1168404Spjd/* 2168404Spjd * CDDL HEADER START 3168404Spjd * 4168404Spjd * The contents of this file are subject to the terms of the 5168404Spjd * Common Development and Distribution License (the "License"). 6168404Spjd * You may not use this file except in compliance with the License. 7168404Spjd * 8168404Spjd * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9168404Spjd * or http://www.opensolaris.org/os/licensing. 10168404Spjd * See the License for the specific language governing permissions 11168404Spjd * and limitations under the License. 12168404Spjd * 13168404Spjd * When distributing Covered Code, include this CDDL HEADER in each 14168404Spjd * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15168404Spjd * If applicable, add the following below this CDDL HEADER, with the 16168404Spjd * fields enclosed by brackets "[]" replaced with your own identifying 17168404Spjd * information: Portions Copyright [yyyy] [name of copyright owner] 18168404Spjd * 19168404Spjd * CDDL HEADER END 20168404Spjd */ 21168404Spjd/* 22219089Spjd * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23228103Smm * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 24239620Smm * Copyright (c) 2012 by Delphix. All rights reserved. 25168404Spjd */ 26168404Spjd 27168404Spjd/* 28168404Spjd * Pool import support functions. 29168404Spjd * 30168404Spjd * To import a pool, we rely on reading the configuration information from the 31168404Spjd * ZFS label of each device. If we successfully read the label, then we 32168404Spjd * organize the configuration information in the following hierarchy: 33168404Spjd * 34168404Spjd * pool guid -> toplevel vdev guid -> label txg 35168404Spjd * 36168404Spjd * Duplicate entries matching this same tuple will be discarded. Once we have 37168404Spjd * examined every device, we pick the best label txg config for each toplevel 38168404Spjd * vdev. We then arrange these toplevel vdevs into a complete pool config, and 39168404Spjd * update any paths that have changed. Finally, we attempt to import the pool 40168404Spjd * using our derived config, and record the results. 41168404Spjd */ 42168404Spjd 43219089Spjd#include <ctype.h> 44168404Spjd#include <devid.h> 45168404Spjd#include <dirent.h> 46168404Spjd#include <errno.h> 47168404Spjd#include <libintl.h> 48219089Spjd#include <stddef.h> 49168404Spjd#include <stdlib.h> 50168404Spjd#include <string.h> 51168404Spjd#include <sys/stat.h> 52168404Spjd#include <unistd.h> 53168404Spjd#include <fcntl.h> 54219089Spjd#include <thread_pool.h> 55168404Spjd#include <libgeom.h> 56168404Spjd 57168404Spjd#include <sys/vdev_impl.h> 58168404Spjd 59168404Spjd#include "libzfs.h" 60168404Spjd#include "libzfs_impl.h" 61168404Spjd 62168404Spjd/* 63168404Spjd * Intermediate structures used to gather configuration information. 64168404Spjd */ 65168404Spjdtypedef struct config_entry { 66168404Spjd uint64_t ce_txg; 67168404Spjd nvlist_t *ce_config; 68168404Spjd struct config_entry *ce_next; 69168404Spjd} config_entry_t; 70168404Spjd 71168404Spjdtypedef struct vdev_entry { 72168404Spjd uint64_t ve_guid; 73168404Spjd config_entry_t *ve_configs; 74168404Spjd struct vdev_entry *ve_next; 75168404Spjd} vdev_entry_t; 76168404Spjd 77168404Spjdtypedef struct pool_entry { 78168404Spjd uint64_t pe_guid; 79168404Spjd vdev_entry_t *pe_vdevs; 80168404Spjd struct pool_entry *pe_next; 81168404Spjd} pool_entry_t; 82168404Spjd 83168404Spjdtypedef struct name_entry { 84168404Spjd char *ne_name; 85168404Spjd uint64_t ne_guid; 86168404Spjd struct name_entry *ne_next; 87168404Spjd} name_entry_t; 88168404Spjd 89168404Spjdtypedef struct pool_list { 90168404Spjd pool_entry_t *pools; 91168404Spjd name_entry_t *names; 92168404Spjd} pool_list_t; 93168404Spjd 94168404Spjdstatic char * 95168404Spjdget_devid(const char *path) 96168404Spjd{ 97168404Spjd int fd; 98168404Spjd ddi_devid_t devid; 99168404Spjd char *minor, *ret; 100168404Spjd 101168404Spjd if ((fd = open(path, O_RDONLY)) < 0) 102168404Spjd return (NULL); 103168404Spjd 104168404Spjd minor = NULL; 105168404Spjd ret = NULL; 106168404Spjd if (devid_get(fd, &devid) == 0) { 107168404Spjd if (devid_get_minor_name(fd, &minor) == 0) 108168404Spjd ret = devid_str_encode(devid, minor); 109168404Spjd if (minor != NULL) 110168404Spjd devid_str_free(minor); 111168404Spjd devid_free(devid); 112168404Spjd } 113168404Spjd (void) close(fd); 114168404Spjd 115168404Spjd return (ret); 116168404Spjd} 117168404Spjd 118219089Spjd 119168404Spjd/* 120168404Spjd * Go through and fix up any path and/or devid information for the given vdev 121168404Spjd * configuration. 