e4066Sahrens * 33fa9e4066Sahrens * pool guid -> toplevel vdev guid -> label txg 34fa9e4066Sahrens * 35fa9e4066Sahrens * Duplicate entries matching this same tuple will be discarded. Once we have 36fa9e4066Sahrens * examined every device, we pick the best label txg config for each toplevel 37fa9e4066Sahrens * vdev. We then arrange these toplevel vdevs into a complete pool config, and 38fa9e4066Sahrens * update any paths that have changed. Finally, we attempt to import the pool 39fa9e4066Sahrens * using our derived config, and record the results. 40fa9e4066Sahrens */ 41fa9e4066Sahrens 42*4f67d755SEric Taylor #include <ctype.h> 43fa9e4066Sahrens #include <devid.h> 44fa9e4066Sahrens #include <dirent.h> 45fa9e4066Sahrens #include <errno.h> 46fa9e4066Sahrens #include <libintl.h> 47*4f67d755SEric Taylor #include <stddef.h> 48fa9e4066Sahrens #include <stdlib.h> 49fa9e4066Sahrens #include <string.h> 50fa9e4066Sahrens #include <sys/stat.h> 51fa9e4066Sahrens #include <unistd.h> 52fa9e4066Sahrens #include <fcntl.h> 53*4f67d755SEric Taylor #include <sys/vtoc.h> 54*4f67d755SEric Taylor #include <sys/dktp/fdisk.h> 55*4f67d755SEric Taylor #include <sys/efi_partition.h> 56*4f67d755SEric Taylor #include <thread_pool.h> 57fa9e4066Sahrens 58fa9e4066Sahrens #include <sys/vdev_impl.h> 59fa9e4066Sahrens 60fa9e4066Sahrens #include "libzfs.h" 61fa9e4066Sahrens #include "libzfs_impl.h" 62fa9e4066Sahrens 63fa9e4066Sahrens /* 64fa9e4066Sahrens * Intermediate structures used to gather configuration information. 65fa9e4066Sahrens */ 66fa9e4066Sahrens typedef struct config_entry { 67fa9e4066Sahrens uint64_t ce_txg; 68fa9e4066Sahrens nvlist_t *ce_config; 69fa9e4066Sahrens struct config_entry *ce_next; 70fa9e4066Sahrens } config_entry_t; 71fa9e4066Sahrens 72fa9e4066Sahrens typedef struct vdev_entry { 73fa9e4066Sahrens uint64_t ve_guid; 74fa9e4066Sahrens config_entry_t *ve_configs; 75fa9e4066Sahrens struct vdev_entry *ve_next; 76fa9e4066Sahrens } vdev_entry_t; 77fa9e4066Sahrens 78fa9e4066Sahrens typedef struct pool_entry { 79fa9e4066Sahrens uint64_t pe_guid; 80fa9e4066Sahrens vdev_entry_t *pe_vdevs; 81fa9e4066Sahrens struct pool_entry *pe_next; 82fa9e4066Sahrens } pool_entry_t; 83fa9e4066Sahrens 84fa9e4066Sahrens typedef struct name_entry { 8599653d4eSeschrock char *ne_name; 86fa9e4066Sahrens uint64_t ne_guid; 87fa9e4066Sahrens struct name_entry *ne_next; 88fa9e4066Sahrens } name_entry_t; 89fa9e4066Sahrens 90fa9e4066Sahrens typedef struct pool_list { 91fa9e4066Sahrens pool_entry_t *pools; 92fa9e4066Sahrens name_entry_t *names; 93fa9e4066Sahrens } pool_list_t; 94fa9e4066Sahrens 95fa9e4066Sahrens static char * 96fa9e4066Sahrens get_devid(const char *path) 97fa9e4066Sahrens { 98fa9e4066Sahrens int fd; 99fa9e4066Sahrens ddi_devid_t devid; 100fa9e4066Sahrens char *minor, *ret; 101fa9e4066Sahrens 102fa9e4066Sahrens if ((fd = open(path, O_RDONLY)) < 0) 103fa9e4066Sahrens return (NULL); 104fa9e4066Sahrens 105fa9e4066Sahrens minor = NULL; 106fa9e4066Sahrens ret = NULL; 107fa9e4066Sahrens if (devid_get(fd, &devid) == 0) { 108fa9e4066Sahrens if (devid_get_minor_name(fd, &minor) == 0) 109fa9e4066Sahrens ret = devid_str_encode(devid, minor); 110fa9e4066Sahrens if (minor != NULL) 111fa9e4066Sahrens devid_str_free(minor); 112fa9e4066Sahrens devid_free(devid); 113fa9e4066Sahrens } 114c67d9675Seschrock (void) close(fd); 115fa9e4066Sahrens 116fa9e4066Sahrens return (ret); 117fa9e4066Sahrens } 118fa9e4066Sahrens 119fa9e4066Sahrens 120fa9e4066Sahrens /* 121fa9e4066Sahrens * Go through and fix up any path and/or devid information for the given vdev 122fa9e4066Sahrens * configuration. 123fa9e4066Sahrens */ 12499653d4eSeschrock static int 125fa9e4066Sahrens fix_paths(nvlist_t *nv, name_entry_t *names) 126fa9e4066Sahrens { 127fa9e4066Sahrens nvlist_t **child; 128fa9e4066Sahrens uint_t c, children; 129fa9e4066Sahrens uint64_t guid; 130c67d9675Seschrock name_entry_t *ne, *best; 131c67d9675Seschrock char *path, *devid; 132c67d9675Seschrock int matched; 133fa9e4066Sahrens 134fa9e4066Sahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 135fa9e4066Sahrens &child, &children) == 0) { 136fa9e4066Sahrens for (c = 0; c < children; c++) 13799653d4eSeschrock if (fix_paths(child[c], names) != 0) 13899653d4eSeschrock return (-1); 13999653d4eSeschrock return (0); 140fa9e4066Sahrens } 141fa9e4066Sahrens 142fa9e4066Sahrens /* 143fa9e4066Sahrens * This is a leaf (file or disk) vdev. In either case, go through 144fa9e4066Sahrens * the name list and see if we find a matching guid. If so, replace 145fa9e4066Sahrens * the path and see if we can calculate a new devid. 146c67d9675Seschrock * 147c67d9675Seschrock * There may be multiple names associated with a particular guid, in 148c67d9675Seschrock * which case we have overlapping slices or multiple paths to the same 149c67d9675Seschrock * disk. If this is the case, then we want to pick the path that is 150c67d9675Seschrock * the most similar to the original, where "most similar" is the number 151c67d9675Seschrock * of matching characters starting from the end of the path. This will 152c67d9675Seschrock * preserve slice numbers even if the disks have been reorganized, and 153c67d9675Seschrock * will also catch preferred disk names if multiple paths exist. 154fa9e4066Sahrens */ 155fa9e4066Sahrens verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); 156c67d9675Seschrock if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) 157c67d9675Seschrock path = NULL; 158c67d9675Seschrock 159c67d9675Seschrock matched = 0; 160c67d9675Seschrock best = NULL; 161c67d9675Seschrock for (ne = names; ne != NULL; ne = ne->ne_next) { 162c67d9675Seschrock if (ne->ne_guid == guid) { 163c67d9675Seschrock const char *src, *dst; 164c67d9675Seschrock int count; 165c67d9675Seschrock 166c67d9675Seschrock if (path == NULL) { 167c67d9675Seschrock best = ne; 168c67d9675Seschrock break; 169c67d9675Seschrock } 170c67d9675Seschrock 171c67d9675Seschrock src = ne->ne_name + strlen(ne->ne_name) - 1; 172c67d9675Seschrock dst = path + strlen(path) - 1; 173c67d9675Seschrock for (count = 0; src >= ne->ne_name && dst >= path; 174c67d9675Seschrock src--, dst--, count++) 175c67d9675Seschrock if (*src != *dst) 176c67d9675Seschrock break; 177c67d9675Seschrock 178c67d9675Seschrock /* 179c67d9675Seschrock * At this point, 'count' is the number of characters 180c67d9675Seschrock * matched from the end. 181c67d9675Seschrock */ 182c67d9675Seschrock if (count > matched || best == NULL) { 183c67d9675Seschrock best = ne; 184c67d9675Seschrock matched = count; 185c67d9675Seschrock } 186c67d9675Seschrock } 187c67d9675Seschrock } 188fa9e4066Sahrens 189c67d9675Seschrock if (best == NULL) 19099653d4eSeschrock return (0); 191fa9e4066Sahrens 19299653d4eSeschrock if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) 19399653d4eSeschrock return (-1); 194fa9e4066Sahrens 195c67d9675Seschrock if ((devid = get_devid(best->ne_name)) == NULL) { 196fa9e4066Sahrens (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); 197fa9e4066Sahrens } else { 19899653d4eSeschrock if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) 19999653d4eSeschrock return (-1); 200fa9e4066Sahrens devid_str_free(devid); 201fa9e4066Sahrens } 20299653d4eSeschrock 20399653d4eSeschrock return (0); 204fa9e4066Sahrens } 205fa9e4066Sahrens 206fa9e4066Sahrens /* 207fa9e4066Sahrens * Add the given configuration to the list of known devices. 208fa9e4066Sahrens */ 20999653d4eSeschrock static int 21099653d4eSeschrock add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, 21199653d4eSeschrock nvlist_t *config) 212fa9e4066Sahrens { 21399653d4eSeschrock uint64_t pool_guid, vdev_guid, top_guid, txg, state; 214fa9e4066Sahrens pool_entry_t *pe; 215fa9e4066Sahrens vdev_entry_t *ve; 216fa9e4066Sahrens config_entry_t *ce; 217fa9e4066Sahrens name_entry_t *ne; 218fa9e4066Sahrens 21999653d4eSeschrock /* 220fa94a07fSbrendan * If this is a hot spare not currently in use or level 2 cache 221fa94a07fSbrendan * device, add it to the list of names to translate, but don't do 222fa94a07fSbrendan * anything else. 22399653d4eSeschrock */ 22499653d4eSeschrock if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 225fa94a07fSbrendan &state) == 0 && 226fa94a07fSbrendan (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) && 22799653d4eSeschrock nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { 22899653d4eSeschrock if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 229ccae0b50Seschrock return (-1); 23099653d4eSeschrock 23199653d4eSeschrock if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 23299653d4eSeschrock free(ne); 23399653d4eSeschrock return (-1); 23499653d4eSeschrock } 23599653d4eSeschrock ne->ne_guid = vdev_guid; 23699653d4eSeschrock ne->ne_next = pl->names; 