122168404Spjd */ 123168404Spjdstatic int 124168404Spjdfix_paths(nvlist_t *nv, name_entry_t *names) 125168404Spjd{ 126168404Spjd nvlist_t **child; 127168404Spjd uint_t c, children; 128168404Spjd uint64_t guid; 129168404Spjd name_entry_t *ne, *best; 130168404Spjd char *path, *devid; 131168404Spjd int matched; 132168404Spjd 133168404Spjd if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 134168404Spjd &child, &children) == 0) { 135168404Spjd for (c = 0; c < children; c++) 136168404Spjd if (fix_paths(child[c], names) != 0) 137168404Spjd return (-1); 138168404Spjd return (0); 139168404Spjd } 140168404Spjd 141168404Spjd /* 142168404Spjd * This is a leaf (file or disk) vdev. In either case, go through 143168404Spjd * the name list and see if we find a matching guid. If so, replace 144168404Spjd * the path and see if we can calculate a new devid. 145168404Spjd * 146168404Spjd * There may be multiple names associated with a particular guid, in 147168404Spjd * which case we have overlapping slices or multiple paths to the same 148168404Spjd * disk. If this is the case, then we want to pick the path that is 149168404Spjd * the most similar to the original, where "most similar" is the number 150168404Spjd * of matching characters starting from the end of the path. This will 151168404Spjd * preserve slice numbers even if the disks have been reorganized, and 152168404Spjd * will also catch preferred disk names if multiple paths exist. 153168404Spjd */ 154168404Spjd verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); 155168404Spjd if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) 156168404Spjd path = NULL; 157168404Spjd 158168404Spjd matched = 0; 159168404Spjd best = NULL; 160168404Spjd for (ne = names; ne != NULL; ne = ne->ne_next) { 161168404Spjd if (ne->ne_guid == guid) { 162168404Spjd const char *src, *dst; 163168404Spjd int count; 164168404Spjd 165168404Spjd if (path == NULL) { 166168404Spjd best = ne; 167168404Spjd break; 168168404Spjd } 169168404Spjd 170168404Spjd src = ne->ne_name + strlen(ne->ne_name) - 1; 171168404Spjd dst = path + strlen(path) - 1; 172168404Spjd for (count = 0; src >= ne->ne_name && dst >= path; 173168404Spjd src--, dst--, count++) 174168404Spjd if (*src != *dst) 175168404Spjd break; 176168404Spjd 177168404Spjd /* 178168404Spjd * At this point, 'count' is the number of characters 179168404Spjd * matched from the end. 180168404Spjd */ 181168404Spjd if (count > matched || best == NULL) { 182168404Spjd best = ne; 183168404Spjd matched = count; 184168404Spjd } 185168404Spjd } 186168404Spjd } 187168404Spjd 188168404Spjd if (best == NULL) 189168404Spjd return (0); 190168404Spjd 191168404Spjd if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) 192168404Spjd return (-1); 193168404Spjd 194168404Spjd if ((devid = get_devid(best->ne_name)) == NULL) { 195168404Spjd (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); 196168404Spjd } else { 197168404Spjd if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) 198168404Spjd return (-1); 199168404Spjd devid_str_free(devid); 200168404Spjd } 201168404Spjd 202168404Spjd return (0); 203168404Spjd} 204168404Spjd 205168404Spjd/* 206168404Spjd * Add the given configuration to the list of known devices. 207168404Spjd */ 208168404Spjdstatic int 209168404Spjdadd_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, 210168404Spjd nvlist_t *config) 211168404Spjd{ 212168404Spjd uint64_t pool_guid, vdev_guid, top_guid, txg, state; 213168404Spjd pool_entry_t *pe; 214168404Spjd vdev_entry_t *ve; 215168404Spjd config_entry_t *ce; 216168404Spjd name_entry_t *ne; 217168404Spjd 218168404Spjd /* 219185029Spjd * If this is a hot spare not currently in use or level 2 cache 220185029Spjd * device, add it to the list of names to translate, but don't do 221185029Spjd * anything else. 222168404Spjd */ 223168404Spjd if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 224185029Spjd &state) == 0 && 225185029Spjd (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) && 226168404Spjd nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { 227168404Spjd if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 228168926Spjd return (-1); 229168404Spjd 230168404Spjd if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 231168404Spjd free(ne); 232168404Spjd return (-1); 233168404Spjd } 234168404Spjd ne->ne_guid = vdev_guid; 235168404Spjd ne->ne_next = pl->names; 236168404Spjd pl->names = ne; 237168404Spjd return (0); 238168404Spjd } 239168404Spjd 240168404Spjd /* 241168404Spjd * If we have a valid config but cannot read any of these fields, then 242168404Spjd * it means we have a half-initialized label. In vdev_label_init() 243168404Spjd * we write a label with txg == 0 so that we can identify the device 244168404Spjd * in case the user refers to the same disk later on. If we fail to 245168404Spjd * create the pool, we'll be left with a label in this state 246168404Spjd * which should not be considered part of a valid pool. 247168404Spjd */ 248168404Spjd if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 249168404Spjd &pool_guid) != 0 || 250168404Spjd nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 251168404Spjd &vdev_guid) != 0 || 252168404Spjd nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, 253168404Spjd &top_guid) != 0 || 254168404Spjd nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 255168404Spjd &txg) != 0 || txg == 0) { 256168404Spjd nvlist_free(config); 257168404Spjd return (0); 258168404Spjd } 259168404Spjd 260168404Spjd /* 261168404Spjd * First, see if we know about this pool. If not, then add it to the 262168404Spjd * list of known pools. 263168404Spjd */ 264168404Spjd for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 265168404Spjd if (pe->pe_guid == pool_guid) 266168404Spjd break; 267168404Spjd } 268168404Spjd 269168404Spjd if (pe == NULL) { 270168404Spjd if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { 271168404Spjd nvlist_free(config); 272168404Spjd return (-1); 273168404Spjd } 274168404Spjd pe->pe_guid = pool_guid; 275168404Spjd pe->pe_next = pl->pools; 276168404Spjd pl->pools = pe; 277168404Spjd } 278168404Spjd 279168404Spjd /* 280168404Spjd * Second, see if we know about this toplevel vdev. Add it if its 281168404Spjd * missing. 282168404Spjd */ 283168404Spjd for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 284168404Spjd if (ve->ve_guid == top_guid) 285168404Spjd break; 286168404Spjd } 287168404Spjd 288168404Spjd if (ve == NULL) { 289168404Spjd if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { 290168404Spjd nvlist_free(config); 291168404Spjd return (-1); 292168404Spjd } 293168404Spjd ve->ve_guid = top_guid; 294168404Spjd ve->ve_next = pe->pe_vdevs; 295168404Spjd pe->pe_vdevs = ve; 296168404Spjd } 297168404Spjd 298168404Spjd /* 299168404Spjd * Third, see if we have a config with a matching transaction group. If 300168404Spjd * so, then we do nothing. Otherwise, add it to the list of known 301168404Spjd * configs. 302168404Spjd */ 303168404Spjd for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { 304168404Spjd if (ce->ce_txg == txg) 305168404Spjd break; 306168404Spjd } 307168404Spjd 308168404Spjd if (ce == NULL) { 309168404Spjd if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { 310168404Spjd nvlist_free(config); 311168404Spjd return (-1); 312168404Spjd } 313168404Spjd ce->ce_txg = txg; 314168404Spjd ce->ce_config = config; 315168404Spjd ce->ce_next = ve->ve_configs; 316168404Spjd ve->ve_configs = ce; 317168404Spjd } else { 318168404Spjd nvlist_free(config); 319168404Spjd } 320168404Spjd 321168404Spjd /* 322168404Spjd * At this point we've successfully added our config to the list of 323168404Spjd * known configs. The last thing to do is add the vdev guid -> path 324168404Spjd * mappings so that we can fix up the configuration as necessary before 325168404Spjd * doing the import. 