23799653d4eSeschrock pl->names = ne; 23899653d4eSeschrock return (0); 23999653d4eSeschrock } 24099653d4eSeschrock 241fa9e4066Sahrens /* 242fa9e4066Sahrens * If we have a valid config but cannot read any of these fields, then 243fa9e4066Sahrens * it means we have a half-initialized label. In vdev_label_init() 244fa9e4066Sahrens * we write a label with txg == 0 so that we can identify the device 245fa9e4066Sahrens * in case the user refers to the same disk later on. If we fail to 246fa9e4066Sahrens * create the pool, we'll be left with a label in this state 247fa9e4066Sahrens * which should not be considered part of a valid pool. 248fa9e4066Sahrens */ 249fa9e4066Sahrens if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 250fa9e4066Sahrens &pool_guid) != 0 || 251fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 252fa9e4066Sahrens &vdev_guid) != 0 || 253fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, 254fa9e4066Sahrens &top_guid) != 0 || 255fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 256fa9e4066Sahrens &txg) != 0 || txg == 0) { 257fa9e4066Sahrens nvlist_free(config); 25899653d4eSeschrock return (0); 259fa9e4066Sahrens } 260fa9e4066Sahrens 261fa9e4066Sahrens /* 262fa9e4066Sahrens * First, see if we know about this pool. If not, then add it to the 263fa9e4066Sahrens * list of known pools. 264fa9e4066Sahrens */ 265fa9e4066Sahrens for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 266fa9e4066Sahrens if (pe->pe_guid == pool_guid) 267fa9e4066Sahrens break; 268fa9e4066Sahrens } 269fa9e4066Sahrens 270fa9e4066Sahrens if (pe == NULL) { 27199653d4eSeschrock if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { 27299653d4eSeschrock nvlist_free(config); 27399653d4eSeschrock return (-1); 27499653d4eSeschrock } 275fa9e4066Sahrens pe->pe_guid = pool_guid; 276fa9e4066Sahrens pe->pe_next = pl->pools; 277fa9e4066Sahrens pl->pools = pe; 278fa9e4066Sahrens } 279fa9e4066Sahrens 280fa9e4066Sahrens /* 281fa9e4066Sahrens * Second, see if we know about this toplevel vdev. Add it if its 282fa9e4066Sahrens * missing. 283fa9e4066Sahrens */ 284fa9e4066Sahrens for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 285fa9e4066Sahrens if (ve->ve_guid == top_guid) 286fa9e4066Sahrens break; 287fa9e4066Sahrens } 288fa9e4066Sahrens 289fa9e4066Sahrens if (ve == NULL) { 29099653d4eSeschrock if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { 29199653d4eSeschrock nvlist_free(config); 29299653d4eSeschrock return (-1); 29399653d4eSeschrock } 294fa9e4066Sahrens ve->ve_guid = top_guid; 295fa9e4066Sahrens ve->ve_next = pe->pe_vdevs; 296fa9e4066Sahrens pe->pe_vdevs = ve; 297fa9e4066Sahrens } 298fa9e4066Sahrens 299fa9e4066Sahrens /* 300fa9e4066Sahrens * Third, see if we have a config with a matching transaction group. If 301fa9e4066Sahrens * so, then we do nothing. Otherwise, add it to the list of known 302fa9e4066Sahrens * configs. 303fa9e4066Sahrens */ 304fa9e4066Sahrens for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { 305fa9e4066Sahrens if (ce->ce_txg == txg) 306fa9e4066Sahrens break; 307fa9e4066Sahrens } 308fa9e4066Sahrens 309fa9e4066Sahrens if (ce == NULL) { 31099653d4eSeschrock if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { 31199653d4eSeschrock nvlist_free(config); 31299653d4eSeschrock return (-1); 31399653d4eSeschrock } 314fa9e4066Sahrens ce->ce_txg = txg; 315fa9e4066Sahrens ce->ce_config = config; 316fa9e4066Sahrens ce->ce_next = ve->ve_configs; 317fa9e4066Sahrens ve->ve_configs = ce; 318fa9e4066Sahrens } else { 319fa9e4066Sahrens nvlist_free(config); 320fa9e4066Sahrens } 321fa9e4066Sahrens 322fa9e4066Sahrens /* 323fa9e4066Sahrens * At this point we've successfully added our config to the list of 324fa9e4066Sahrens * known configs. The last thing to do is add the vdev guid -> path 325fa9e4066Sahrens * mappings so that we can fix up the configuration as necessary before 326fa9e4066Sahrens * doing the import. 327fa9e4066Sahrens */ 32899653d4eSeschrock if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 32999653d4eSeschrock return (-1); 33099653d4eSeschrock 33199653d4eSeschrock if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 33299653d4eSeschrock free(ne); 33399653d4eSeschrock return (-1); 33499653d4eSeschrock } 335fa9e4066Sahrens 336fa9e4066Sahrens ne->ne_guid = vdev_guid; 337fa9e4066Sahrens ne->ne_next = pl->names; 338fa9e4066Sahrens pl->names = ne; 33999653d4eSeschrock 34099653d4eSeschrock return (0); 341fa9e4066Sahrens } 342fa9e4066Sahrens 343eaca9bbdSeschrock /* 344eaca9bbdSeschrock * Returns true if the named pool matches the given GUID. 345eaca9bbdSeschrock */ 34694de1d4cSeschrock static int 34794de1d4cSeschrock pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, 34894de1d4cSeschrock boolean_t *isactive) 349eaca9bbdSeschrock { 350eaca9bbdSeschrock zpool_handle_t *zhp; 351eaca9bbdSeschrock uint64_t theguid; 352eaca9bbdSeschrock 35394de1d4cSeschrock if (zpool_open_silent(hdl, name, &zhp) != 0) 35494de1d4cSeschrock return (-1); 35594de1d4cSeschrock 35694de1d4cSeschrock if (zhp == NULL) { 35794de1d4cSeschrock *isactive = B_FALSE; 35894de1d4cSeschrock return (0); 35994de1d4cSeschrock } 360eaca9bbdSeschrock 361eaca9bbdSeschrock verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, 362eaca9bbdSeschrock &theguid) == 0); 363eaca9bbdSeschrock 364eaca9bbdSeschrock zpool_close(zhp); 365eaca9bbdSeschrock 36694de1d4cSeschrock *isactive = (theguid == guid); 36794de1d4cSeschrock return (0); 368eaca9bbdSeschrock } 369eaca9bbdSeschrock 3702f8aaab3Seschrock static nvlist_t * 3712f8aaab3Seschrock refresh_config(libzfs_handle_t *hdl, nvlist_t *config) 3722f8aaab3Seschrock { 3732f8aaab3Seschrock nvlist_t *nvl; 3742f8aaab3Seschrock zfs_cmd_t zc = { 0 }; 3752f8aaab3Seschrock int err; 3762f8aaab3Seschrock 3772f8aaab3Seschrock if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) 3782f8aaab3Seschrock return (NULL); 3792f8aaab3Seschrock 3802f8aaab3Seschrock if (zcmd_alloc_dst_nvlist(hdl, &zc, 3812f8aaab3Seschrock zc.zc_nvlist_conf_size * 2) != 0) { 3822f8aaab3Seschrock zcmd_free_nvlists(&zc); 3832f8aaab3Seschrock return (NULL); 3842f8aaab3Seschrock } 3852f8aaab3Seschrock 3862f8aaab3Seschrock while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, 3872f8aaab3Seschrock &zc)) != 0 && errno == ENOMEM) { 3882f8aaab3Seschrock if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3892f8aaab3Seschrock zcmd_free_nvlists(&zc); 3902f8aaab3Seschrock return (NULL); 3912f8aaab3Seschrock } 3922f8aaab3Seschrock } 3932f8aaab3Seschrock 3942f8aaab3Seschrock if (err) { 3952f8aaab3Seschrock zcmd_free_nvlists(&zc); 3962f8aaab3Seschrock return (NULL); 3972f8aaab3Seschrock } 3982f8aaab3Seschrock 3992f8aaab3Seschrock if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) { 4002f8aaab3Seschrock zcmd_free_nvlists(&zc); 4012f8aaab3Seschrock return (NULL); 4022f8aaab3Seschrock } 4032f8aaab3Seschrock 4042f8aaab3Seschrock zcmd_free_nvlists(&zc); 4052f8aaab3Seschrock return (nvl); 4062f8aaab3Seschrock } 4072f8aaab3Seschrock 40888ecc943SGeorge Wilson /* 40988ecc943SGeorge Wilson * Determine if the vdev id is a hole in the namespace. 41088ecc943SGeorge Wilson */ 41188ecc943SGeorge Wilson boolean_t 41288ecc943SGeorge Wilson vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id) 41388ecc943SGeorge Wilson { 41488ecc943SGeorge Wilson for (int c = 0; c < holes; c++) { 41588ecc943SGeorge Wilson 41688ecc943SGeorge Wilson /* Top-level is a hole */ 41788ecc943SGeorge Wilson if (hole_array[c] == id) 41888ecc943SGeorge Wilson return (B_TRUE); 41988ecc943SGeorge Wilson } 42088ecc943SGeorge Wilson return (B_FALSE); 42188ecc943SGeorge Wilson } 42288ecc943SGeorge Wilson 423fa9e4066Sahrens /* 424fa9e4066Sahrens * Convert our list of pools into the definitive set of configurations. We 425fa9e4066Sahrens * start by picking the best config for each toplevel vdev. Once that's done, 426fa9e4066Sahrens * we assemble the toplevel vdevs into a full config for the pool. We make a 427fa9e4066Sahrens * pass to fix up any incorrect paths, and then add it to the main list to 428fa9e4066Sahrens * return to the user. 429fa9e4066Sahrens */ 430fa9e4066Sahrens static nvlist_t * 4313a57275aSck get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok) 432fa9e4066Sahrens { 43399653d4eSeschrock pool_entry_t *pe; 43499653d4eSeschrock vdev_entry_t *ve; 43599653d4eSeschrock config_entry_t *ce; 43699653d4eSeschrock nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot; 437fa94a07fSbrendan nvlist_t **spares, **l2cache; 438fa94a07fSbrendan uint_t i, nspares, nl2cache; 43999653d4eSeschrock boolean_t config_seen; 440fa9e4066Sahrens uint64_t best_txg; 44195173954Sek char *name, *hostname; 44299653d4eSeschrock uint64_t version, guid; 44399653d4eSeschrock uint_t children = 0; 44499653d4eSeschrock nvlist_t **child = NULL; 44588ecc943SGeorge Wilson uint_t holes; 44688ecc943SGeorge Wilson uint64_t *hole_array, max_id; 44799653d4eSeschrock uint_t c; 44894de1d4cSeschrock boolean_t isactive; 44995173954Sek uint64_t hostid; 4502f8aaab3Seschrock nvlist_t *nvl; 45124e697d4Sck boolean_t found_one = B_FALSE; 45288ecc943SGeorge Wilson boolean_t valid_top_config = B_FALSE; 453fa9e4066Sahrens 45499653d4eSeschrock if (nvlist_alloc(&ret, 0, 0) != 0) 45599653d4eSeschrock goto nomem; 456fa9e4066Sahrens 45799653d4eSeschrock for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 45888ecc943SGeorge Wilson uint64_t id, max_txg = 0; 459fa9e4066Sahrens 46099653d4eSeschrock if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) 46199653d4eSeschrock goto nomem; 46299653d4eSeschrock config_seen = B_FALSE; 463fa9e4066Sahrens 464fa9e4066Sahrens /* 465fa9e4066Sahrens * Iterate over all toplevel vdevs. Grab the pool configuration 466fa9e4066Sahrens * from the first one we find, and then go through the rest and 467fa9e4066Sahrens * add them as necessary to the 'vdevs' member of the config. 468fa9e4066Sahrens */ 46999653d4eSeschrock for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 470fa9e4066Sahrens 471fa9e4066Sahrens /* 472fa9e4066Sahrens * Determine the best configuration for this vdev by 473fa9e4066Sahrens * selecting the config with the latest transaction 474fa9e4066Sahrens * group. 475fa9e4066Sahrens */ 476fa9e4066Sahrens best_txg = 0; 477fa9e4066Sahrens for (ce = ve->ve_configs; ce != NULL; 478fa9e4066Sahrens ce = ce->ce_next) { 479fa9e4066Sahrens 48099653d4eSeschrock if (ce->ce_txg > best_txg) { 481fa9e4066Sahrens tmp = ce->ce_config; 48299653d4eSeschrock best_txg = ce->ce_txg; 48399653d4eSeschrock } 484fa9e4066Sahrens } 485fa9e4066Sahrens 48688ecc943SGeorge Wilson /* 48788ecc943SGeorge Wilson * We rely on the fact that the max txg for the 48888ecc943SGeorge Wilson * pool will contain the most up-to-date information 48988ecc943SGeorge Wilson * about the valid top-levels in the vdev namespace. 49088ecc943SGeorge Wilson */ 49188ecc943SGeorge Wilson if (best_txg > max_txg) { 49288ecc943SGeorge Wilson (void) nvlist_remove(config, 49388ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, 49488ecc943SGeorge Wilson DATA_TYPE_UINT64); 49588ecc943SGeorge Wilson (void) nvlist_remove(config, 49688ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, 49788ecc943SGeorge Wilson DATA_TYPE_UINT64_ARRAY); 49888ecc943SGeorge Wilson 49988ecc943SGeorge Wilson max_txg = best_txg; 50088ecc943SGeorge Wilson hole_array = NULL; 50188ecc943SGeorge Wilson holes = 0; 50288ecc943SGeorge Wilson max_id = 0; 50388ecc943SGeorge Wilson valid_top_config = B_FALSE; 50488ecc943SGeorge Wilson 50588ecc943SGeorge Wilson if (nvlist_lookup_uint64(tmp, 50688ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) { 50788ecc943SGeorge Wilson verify(nvlist_add_uint64(config, 50888ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, 50988ecc943SGeorge Wilson max_id) == 0); 51088ecc943SGeorge Wilson valid_top_config = B_TRUE; 51188ecc943SGeorge Wilson } 51288ecc943SGeorge Wilson 51388ecc943SGeorge Wilson if (nvlist_lookup_uint64_array(tmp, 51488ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, &hole_array, 51588ecc943SGeorge Wilson &holes) == 0) { 51688ecc943SGeorge Wilson verify(nvlist_add_uint64_array(config, 51788ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, 51888ecc943SGeorge Wilson hole_array, holes) == 0); 51988ecc943SGeorge Wilson } 52088ecc943SGeorge Wilson } 52188ecc943SGeorge Wilson 522fa9e4066Sahrens if (!config_seen) { 523fa9e4066Sahrens /* 524fa9e4066Sahrens * Copy the relevant pieces of data to the pool 525fa9e4066Sahrens * configuration: 526fa9e4066Sahrens * 52799653d4eSeschrock * version 528fa9e4066Sahrens * pool guid 529fa9e4066Sahrens * name 530fa9e4066Sahrens * pool state 53195173954Sek * hostid (if available) 53295173954Sek * hostname (if available) 533fa9e4066Sahrens */ 534fa9e4066Sahrens uint64_t state; 535fa9e4066Sahrens 53699653d4eSeschrock verify(nvlist_lookup_uint64(tmp, 53799653d4eSeschrock ZPOOL_CONFIG_VERSION, &version) == 0); 53899653d4eSeschrock if (nvlist_add_uint64(config, 53999653d4eSeschrock ZPOOL_CONFIG_VERSION, version) != 0) 54099653d4eSeschrock goto nomem; 541fa9e4066Sahrens verify(nvlist_lookup_uint64(tmp, 542fa9e4066Sahrens ZPOOL_CONFIG_POOL_GUID, &guid) == 0); 54399653d4eSeschrock if (nvlist_add_uint64(config, 54499653d4eSeschrock ZPOOL_CONFIG_POOL_GUID, guid) != 0) 54599653d4eSeschrock goto nomem; 546fa9e4066Sahrens verify(nvlist_lookup_string(tmp, 547fa9e4066Sahrens ZPOOL_CONFIG_POOL_NAME, &name) == 0); 54899653d4eSeschrock if (nvlist_add_string(config, 54999653d4eSeschrock ZPOOL_CONFIG_POOL_NAME, name) != 0) 55099653d4eSeschrock goto nomem; 551fa9e4066Sahrens verify(nvlist_lookup_uint64(tmp, 552fa9e4066Sahrens ZPOOL_CONFIG_POOL_STATE, &state) == 0); 55399653d4eSeschrock if (nvlist_add_uint64(config, 55499653d4eSeschrock ZPOOL_CONFIG_POOL_STATE, state) != 0) 55599653d4eSeschrock goto nomem; 55695173954Sek hostid = 0; 55795173954Sek if (nvlist_lookup_uint64(tmp, 55895173954Sek ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 55995173954Sek if (nvlist_add_uint64(config, 56095173954Sek ZPOOL_CONFIG_HOSTID, hostid) != 0) 56195173954Sek goto nomem; 56295173954Sek verify(nvlist_lookup_string(tmp, 56395173954Sek ZPOOL_CONFIG_HOSTNAME, 56495173954Sek &hostname) == 0); 56595173954Sek if (nvlist_add_string(config, 56695173954Sek ZPOOL_CONFIG_HOSTNAME, 56795173954Sek hostname) != 0) 56895173954Sek goto nomem; 56995173954Sek } 570fa9e4066Sahrens 57199653d4eSeschrock config_seen = B_TRUE; 572fa9e4066Sahrens } 573fa9e4066Sahrens 574fa9e4066Sahrens /* 575fa9e4066Sahrens * Add this top-level vdev to the child array. 576fa9e4066Sahrens */ 577fa9e4066Sahrens verify(nvlist_lookup_nvlist(tmp, 578fa9e4066Sahrens ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); 579fa9e4066Sahrens verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, 580fa9e4066Sahrens &id) == 0); 58188ecc943SGeorge Wilson 582fa9e4066Sahrens if (id >= children) { 583fa9e4066Sahrens nvlist_t **newchild; 584fa9e4066Sahrens 58599653d4eSeschrock newchild = zfs_alloc(hdl, (id + 1) * 586fa9e4066Sahrens sizeof (nvlist_t *)); 58799653d4eSeschrock if (newchild == NULL) 58899653d4eSeschrock goto nomem; 589fa9e4066Sahrens 590fa9e4066Sahrens for (c = 0; c < children; c++) 591fa9e4066Sahrens newchild[c] = child[c]; 592fa9e4066Sahrens 593fa9e4066Sahrens free(child); 594fa9e4066Sahrens child = newchild; 595fa9e4066Sahrens children = id + 1; 596fa9e4066Sahrens } 59799653d4eSeschrock if (nvlist_dup(nvtop, &child[id], 0) != 0) 59899653d4eSeschrock goto nomem; 599fa9e4066Sahrens 600fa9e4066Sahrens } 601fa9e4066Sahrens 60288ecc943SGeorge Wilson /* 60388ecc943SGeorge Wilson * If we have information about all the top-levels then 60488ecc943SGeorge Wilson * clean up the nvlist which we've constructed. This 60588ecc943SGeorge Wilson * means removing any extraneous devices that are 60688ecc943SGeorge Wilson * beyond the valid range or adding devices to the end 60788ecc943SGeorge Wilson * of our array which appear to be missing. 60888ecc943SGeorge Wilson */ 60988ecc943SGeorge Wilson if (valid_top_config) { 61088ecc943SGeorge Wilson if (max_id < children) { 61188ecc943SGeorge Wilson for (c = max_id; c < children; c++) 61288ecc943SGeorge Wilson nvlist_free(child[c]); 61388ecc943SGeorge Wilson children = max_id; 61488ecc943SGeorge Wilson } else if (max_id > children) { 61588ecc943SGeorge Wilson nvlist_t **newchild; 61688ecc943SGeorge Wilson 61788ecc943SGeorge Wilson newchild = zfs_alloc(hdl, (max_id) * 61888ecc943SGeorge Wilson sizeof (nvlist_t *)); 61988ecc943SGeorge Wilson if (newchild == NULL) 62088ecc943SGeorge Wilson goto nomem; 62188ecc943SGeorge Wilson 62288ecc943SGeorge Wilson for (c = 0; c < children; c++) 62388ecc943SGeorge Wilson newchild[c] = child[c]; 62488ecc943SGeorge Wilson 62588ecc943SGeorge Wilson free(child); 62688ecc943SGeorge Wilson child = newchild; 62788ecc943SGeorge Wilson children = max_id; 62888ecc943SGeorge Wilson } 62988ecc943SGeorge Wilson } 63088ecc943SGeorge Wilson 631fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 632fa9e4066Sahrens &guid) == 0); 633fa9e4066Sahrens 63488ecc943SGeorge Wilson /* 63588ecc943SGeorge Wilson * The vdev namespace may contain holes as a result of 63688ecc943SGeorge Wilson * device removal. We must add them back into the vdev 63788ecc943SGeorge Wilson * tree before we process any missing devices. 63888ecc943SGeorge Wilson */ 63988ecc943SGeorge Wilson if (holes > 0) { 64088ecc943SGeorge Wilson ASSERT(valid_top_config); 64188ecc943SGeorge Wilson 64288ecc943SGeorge Wilson for (c = 0; c < children; c++) { 64388ecc943SGeorge Wilson nvlist_t *holey; 64488ecc943SGeorge Wilson 64588ecc943SGeorge Wilson if (child[c] != NULL || 64688ecc943SGeorge Wilson !vdev_is_hole(hole_array, holes, c)) 64788ecc943SGeorge Wilson continue; 64888ecc943SGeorge Wilson 64988ecc943SGeorge Wilson if (nvlist_alloc(&holey, NV_UNIQUE_NAME, 65088ecc943SGeorge Wilson 0) != 0) 65188ecc943SGeorge Wilson goto nomem; 65288ecc943SGeorge Wilson 65388ecc943SGeorge Wilson /* 65488ecc943SGeorge Wilson * Holes in the namespace are treated as 65588ecc943SGeorge Wilson * "hole" top-level vdevs and have a 65688ecc943SGeorge Wilson * special flag set on them. 65788ecc943SGeorge Wilson */ 65888ecc943SGeorge Wilson if (nvlist_add_string(holey, 65988ecc943SGeorge Wilson ZPOOL_CONFIG_TYPE, 66088ecc943SGeorge Wilson VDEV_TYPE_HOLE) != 0 || 66188ecc943SGeorge Wilson nvlist_add_uint64(holey, 66288ecc943SGeorge Wilson ZPOOL_CONFIG_ID, c) != 0 || 66388ecc943SGeorge Wilson nvlist_add_uint64(holey, 66488ecc943SGeorge Wilson ZPOOL_CONFIG_GUID, 0ULL) != 0) 66588ecc943SGeorge Wilson goto nomem; 66688ecc943SGeorge Wilson child[c] = holey; 66788ecc943SGeorge Wilson } 66888ecc943SGeorge Wilson } 66988ecc943SGeorge Wilson 670fa9e4066Sahrens /* 671fa9e4066Sahrens * Look for any missing top-level vdevs. If this is the case, 672fa9e4066Sahrens * create a faked up 'missing' vdev as a placeholder. We cannot 673fa9e4066Sahrens * simply compress the child array, because the kernel performs 674fa9e4066Sahrens * certain checks to make sure the vdev IDs match their location 675fa9e4066Sahrens * in the configuration. 676fa9e4066Sahrens */ 67788ecc943SGeorge Wilson for (c = 0; c < children; c++) { 678fa9e4066Sahrens if (child[c] == NULL) { 679fa9e4066Sahrens nvlist_t *missing; 68099653d4eSeschrock if (nvlist_alloc(&missing, NV_UNIQUE_NAME, 68199653d4eSeschrock 0) != 0) 68299653d4eSeschrock goto nomem; 68399653d4eSeschrock if (nvlist_add_string(missing, 68499653d4eSeschrock ZPOOL_CONFIG_TYPE, 68599653d4eSeschrock VDEV_TYPE_MISSING) != 0 || 68699653d4eSeschrock nvlist_add_uint64(missing, 68799653d4eSeschrock ZPOOL_CONFIG_ID, c) != 0 || 68899653d4eSeschrock nvlist_add_uint64(missing, 68999653d4eSeschrock ZPOOL_CONFIG_GUID, 0ULL) != 0) { 69099653d4eSeschrock nvlist_free(missing); 69199653d4eSeschrock goto nomem; 69299653d4eSeschrock } 693fa9e4066Sahrens child[c] = missing; 694fa9e4066Sahrens } 69588ecc943SGeorge Wilson } 696fa9e4066Sahrens 697fa9e4066Sahrens /* 698fa9e4066Sahrens * Put all of this pool's top-level vdevs into a root vdev. 699fa9e4066Sahrens */ 70099653d4eSeschrock if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) 70199653d4eSeschrock goto nomem; 70299653d4eSeschrock if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, 70399653d4eSeschrock VDEV_TYPE_ROOT) != 0 || 70499653d4eSeschrock nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || 70599653d4eSeschrock nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || 70699653d4eSeschrock nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 70799653d4eSeschrock child, children) != 0) { 70899653d4eSeschrock nvlist_free(nvroot); 70999653d4eSeschrock goto nomem; 71099653d4eSeschrock } 711fa9e4066Sahrens 712fa9e4066Sahrens for (c = 0; c < children; c++) 713fa9e4066Sahrens nvlist_free(child[c]); 714fa9e4066Sahrens free(child); 71599653d4eSeschrock children = 0; 71699653d4eSeschrock child = NULL; 717fa9e4066Sahrens 718fa9e4066Sahrens /* 719fa9e4066Sahrens * Go through and fix up any paths and/or devids based on our 720fa9e4066Sahrens * known list of vdev GUID -> path mappings. 721fa9e4066Sahrens */ 72299653d4eSeschrock if (fix_paths(nvroot, pl->names) != 0) { 72399653d4eSeschrock nvlist_free(nvroot); 72499653d4eSeschrock goto nomem; 72599653d4eSeschrock } 726fa9e4066Sahrens 727fa9e4066Sahrens /* 728fa9e4066Sahrens * Add the root vdev to this pool's configuration. 729fa9e4066Sahrens */ 73099653d4eSeschrock if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 73199653d4eSeschrock nvroot) != 0) { 73299653d4eSeschrock nvlist_free(nvroot); 73399653d4eSeschrock goto nomem; 73499653d4eSeschrock } 735fa9e4066Sahrens nvlist_free(nvroot); 736fa9e4066Sahrens 7373a57275aSck /* 7383a57275aSck * zdb uses this path to report on active pools that were 7393a57275aSck * imported or created using -R. 7403a57275aSck */ 7413a57275aSck if (active_ok) 7423a57275aSck goto add_pool; 7433a57275aSck 744fa9e4066Sahrens /* 745fa9e4066Sahrens * Determine if this pool is currently active, in which case we 746fa9e4066Sahrens * can't actually import it. 747fa9e4066Sahrens */ 748fa9e4066Sahrens verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 749fa9e4066Sahrens &name) == 0); 750fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 751fa9e4066Sahrens &guid) == 0); 752fa9e4066Sahrens 75394de1d4cSeschrock if (pool_active(hdl, name, guid, &isactive) != 0) 75494de1d4cSeschrock goto error; 75594de1d4cSeschrock 7560192a278Seschrock if (isactive) { 757fa9e4066Sahrens nvlist_free(config); 75899653d4eSeschrock config = NULL; 759fa9e4066Sahrens continue; 760fa9e4066Sahrens } 761fa9e4066Sahrens 76288ecc943SGeorge Wilson if ((nvl = refresh_config(hdl, config)) == NULL) { 76388ecc943SGeorge Wilson nvlist_free(config); 76488ecc943SGeorge Wilson config = NULL; 76588ecc943SGeorge Wilson continue; 76688ecc943SGeorge Wilson } 767fa9e4066Sahrens 768fa9e4066Sahrens nvlist_free(config); 7692f8aaab3Seschrock config = nvl; 770fa9e4066Sahrens 77199653d4eSeschrock /* 77299653d4eSeschrock * Go through and update the paths for spares, now that we have 77399653d4eSeschrock * them. 77499653d4eSeschrock */ 77599653d4eSeschrock verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 77699653d4eSeschrock &nvroot) == 0); 77799653d4eSeschrock if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 77899653d4eSeschrock &spares, &nspares) == 0) { 77999653d4eSeschrock for (i = 0; i < nspares; i++) { 78099653d4eSeschrock if (fix_paths(spares[i], pl->names) != 0) 78199653d4eSeschrock goto nomem; 78299653d4eSeschrock } 78399653d4eSeschrock } 78499653d4eSeschrock 785fa94a07fSbrendan /* 786fa94a07fSbrendan * Update the paths for l2cache devices. 787fa94a07fSbrendan */ 788fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 789fa94a07fSbrendan &l2cache, &nl2cache) == 0) { 790fa94a07fSbrendan for (i = 0; i < nl2cache; i++) { 791fa94a07fSbrendan if (fix_paths(l2cache[i], pl->names) != 0) 792fa94a07fSbrendan goto nomem; 793fa94a07fSbrendan } 794fa94a07fSbrendan } 795fa94a07fSbrendan 79695173954Sek /* 79795173954Sek * Restore the original information read from the actual label. 79895173954Sek */ 79995173954Sek (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, 80095173954Sek DATA_TYPE_UINT64); 80195173954Sek (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, 80295173954Sek DATA_TYPE_STRING); 80395173954Sek if (hostid != 0) { 80495173954Sek verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 80595173954Sek hostid) == 0); 80695173954Sek verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 80795173954Sek hostname) == 0); 80895173954Sek } 80995173954Sek 8103a57275aSck add_pool: 811fa9e4066Sahrens /* 812fa9e4066Sahrens * Add this pool to the list of configs. 813fa9e4066Sahrens */ 814e9dbad6fSeschrock verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 815e9dbad6fSeschrock &name) == 0); 81699653d4eSeschrock if (nvlist_add_nvlist(ret, name, config) != 0) 81799653d4eSeschrock goto nomem; 818fa9e4066Sahrens 81924e697d4Sck found_one = B_TRUE; 820fa9e4066Sahrens nvlist_free(config); 82199653d4eSeschrock config = NULL; 822fa9e4066Sahrens } 823fa9e4066Sahrens 82424e697d4Sck if (!found_one) { 82524e697d4Sck nvlist_free(ret); 82624e697d4Sck ret = NULL; 82724e697d4Sck } 82824e697d4Sck 829fa9e4066Sahrens return (ret); 83099653d4eSeschrock 83199653d4eSeschrock nomem: 83299653d4eSeschrock (void) no_memory(hdl); 83399653d4eSeschrock error: 83494de1d4cSeschrock nvlist_free(config); 83594de1d4cSeschrock nvlist_free(ret); 83699653d4eSeschrock for (c = 0; c < children; c++) 83799653d4eSeschrock nvlist_free(child[c]); 83894de1d4cSeschrock free(child); 83999653d4eSeschrock 84099653d4eSeschrock return (NULL); 841fa9e4066Sahrens } 842fa9e4066Sahrens 843fa9e4066Sahrens /* 844fa9e4066Sahrens * Return the offset of the given label. 845fa9e4066Sahrens */ 846fa9e4066Sahrens static uint64_t 847e7437265Sahrens label_offset(uint64_t size, int l) 848fa9e4066Sahrens { 849e7437265Sahrens ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0); 850fa9e4066Sahrens return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 851fa9e4066Sahrens 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); 852fa9e4066Sahrens } 853fa9e4066Sahrens 854fa9e4066Sahrens /* 855fa9e4066Sahrens * Given a file descriptor, read the label information and return an nvlist 856fa9e4066Sahrens * describing the configuration, if there is one. 857fa9e4066Sahrens */ 85899653d4eSeschrock int 85999653d4eSeschrock zpool_read_label(int fd, nvlist_t **config) 860fa9e4066Sahrens { 861fa9e4066Sahrens struct stat64 statbuf; 862fa9e4066Sahrens int l; 863fa9e4066Sahrens vdev_label_t *label; 864e7437265Sahrens uint64_t state, txg, size; 865fa9e4066Sahrens 86699653d4eSeschrock *config = NULL; 86799653d4eSeschrock 868fa9e4066Sahrens if (fstat64(fd, &statbuf) == -1) 86999653d4eSeschrock return (0); 870e7437265Sahrens size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 871fa9e4066Sahrens 87299653d4eSeschrock if ((label = malloc(sizeof (vdev_label_t))) == NULL) 87399653d4eSeschrock return (-1); 874fa9e4066Sahrens 875fa9e4066Sahrens for (l = 0; l < VDEV_LABELS; l++) { 876c5904d13Seschrock if (pread64(fd, label, sizeof (vdev_label_t), 877e7437265Sahrens label_offset(size, l)) != sizeof (vdev_label_t)) 878fa9e4066Sahrens continue; 879fa9e4066Sahrens 880fa9e4066Sahrens if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 88199653d4eSeschrock sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) 882fa9e4066Sahrens continue; 883fa9e4066Sahrens 88499653d4eSeschrock if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 885fa94a07fSbrendan &state) != 0 || state > POOL_STATE_L2CACHE) { 88699653d4eSeschrock nvlist_free(*config); 887fa9e4066Sahrens continue; 888fa9e4066Sahrens } 889fa9e4066Sahrens 890fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 89199653d4eSeschrock (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 89299653d4eSeschrock &txg) != 0 || txg == 0)) { 89399653d4eSeschrock nvlist_free(*config); 894fa9e4066Sahrens continue; 895fa9e4066Sahrens } 896fa9e4066Sahrens 897fa9e4066Sahrens free(label); 89899653d4eSeschrock return (0); 899fa9e4066Sahrens } 900fa9e4066Sahrens 901fa9e4066Sahrens free(label); 90299653d4eSeschrock *config = NULL; 90399653d4eSeschrock return (0); 904fa9e4066Sahrens } 905fa9e4066Sahrens 906*4f67d755SEric Taylor typedef struct rdsk_node { 907*4f67d755SEric Taylor char *rn_name; 908*4f67d755SEric Taylor int rn_dfd; 909*4f67d755SEric Taylor libzfs_handle_t *rn_hdl; 910*4f67d755SEric Taylor nvlist_t *rn_config; 911*4f67d755SEric Taylor avl_tree_t *rn_avl; 912*4f67d755SEric Taylor avl_node_t rn_node; 913*4f67d755SEric Taylor boolean_t rn_nozpool; 914*4f67d755SEric Taylor } rdsk_node_t; 915*4f67d755SEric Taylor 916*4f67d755SEric Taylor static int 917*4f67d755SEric Taylor slice_cache_compare(const void *arg1, const void *arg2) 918*4f67d755SEric Taylor { 919*4f67d755SEric Taylor const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; 920*4f67d755SEric Taylor const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; 921*4f67d755SEric Taylor char *nm1slice, *nm2slice; 922*4f67d755SEric Taylor int rv; 923*4f67d755SEric Taylor 924*4f67d755SEric Taylor /* 925*4f67d755SEric Taylor * slices zero and two are the most likely to provide results, 926*4f67d755SEric Taylor * so put those first 927*4f67d755SEric Taylor */ 928*4f67d755SEric Taylor nm1slice = strstr(nm1, "s0"); 929*4f67d755SEric Taylor nm2slice = strstr(nm2, "s0"); 930*4f67d755SEric Taylor if (nm1slice && !nm2slice) { 931*4f67d755SEric Taylor return (-1); 932*4f67d755SEric Taylor } 933*4f67d755SEric Taylor if (!nm1slice && nm2slice) { 934*4f67d755SEric Taylor return (1); 935*4f67d755SEric Taylor } 936*4f67d755SEric Taylor nm1slice = strstr(nm1, "s2"); 937*4f67d755SEric Taylor nm2slice = strstr(nm2, "s2"); 938*4f67d755SEric Taylor if (nm1slice && !nm2slice) { 939*4f67d755SEric Taylor return (-1); 940*4f67d755SEric Taylor } 941*4f67d755SEric Taylor if (!nm1slice && nm2slice) { 942*4f67d755SEric Taylor return (1); 943*4f67d755SEric Taylor } 944*4f67d755SEric Taylor 945*4f67d755SEric Taylor rv = strcmp(nm1, nm2); 946*4f67d755SEric Taylor if (rv == 0) 947*4f67d755SEric Taylor return (0); 948*4f67d755SEric Taylor return (rv > 0 ? 1 : -1); 949*4f67d755SEric Taylor } 950*4f67d755SEric Taylor 951*4f67d755SEric Taylor static void 952*4f67d755SEric Taylor check_one_slice(avl_tree_t *r, char *diskname, uint_t partno, 953*4f67d755SEric Taylor diskaddr_t size, uint_t blksz) 954*4f67d755SEric Taylor { 955*4f67d755SEric Taylor rdsk_node_t tmpnode; 956*4f67d755SEric Taylor rdsk_node_t *node; 957*4f67d755SEric Taylor char sname[MAXNAMELEN]; 958*4f67d755SEric Taylor 959*4f67d755SEric Taylor tmpnode.rn_name = &sname[0]; 960*4f67d755SEric Taylor (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u", 961*4f67d755SEric Taylor diskname, partno); 962*4f67d755SEric Taylor /* too small to contain a zpool? */ 963*4f67d755SEric Taylor if ((size < (SPA_MINDEVSIZE / blksz)) && 964*4f67d755SEric Taylor (node = avl_find(r, &tmpnode, NULL))) 965*4f67d755SEric Taylor node->rn_nozpool = B_TRUE; 966*4f67d755SEric Taylor } 967*4f67d755SEric Taylor 968*4f67d755SEric Taylor static void 969*4f67d755SEric Taylor nozpool_all_slices(avl_tree_t *r, const char *sname) 970*4f67d755SEric Taylor { 971*4f67d755SEric Taylor char diskname[MAXNAMELEN]; 972*4f67d755SEric Taylor char *ptr; 973*4f67d755SEric Taylor int i; 974*4f67d755SEric Taylor 975*4f67d755SEric Taylor (void) strncpy(diskname, sname, MAXNAMELEN); 976*4f67d755SEric Taylor if (((ptr = strrchr(diskname, 's')) == NULL) && 977*4f67d755SEric Taylor ((ptr = strrchr(diskname, 'p')) == NULL)) 978*4f67d755SEric Taylor return; 979*4f67d755SEric Taylor ptr[0] = 's'; 980*4f67d755SEric Taylor ptr[1] = '\0'; 981*4f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 982*4f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 983*4f67d755SEric Taylor ptr[0] = 'p'; 984*4f67d755SEric Taylor for (i = 0; i <= FD_NUMPART; i++) 985*4f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 986*4f67d755SEric Taylor } 987*4f67d755SEric Taylor 988*4f67d755SEric Taylor static void 989*4f67d755SEric Taylor check_slices(avl_tree_t *r, int fd, const char *sname) 990*4f67d755SEric Taylor { 991*4f67d755SEric Taylor struct extvtoc vtoc; 992*4f67d755SEric Taylor struct dk_gpt *gpt; 993*4f67d755SEric Taylor char diskname[MAXNAMELEN]; 994*4f67d755SEric Taylor char *ptr; 995*4f67d755SEric Taylor int i; 996*4f67d755SEric Taylor 997*4f67d755SEric Taylor (void) strncpy(diskname, sname, MAXNAMELEN); 998*4f67d755SEric Taylor if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1])) 999*4f67d755SEric Taylor return; 1000*4f67d755SEric Taylor ptr[1] = '\0'; 1001*4f67d755SEric Taylor 1002*4f67d755SEric Taylor if (read_extvtoc(fd, &vtoc) >= 0) { 1003*4f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 1004*4f67d755SEric Taylor check_one_slice(r, diskname, i, 1005*4f67d755SEric Taylor vtoc.v_part[i].p_size, vtoc.v_sectorsz); 1006*4f67d755SEric Taylor } else if (efi_alloc_and_read(fd, &gpt) >= 0) { 1007*4f67d755SEric Taylor /* 1008*4f67d755SEric Taylor * on x86 we'll still have leftover links that point 1009*4f67d755SEric Taylor * to slices s[9-15], so use NDKMAP instead 1010*4f67d755SEric Taylor */ 1011*4f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 1012*4f67d755SEric Taylor check_one_slice(r, diskname, i, 1013*4f67d755SEric Taylor gpt->efi_parts[i].p_size, gpt->efi_lbasize); 1014*4f67d755SEric Taylor /* nodes p[1-4] are never used with EFI labels */ 1015*4f67d755SEric Taylor ptr[0] = 'p'; 1016*4f67d755SEric Taylor for (i = 1; i <= FD_NUMPART; i++) 1017*4f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 1018*4f67d755SEric Taylor efi_free(gpt); 1019*4f67d755SEric Taylor } 1020*4f67d755SEric Taylor } 1021*4f67d755SEric Taylor 1022*4f67d755SEric Taylor static void 1023*4f67d755SEric Taylor zpool_open_func(void *arg) 1024*4f67d755SEric Taylor { 1025*4f67d755SEric Taylor rdsk_node_t *rn = arg; 1026*4f67d755SEric Taylor struct stat64 statbuf; 1027*4f67d755SEric Taylor nvlist_t *config; 1028*4f67d755SEric Taylor int fd; 1029*4f67d755SEric Taylor 1030*4f67d755SEric Taylor if (rn->rn_nozpool) 1031*4f67d755SEric Taylor return; 1032*4f67d755SEric Taylor if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { 1033*4f67d755SEric Taylor /* symlink to a device that's no longer there */ 1034*4f67d755SEric Taylor if (errno == ENOENT) 1035*4f67d755SEric Taylor nozpool_all_slices(rn->rn_avl, rn->rn_name); 1036*4f67d755SEric Taylor return; 1037*4f67d755SEric Taylor } 1038*4f67d755SEric Taylor /* 1039*4f67d755SEric Taylor * Ignore failed stats. We only want regular 1040*4f67d755SEric Taylor * files, character devs and block devs. 1041*4f67d755SEric Taylor */ 1042*4f67d755SEric Taylor if (fstat64(fd, &statbuf) != 0 || 1043*4f67d755SEric Taylor (!S_ISREG(statbuf.st_mode) && 1044*4f67d755SEric Taylor !S_ISCHR(statbuf.st_mode) && 1045*4f67d755SEric Taylor !S_ISBLK(statbuf.st_mode))) { 1046*4f67d755SEric Taylor (void) close(fd); 1047*4f67d755SEric Taylor return; 1048*4f67d755SEric Taylor } 1049*4f67d755SEric Taylor /* this file is too small to hold a zpool */ 1050*4f67d755SEric Taylor if (S_ISREG(statbuf.st_mode) && 1051*4f67d755SEric Taylor statbuf.st_size < SPA_MINDEVSIZE) { 1052*4f67d755SEric Taylor (void) close(fd); 1053*4f67d755SEric Taylor return; 1054*4f67d755SEric Taylor } else if (!S_ISREG(statbuf.st_mode)) { 1055*4f67d755SEric Taylor /* 1056*4f67d755SEric Taylor * Try to read the disk label first so we don't have to 1057*4f67d755SEric Taylor * open a bunch of minor nodes that can't have a zpool. 1058*4f67d755SEric Taylor */ 1059*4f67d755SEric Taylor check_slices(rn->rn_avl, fd, rn->rn_name); 1060*4f67d755SEric Taylor } 1061*4f67d755SEric Taylor 1062*4f67d755SEric Taylor if ((zpool_read_label(fd, &config)) != 0) { 1063*4f67d755SEric Taylor (void) close(fd); 1064*4f67d755SEric Taylor (void) no_memory(rn->rn_hdl); 1065*4f67d755SEric Taylor return; 1066*4f67d755SEric Taylor } 1067*4f67d755SEric Taylor (void) close(fd); 1068*4f67d755SEric Taylor 1069*4f67d755SEric Taylor 1070*4f67d755SEric Taylor rn->rn_config = config; 1071*4f67d755SEric Taylor if (config != NULL) { 1072*4f67d755SEric Taylor assert(rn->rn_nozpool == B_FALSE); 1073*4f67d755SEric Taylor } 1074*4f67d755SEric Taylor } 1075*4f67d755SEric Taylor 1076096d22d4SEric Schrock /* 1077096d22d4SEric Schrock * Given a file descriptor, clear (zero) the label information. This function 1078096d22d4SEric Schrock * is currently only used in the appliance stack as part of the ZFS sysevent 1079096d22d4SEric Schrock * module. 1080096d22d4SEric Schrock */ 1081096d22d4SEric Schrock int 1082096d22d4SEric Schrock zpool_clear_label(int fd) 1083096d22d4SEric Schrock { 1084096d22d4SEric Schrock struct stat64 statbuf; 1085096d22d4SEric Schrock int l; 1086096d22d4SEric Schrock vdev_label_t *label; 1087096d22d4SEric Schrock uint64_t size; 1088096d22d4SEric Schrock 1089096d22d4SEric Schrock if (fstat64(fd, &statbuf) == -1) 1090096d22d4SEric Schrock return (0); 1091096d22d4SEric Schrock size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 1092096d22d4SEric Schrock 1093096d22d4SEric Schrock if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) 1094096d22d4SEric Schrock return (-1); 1095096d22d4SEric Schrock 1096096d22d4SEric Schrock for (l = 0; l < VDEV_LABELS; l++) { 1097096d22d4SEric Schrock if (pwrite64(fd, label, sizeof (vdev_label_t), 1098096d22d4SEric Schrock label_offset(size, l)) != sizeof (vdev_label_t)) 1099096d22d4SEric Schrock return (-1); 1100096d22d4SEric Schrock } 1101096d22d4SEric Schrock 1102096d22d4SEric Schrock free(label); 1103096d22d4SEric Schrock return (0); 1104096d22d4SEric Schrock } 1105096d22d4SEric Schrock 1106fa9e4066Sahrens /* 1107fa9e4066Sahrens * Given a list of directories to search, find all pools stored on disk. This 1108fa9e4066Sahrens * includes partial pools which are not available to import. If no args are 1109fa9e4066Sahrens * given (argc is 0), then the default directory (/dev/dsk) is searched. 111024e697d4Sck * poolname or guid (but not both) are provided by the caller when trying 111124e697d4Sck * to import a specific pool. 1112fa9e4066Sahrens */ 111324e697d4Sck static nvlist_t * 111424e697d4Sck zpool_find_import_impl(libzfs_handle_t *hdl, int argc, char **argv, 111524e697d4Sck boolean_t active_ok, char *poolname, uint64_t guid) 1116fa9e4066Sahrens { 1117fa9e4066Sahrens int i; 1118ccae0b50Seschrock DIR *dirp = NULL; 1119fa9e4066Sahrens struct dirent64 *dp; 1120fa9e4066Sahrens char path[MAXPATHLEN]; 1121842c5645Sjwadams char *end; 1122842c5645Sjwadams size_t pathleft; 1123*4f67d755SEric Taylor nvlist_t *ret = NULL; 1124fa9e4066Sahrens static char *default_dir = "/dev/dsk"; 1125fa9e4066Sahrens pool_list_t pools = { 0 }; 112699653d4eSeschrock pool_entry_t *pe, *penext; 112799653d4eSeschrock vdev_entry_t *ve, *venext; 112899653d4eSeschrock config_entry_t *ce, *cenext; 112999653d4eSeschrock name_entry_t *ne, *nenext; 1130*4f67d755SEric Taylor avl_tree_t slice_cache; 1131*4f67d755SEric Taylor rdsk_node_t *slice; 1132*4f67d755SEric Taylor void *cookie; 113399653d4eSeschrock 113424e697d4Sck verify(poolname == NULL || guid == 0); 113524e697d4Sck 1136fa9e4066Sahrens if (argc == 0) { 1137fa9e4066Sahrens argc = 1; 1138fa9e4066Sahrens argv = &default_dir; 1139fa9e4066Sahrens } 1140fa9e4066Sahrens 1141fa9e4066Sahrens /* 1142fa9e4066Sahrens * Go through and read the label configuration information from every 1143fa9e4066Sahrens * possible device, organizing the information according to pool GUID 1144fa9e4066Sahrens * and toplevel GUID. 1145fa9e4066Sahrens */ 1146fa9e4066Sahrens for (i = 0; i < argc; i++) { 1147*4f67d755SEric Taylor tpool_t *t; 1148842c5645Sjwadams char *rdsk; 1149842c5645Sjwadams int dfd; 1150842c5645Sjwadams 1151842c5645Sjwadams /* use realpath to normalize the path */ 1152842c5645Sjwadams if (realpath(argv[i], path) == 0) { 1153ece3d9b3Slling (void) zfs_error_fmt(hdl, EZFS_BADPATH, 115499653d4eSeschrock dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1155fa9e4066Sahrens argv[i]); 115699653d4eSeschrock goto error; 1157fa9e4066Sahrens } 1158842c5645Sjwadams end = &path[strlen(path)]; 1159842c5645Sjwadams *end++ = '/'; 1160842c5645Sjwadams *end = 0; 1161842c5645Sjwadams pathleft = &path[sizeof (path)] - end; 1162842c5645Sjwadams 1163842c5645Sjwadams /* 1164842c5645Sjwadams * Using raw devices instead of block devices when we're 1165842c5645Sjwadams * reading the labels skips a bunch of slow operations during 1166842c5645Sjwadams * close(2) processing, so we replace /dev/dsk with /dev/rdsk. 1167842c5645Sjwadams */ 1168842c5645Sjwadams if (strcmp(path, "/dev/dsk/") == 0) 1169842c5645Sjwadams rdsk = "/dev/rdsk/"; 1170842c5645Sjwadams else 1171842c5645Sjwadams rdsk = path; 1172fa9e4066Sahrens 1173842c5645Sjwadams if ((dfd = open64(rdsk, O_RDONLY)) < 0 || 1174842c5645Sjwadams (dirp = fdopendir(dfd)) == NULL) { 117599653d4eSeschrock zfs_error_aux(hdl, strerror(errno)); 1176ece3d9b3Slling (void) zfs_error_fmt(hdl, EZFS_BADPATH, 117799653d4eSeschrock dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1178842c5645Sjwadams rdsk); 117999653d4eSeschrock goto error; 1180fa9e4066Sahrens } 1181fa9e4066Sahrens 1182*4f67d755SEric Taylor avl_create(&slice_cache, slice_cache_compare, 1183*4f67d755SEric Taylor sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node)); 1184fa9e4066Sahrens /* 1185fa9e4066Sahrens * This is not MT-safe, but we have no MT consumers of libzfs 1186fa9e4066Sahrens */ 1187fa9e4066Sahrens while ((dp = readdir64(dirp)) != NULL) { 1188842c5645Sjwadams const char *name = dp->d_name; 1189842c5645Sjwadams if (name[0] == '.' && 1190842c5645Sjwadams (name[1] == 0 || (name[1] == '.' && name[2] == 0))) 1191fa9e4066Sahrens continue; 1192fa9e4066Sahrens 1193*4f67d755SEric Taylor slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 1194*4f67d755SEric Taylor slice->rn_name = zfs_strdup(hdl, name); 1195*4f67d755SEric Taylor slice->rn_avl = &slice_cache; 1196*4f67d755SEric Taylor slice->rn_dfd = dfd; 1197*4f67d755SEric Taylor slice->rn_hdl = hdl; 1198*4f67d755SEric Taylor slice->rn_nozpool = B_FALSE; 1199*4f67d755SEric Taylor avl_add(&slice_cache, slice); 1200*4f67d755SEric Taylor } 1201*4f67d755SEric Taylor /* 1202*4f67d755SEric Taylor * create a thread pool to do all of this in parallel; 1203*4f67d755SEric Taylor * rn_nozpool is not protected, so this is racy in that 1204*4f67d755SEric Taylor * multiple tasks could decide that the same slice can 1205*4f67d755SEric Taylor * not hold a zpool, which is benign. Also choose 1206*4f67d755SEric Taylor * double the number of processors; we hold a lot of 1207*4f67d755SEric Taylor * locks in the kernel, so going beyond this doesn't 1208*4f67d755SEric Taylor * buy us much. 1209*4f67d755SEric Taylor */ 1210*4f67d755SEric Taylor t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN), 1211*4f67d755SEric Taylor 0, NULL); 1212*4f67d755SEric Taylor for (slice = avl_first(&slice_cache); slice; 1213*4f67d755SEric Taylor (slice = avl_walk(&slice_cache, slice, 1214*4f67d755SEric Taylor AVL_AFTER))) 1215*4f67d755SEric Taylor (void) tpool_dispatch(t, zpool_open_func, slice); 1216*4f67d755SEric Taylor tpool_wait(t); 1217*4f67d755SEric Taylor tpool_destroy(t); 1218*4f67d755SEric Taylor 1219*4f67d755SEric Taylor cookie = NULL; 1220*4f67d755SEric Taylor while ((slice = avl_destroy_nodes(&slice_cache, 1221*4f67d755SEric Taylor &cookie)) != NULL) { 1222*4f67d755SEric Taylor if (slice->rn_config != NULL) { 1223*4f67d755SEric Taylor nvlist_t *config = slice->rn_config; 122424e697d4Sck boolean_t matched = B_TRUE; 122524e697d4Sck 122624e697d4Sck if (poolname != NULL) { 122724e697d4Sck char *pname; 12287b0e68eaSck 12297b0e68eaSck matched = nvlist_lookup_string(config, 123024e697d4Sck ZPOOL_CONFIG_POOL_NAME, 12317b0e68eaSck &pname) == 0 && 12327b0e68eaSck strcmp(poolname, pname) == 0; 123324e697d4Sck } else if (guid != 0) { 123424e697d4Sck uint64_t this_guid; 1235913f2d80Sck 1236913f2d80Sck matched = nvlist_lookup_uint64(config, 123724e697d4Sck ZPOOL_CONFIG_POOL_GUID, 1238913f2d80Sck &this_guid) == 0 && 1239913f2d80Sck guid == this_guid; 124024e697d4Sck } 124124e697d4Sck if (!matched) { 124224e697d4Sck nvlist_free(config); 124324e697d4Sck config = NULL; 124424e697d4Sck continue; 124524e697d4Sck } 1246842c5645Sjwadams /* use the non-raw path for the config */ 1247*4f67d755SEric Taylor (void) strlcpy(end, slice->rn_name, pathleft); 124899653d4eSeschrock if (add_config(hdl, &pools, path, config) != 0) 124999653d4eSeschrock goto error; 1250842c5645Sjwadams } 1251*4f67d755SEric Taylor free(slice->rn_name); 1252*4f67d755SEric Taylor free(slice); 1253fa9e4066Sahrens } 1254*4f67d755SEric Taylor avl_destroy(&slice_cache); 1255ccae0b50Seschrock 1256ccae0b50Seschrock (void) closedir(dirp); 1257ccae0b50Seschrock dirp = NULL; 1258fa9e4066Sahrens } 1259fa9e4066Sahrens 12603a57275aSck ret = get_configs(hdl, &pools, active_ok); 126199653d4eSeschrock 126299653d4eSeschrock error: 126399653d4eSeschrock for (pe = pools.pools; pe != NULL; pe = penext) { 126499653d4eSeschrock penext = pe->pe_next; 126599653d4eSeschrock for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { 126699653d4eSeschrock venext = ve->ve_next; 126799653d4eSeschrock for (ce = ve->ve_configs; ce != NULL; ce = cenext) { 126899653d4eSeschrock cenext = ce->ce_next; 126999653d4eSeschrock if (ce->ce_config) 127099653d4eSeschrock nvlist_free(ce->ce_config); 127199653d4eSeschrock free(ce); 127299653d4eSeschrock } 127399653d4eSeschrock free(ve); 127499653d4eSeschrock } 127599653d4eSeschrock free(pe); 127699653d4eSeschrock } 127799653d4eSeschrock 127899653d4eSeschrock for (ne = pools.names; ne != NULL; ne = nenext) { 127999653d4eSeschrock nenext = ne->ne_next; 128099653d4eSeschrock if (ne->ne_name) 128199653d4eSeschrock free(ne->ne_name); 128299653d4eSeschrock free(ne); 128399653d4eSeschrock } 128499653d4eSeschrock 1285ccae0b50Seschrock if (dirp) 1286ccae0b50Seschrock (void) closedir(dirp); 1287fa9e4066Sahrens 1288fa9e4066Sahrens return (ret); 1289fa9e4066Sahrens } 1290fa9e4066Sahrens 129124e697d4Sck nvlist_t * 129224e697d4Sck zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) 129324e697d4Sck { 129424e697d4Sck return (zpool_find_import_impl(hdl, argc, argv, B_FALSE, NULL, 0)); 129524e697d4Sck } 129624e697d4Sck 129724e697d4Sck nvlist_t * 129824e697d4Sck zpool_find_import_byname(libzfs_handle_t *hdl, int argc, char **argv, 129924e697d4Sck char *pool) 130024e697d4Sck { 130124e697d4Sck return (zpool_find_import_impl(hdl, argc, argv, B_FALSE, pool, 0)); 130224e697d4Sck } 130324e697d4Sck 130424e697d4Sck nvlist_t * 130524e697d4Sck zpool_find_import_byguid(libzfs_handle_t *hdl, int argc, char **argv, 130624e697d4Sck uint64_t guid) 130724e697d4Sck { 130824e697d4Sck return (zpool_find_import_impl(hdl, argc, argv, B_FALSE, NULL, guid)); 130924e697d4Sck } 131024e697d4Sck 131124e697d4Sck nvlist_t * 131224e697d4Sck zpool_find_import_activeok(libzfs_handle_t *hdl, int argc, char **argv) 131324e697d4Sck { 131424e697d4Sck return (zpool_find_import_impl(hdl, argc, argv, B_TRUE, NULL, 0)); 131524e697d4Sck } 131624e697d4Sck 13172f8aaab3Seschrock /* 13182f8aaab3Seschrock * Given a cache file, return the contents as a list of importable pools. 131924e697d4Sck * poolname or guid (but not both) are provided by the caller when trying 132024e697d4Sck * to import a specific pool. 13212f8aaab3Seschrock */ 13222f8aaab3Seschrock nvlist_t * 13233a57275aSck zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, 1324e829d913Sck char *poolname, uint64_t guid) 13252f8aaab3Seschrock { 13262f8aaab3Seschrock char *buf; 13272f8aaab3Seschrock int fd; 13282f8aaab3Seschrock struct stat64 statbuf; 13292f8aaab3Seschrock nvlist_t *raw, *src, *dst; 13302f8aaab3Seschrock nvlist_t *pools; 13312f8aaab3Seschrock nvpair_t *elem; 13322f8aaab3Seschrock char *name; 133324e697d4Sck uint64_t this_guid; 13342f8aaab3Seschrock boolean_t active; 13352f8aaab3Seschrock 133624e697d4Sck verify(poolname == NULL || guid == 0); 133724e697d4Sck 13382f8aaab3Seschrock if ((fd = open(cachefile, O_RDONLY)) < 0) { 13392f8aaab3Seschrock zfs_error_aux(hdl, "%s", strerror(errno)); 13402f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13412f8aaab3Seschrock dgettext(TEXT_DOMAIN, "failed to open cache file")); 13422f8aaab3Seschrock return (NULL); 13432f8aaab3Seschrock } 13442f8aaab3Seschrock 13452f8aaab3Seschrock if (fstat64(fd, &statbuf) != 0) { 13462f8aaab3Seschrock zfs_error_aux(hdl, "%s", strerror(errno)); 13472f8aaab3Seschrock (void) close(fd); 13482f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13492f8aaab3Seschrock dgettext(TEXT_DOMAIN, "failed to get size of cache file")); 13502f8aaab3Seschrock return (NULL); 13512f8aaab3Seschrock } 13522f8aaab3Seschrock 13532f8aaab3Seschrock if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { 13542f8aaab3Seschrock (void) close(fd); 13552f8aaab3Seschrock return (NULL); 13562f8aaab3Seschrock } 13572f8aaab3Seschrock 13582f8aaab3Seschrock if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 13592f8aaab3Seschrock (void) close(fd); 13602f8aaab3Seschrock free(buf); 13612f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13622f8aaab3Seschrock dgettext(TEXT_DOMAIN, 13632f8aaab3Seschrock "failed to read cache file contents")); 13642f8aaab3Seschrock return (NULL); 13652f8aaab3Seschrock } 13662f8aaab3Seschrock 13672f8aaab3Seschrock (void) close(fd); 13682f8aaab3Seschrock 13692f8aaab3Seschrock if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { 13702f8aaab3Seschrock free(buf); 13712f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13722f8aaab3Seschrock dgettext(TEXT_DOMAIN, 13732f8aaab3Seschrock "invalid or corrupt cache file contents")); 13742f8aaab3Seschrock return (NULL); 13752f8aaab3Seschrock } 13762f8aaab3Seschrock 13772f8aaab3Seschrock free(buf); 13782f8aaab3Seschrock 13792f8aaab3Seschrock /* 13802f8aaab3Seschrock * Go through and get the current state of the pools and refresh their 13812f8aaab3Seschrock * state. 13822f8aaab3Seschrock */ 13832f8aaab3Seschrock if (nvlist_alloc(&pools, 0, 0) != 0) { 13842f8aaab3Seschrock (void) no_memory(hdl); 13852f8aaab3Seschrock nvlist_free(raw); 13862f8aaab3Seschrock return (NULL); 13872f8aaab3Seschrock } 13882f8aaab3Seschrock 13892f8aaab3Seschrock elem = NULL; 13902f8aaab3Seschrock while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { 13912f8aaab3Seschrock verify(nvpair_value_nvlist(elem, &src) == 0); 13922f8aaab3Seschrock 13932f8aaab3Seschrock verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME, 13942f8aaab3Seschrock &name) == 0); 139524e697d4Sck if (poolname != NULL && strcmp(poolname, name) != 0) 139624e697d4Sck continue; 139724e697d4Sck 13982f8aaab3Seschrock verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 139924e697d4Sck &this_guid) == 0); 140024e697d4Sck if (guid != 0) { 140124e697d4Sck verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 140224e697d4Sck &this_guid) == 0); 140324e697d4Sck if (guid != this_guid) 140424e697d4Sck continue; 140524e697d4Sck } 14062f8aaab3Seschrock 1407e829d913Sck if (pool_active(hdl, name, this_guid, &active) != 0) { 1408e829d913Sck nvlist_free(raw); 1409e829d913Sck nvlist_free(pools); 1410e829d913Sck return (NULL); 1411e829d913Sck } 14122f8aaab3Seschrock 1413e829d913Sck if (active) 1414e829d913Sck continue; 14152f8aaab3Seschrock 1416e829d913Sck if ((dst = refresh_config(hdl, src)) == NULL) { 1417e829d913Sck nvlist_free(raw); 1418e829d913Sck nvlist_free(pools); 1419e829d913Sck return (NULL); 1420e829d913Sck } 14212f8aaab3Seschrock 1422e829d913Sck if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { 1423e829d913Sck (void) no_memory(hdl); 14242f8aaab3Seschrock nvlist_free(dst); 1425e829d913Sck nvlist_free(raw); 1426e829d913Sck nvlist_free(pools); 1427e829d913Sck return (NULL); 14282f8aaab3Seschrock } 1429e829d913Sck nvlist_free(dst); 14302f8aaab3Seschrock } 14312f8aaab3Seschrock 14322f8aaab3Seschrock nvlist_free(raw); 14332f8aaab3Seschrock return (pools); 14342f8aaab3Seschrock } 14352f8aaab3Seschrock 14362f8aaab3Seschrock 143799653d4eSeschrock boolean_t 1438fa9e4066Sahrens find_guid(nvlist_t *nv, uint64_t guid) 1439fa9e4066Sahrens { 1440fa9e4066Sahrens uint64_t tmp; 1441fa9e4066Sahrens nvlist_t **child; 1442fa9e4066Sahrens uint_t c, children; 1443fa9e4066Sahrens 1444fa9e4066Sahrens verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); 1445fa9e4066Sahrens if (tmp == guid) 144699653d4eSeschrock return (B_TRUE); 1447fa9e4066Sahrens 1448fa9e4066Sahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1449fa9e4066Sahrens &child, &children) == 0) { 1450fa9e4066Sahrens for (c = 0; c < children; c++) 1451fa9e4066Sahrens if (find_guid(child[c], guid)) 145299653d4eSeschrock return (B_TRUE); 145399653d4eSeschrock } 145499653d4eSeschrock 145599653d4eSeschrock return (B_FALSE); 145699653d4eSeschrock } 145799653d4eSeschrock 1458fa94a07fSbrendan typedef struct aux_cbdata { 1459fa94a07fSbrendan const char *cb_type; 146099653d4eSeschrock uint64_t cb_guid; 146199653d4eSeschrock zpool_handle_t *cb_zhp; 1462fa94a07fSbrendan } aux_cbdata_t; 146399653d4eSeschrock 146499653d4eSeschrock static int 1465fa94a07fSbrendan find_aux(zpool_handle_t *zhp, void *data) 146699653d4eSeschrock { 1467fa94a07fSbrendan aux_cbdata_t *cbp = data; 1468fa94a07fSbrendan nvlist_t **list; 1469fa94a07fSbrendan uint_t i, count; 147099653d4eSeschrock uint64_t guid; 147199653d4eSeschrock nvlist_t *nvroot; 147299653d4eSeschrock 147399653d4eSeschrock verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 147499653d4eSeschrock &nvroot) == 0); 147599653d4eSeschrock 1476fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, 1477fa94a07fSbrendan &list, &count) == 0) { 1478fa94a07fSbrendan for (i = 0; i < count; i++) { 1479fa94a07fSbrendan verify(nvlist_lookup_uint64(list[i], 148099653d4eSeschrock ZPOOL_CONFIG_GUID, &guid) == 0); 148199653d4eSeschrock if (guid == cbp->cb_guid) { 148299653d4eSeschrock cbp->cb_zhp = zhp; 148399653d4eSeschrock return (1); 148499653d4eSeschrock } 148599653d4eSeschrock } 1486fa9e4066Sahrens } 1487fa9e4066Sahrens 148899653d4eSeschrock zpool_close(zhp); 148999653d4eSeschrock return (0); 1490fa9e4066Sahrens } 1491fa9e4066Sahrens 1492fa9e4066Sahrens /* 149399653d4eSeschrock * Determines if the pool is in use. If so, it returns true and the state of 1494fa9e4066Sahrens * the pool as well as the name of the pool. Both strings are allocated and 1495fa9e4066Sahrens * must be freed by the caller. 1496fa9e4066Sahrens */ 1497fa9e4066Sahrens int 149899653d4eSeschrock zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, 149999653d4eSeschrock boolean_t *inuse) 1500fa9e4066Sahrens { 1501fa9e4066Sahrens nvlist_t *config; 1502fa9e4066Sahrens char *name; 150399653d4eSeschrock boolean_t ret; 1504fa9e4066Sahrens uint64_t guid, vdev_guid; 1505fa9e4066Sahrens zpool_handle_t *zhp; 1506fa9e4066Sahrens nvlist_t *pool_config; 150739c23413Seschrock uint64_t stateval, isspare; 1508fa94a07fSbrendan aux_cbdata_t cb = { 0 }; 150994de1d4cSeschrock boolean_t isactive; 151099653d4eSeschrock 151199653d4eSeschrock *inuse = B_FALSE; 1512fa9e4066Sahrens 151399653d4eSeschrock if (zpool_read_label(fd, &config) != 0) { 151499653d4eSeschrock (void) no_memory(hdl); 151599653d4eSeschrock return (-1); 151699653d4eSeschrock } 151799653d4eSeschrock 151899653d4eSeschrock if (config == NULL) 151999653d4eSeschrock return (0); 1520fa9e4066Sahrens 1521fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 152246a2abf2Seschrock &stateval) == 0); 1523fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 1524fa9e4066Sahrens &vdev_guid) == 0); 1525fa9e4066Sahrens 1526fa94a07fSbrendan if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { 152799653d4eSeschrock verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 152899653d4eSeschrock &name) == 0); 152999653d4eSeschrock verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 153099653d4eSeschrock &guid) == 0); 153199653d4eSeschrock } 153299653d4eSeschrock 153346a2abf2Seschrock switch (stateval) { 1534fa9e4066Sahrens case POOL_STATE_EXPORTED: 153599653d4eSeschrock ret = B_TRUE; 1536fa9e4066Sahrens break; 1537fa9e4066Sahrens 1538fa9e4066Sahrens case POOL_STATE_ACTIVE: 1539fa9e4066Sahrens /* 1540fa9e4066Sahrens * For an active pool, we have to determine if it's really part 1541eaca9bbdSeschrock * of a currently active pool (in which case the pool will exist 1542eaca9bbdSeschrock * and the guid will be the same), or whether it's part of an 1543eaca9bbdSeschrock * active pool that was disconnected without being explicitly 1544eaca9bbdSeschrock * exported. 1545fa9e4066Sahrens */ 154694de1d4cSeschrock if (pool_active(hdl, name, guid, &isactive) != 0) { 154794de1d4cSeschrock nvlist_free(config); 154894de1d4cSeschrock return (-1); 154994de1d4cSeschrock } 155094de1d4cSeschrock 155194de1d4cSeschrock if (isactive) { 1552fa9e4066Sahrens /* 1553fa9e4066Sahrens * Because the device may have been removed while 1554fa9e4066Sahrens * offlined, we only report it as active if the vdev is 1555fa9e4066Sahrens * still present in the config. Otherwise, pretend like 1556fa9e4066Sahrens * it's not in use. 1557fa9e4066Sahrens */ 155899653d4eSeschrock if ((zhp = zpool_open_canfail(hdl, name)) != NULL && 1559088e9d47Seschrock (pool_config = zpool_get_config(zhp, NULL)) 1560088e9d47Seschrock != NULL) { 1561fa9e4066Sahrens nvlist_t *nvroot; 1562fa9e4066Sahrens 1563fa9e4066Sahrens verify(nvlist_lookup_nvlist(pool_config, 1564fa9e4066Sahrens ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 156546a2abf2Seschrock ret = find_guid(nvroot, vdev_guid); 1566fa9e4066Sahrens } else { 156799653d4eSeschrock ret = B_FALSE; 1568fa9e4066Sahrens } 156999653d4eSeschrock 157039c23413Seschrock /* 157139c23413Seschrock * If this is an active spare within another pool, we 157239c23413Seschrock * treat it like an unused hot spare. This allows the 157339c23413Seschrock * user to create a pool with a hot spare that currently 157439c23413Seschrock * in use within another pool. Since we return B_TRUE, 157539c23413Seschrock * libdiskmgt will continue to prevent generic consumers 157639c23413Seschrock * from using the device. 157739c23413Seschrock */ 157839c23413Seschrock if (ret && nvlist_lookup_uint64(config, 157939c23413Seschrock ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) 158039c23413Seschrock stateval = POOL_STATE_SPARE; 158139c23413Seschrock 158299653d4eSeschrock if (zhp != NULL) 158399653d4eSeschrock zpool_close(zhp); 1584fa9e4066Sahrens } else { 158546a2abf2Seschrock stateval = POOL_STATE_POTENTIALLY_ACTIVE; 158699653d4eSeschrock ret = B_TRUE; 158799653d4eSeschrock } 158899653d4eSeschrock break; 158999653d4eSeschrock 159099653d4eSeschrock case POOL_STATE_SPARE: 159199653d4eSeschrock /* 159299653d4eSeschrock * For a hot spare, it can be either definitively in use, or 159399653d4eSeschrock * potentially active. To determine if it's in use, we iterate 159499653d4eSeschrock * over all pools in the system and search for one with a spare 159599653d4eSeschrock * with a matching guid. 159699653d4eSeschrock * 159799653d4eSeschrock * Due to the shared nature of spares, we don't actually report 159899653d4eSeschrock * the potentially active case as in use. This means the user 159999653d4eSeschrock * can freely create pools on the hot spares of exported pools, 160099653d4eSeschrock * but to do otherwise makes the resulting code complicated, and 160199653d4eSeschrock * we end up having to deal with this case anyway. 160299653d4eSeschrock */ 160399653d4eSeschrock cb.cb_zhp = NULL; 160499653d4eSeschrock cb.cb_guid = vdev_guid; 1605fa94a07fSbrendan cb.cb_type = ZPOOL_CONFIG_SPARES; 1606fa94a07fSbrendan if (zpool_iter(hdl, find_aux, &cb) == 1) { 1607fa94a07fSbrendan name = (char *)zpool_get_name(cb.cb_zhp); 1608fa94a07fSbrendan ret = TRUE; 1609fa94a07fSbrendan } else { 1610fa94a07fSbrendan ret = FALSE; 1611fa94a07fSbrendan } 1612fa94a07fSbrendan break; 1613fa94a07fSbrendan 1614fa94a07fSbrendan case POOL_STATE_L2CACHE: 1615fa94a07fSbrendan 1616fa94a07fSbrendan /* 1617fa94a07fSbrendan * Check if any pool is currently using this l2cache device. 1618fa94a07fSbrendan */ 1619fa94a07fSbrendan cb.cb_zhp = NULL; 1620fa94a07fSbrendan cb.cb_guid = vdev_guid; 1621fa94a07fSbrendan cb.cb_type = ZPOOL_CONFIG_L2CACHE; 1622fa94a07fSbrendan if (zpool_iter(hdl, find_aux, &cb) == 1) { 162399653d4eSeschrock name = (char *)zpool_get_name(cb.cb_zhp); 1624fa9e4066Sahrens ret = TRUE; 162599653d4eSeschrock } else { 162699653d4eSeschrock ret = FALSE; 1627fa9e4066Sahrens } 1628fa9e4066Sahrens break; 1629fa9e4066Sahrens 1630fa9e4066Sahrens default: 163199653d4eSeschrock ret = B_FALSE; 1632fa9e4066Sahrens } 1633fa9e4066Sahrens 163446a2abf2Seschrock 163546a2abf2Seschrock if (ret) { 163699653d4eSeschrock if ((*namestr = zfs_strdup(hdl, name)) == NULL) { 1637c6765aabSeschrock if (cb.cb_zhp) 1638c6765aabSeschrock zpool_close(cb.cb_zhp); 163999653d4eSeschrock nvlist_free(config); 164099653d4eSeschrock return (-1); 164199653d4eSeschrock } 164246a2abf2Seschrock *state = (pool_state_t)stateval; 164346a2abf2Seschrock } 164446a2abf2Seschrock 164599653d4eSeschrock if (cb.cb_zhp) 164699653d4eSeschrock zpool_close(cb.cb_zhp); 164799653d4eSeschrock 1648fa9e4066Sahrens nvlist_free(config); 164999653d4eSeschrock *inuse = ret; 165099653d4eSeschrock return (0); 1651fa9e4066Sahrens } 1652