326168404Spjd */ 327168404Spjd if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 328168404Spjd return (-1); 329168404Spjd 330168404Spjd if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 331168404Spjd free(ne); 332168404Spjd return (-1); 333168404Spjd } 334168404Spjd 335168404Spjd ne->ne_guid = vdev_guid; 336168404Spjd ne->ne_next = pl->names; 337168404Spjd pl->names = ne; 338168404Spjd 339168404Spjd return (0); 340168404Spjd} 341168404Spjd 342168404Spjd/* 343168404Spjd * Returns true if the named pool matches the given GUID. 344168404Spjd */ 345168404Spjdstatic int 346168404Spjdpool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, 347168404Spjd boolean_t *isactive) 348168404Spjd{ 349168404Spjd zpool_handle_t *zhp; 350168404Spjd uint64_t theguid; 351168404Spjd 352168404Spjd if (zpool_open_silent(hdl, name, &zhp) != 0) 353168404Spjd return (-1); 354168404Spjd 355168404Spjd if (zhp == NULL) { 356168404Spjd *isactive = B_FALSE; 357168404Spjd return (0); 358168404Spjd } 359168404Spjd 360168404Spjd verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, 361168404Spjd &theguid) == 0); 362168404Spjd 363168404Spjd zpool_close(zhp); 364168404Spjd 365168404Spjd *isactive = (theguid == guid); 366168404Spjd return (0); 367168404Spjd} 368168404Spjd 369185029Spjdstatic nvlist_t * 370185029Spjdrefresh_config(libzfs_handle_t *hdl, nvlist_t *config) 371185029Spjd{ 372185029Spjd nvlist_t *nvl; 373185029Spjd zfs_cmd_t zc = { 0 }; 374185029Spjd int err; 375185029Spjd 376185029Spjd if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) 377185029Spjd return (NULL); 378185029Spjd 379185029Spjd if (zcmd_alloc_dst_nvlist(hdl, &zc, 380185029Spjd zc.zc_nvlist_conf_size * 2) != 0) { 381185029Spjd zcmd_free_nvlists(&zc); 382185029Spjd return (NULL); 383185029Spjd } 384185029Spjd 385185029Spjd while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, 386185029Spjd &zc)) != 0 && errno == ENOMEM) { 387185029Spjd if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 388185029Spjd zcmd_free_nvlists(&zc); 389185029Spjd return (NULL); 390185029Spjd } 391185029Spjd } 392185029Spjd 393185029Spjd if (err) { 394185029Spjd zcmd_free_nvlists(&zc); 395185029Spjd return (NULL); 396185029Spjd } 397185029Spjd 398185029Spjd if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) { 399185029Spjd zcmd_free_nvlists(&zc); 400185029Spjd return (NULL); 401185029Spjd } 402185029Spjd 403185029Spjd zcmd_free_nvlists(&zc); 404185029Spjd return (nvl); 405185029Spjd} 406185029Spjd 407168404Spjd/* 408219089Spjd * Determine if the vdev id is a hole in the namespace. 409219089Spjd */ 410219089Spjdboolean_t 411219089Spjdvdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id) 412219089Spjd{ 413219089Spjd for (int c = 0; c < holes; c++) { 414219089Spjd 415219089Spjd /* Top-level is a hole */ 416219089Spjd if (hole_array[c] == id) 417219089Spjd return (B_TRUE); 418219089Spjd } 419219089Spjd return (B_FALSE); 420219089Spjd} 421219089Spjd 422219089Spjd/* 423168404Spjd * Convert our list of pools into the definitive set of configurations. We 424168404Spjd * start by picking the best config for each toplevel vdev. Once that's done, 425168404Spjd * we assemble the toplevel vdevs into a full config for the pool. We make a 426168404Spjd * pass to fix up any incorrect paths, and then add it to the main list to 427168404Spjd * return to the user. 428168404Spjd */ 429168404Spjdstatic nvlist_t * 430185029Spjdget_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok) 431168404Spjd{ 432168404Spjd pool_entry_t *pe; 433168404Spjd vdev_entry_t *ve; 434168404Spjd config_entry_t *ce; 435168404Spjd nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot; 436185029Spjd nvlist_t **spares, **l2cache; 437185029Spjd uint_t i, nspares, nl2cache; 438168404Spjd boolean_t config_seen; 439168404Spjd uint64_t best_txg; 440239620Smm char *name, *hostname; 441239620Smm uint64_t guid; 442168404Spjd uint_t children = 0; 443168404Spjd nvlist_t **child = NULL; 444219089Spjd uint_t holes; 445219089Spjd uint64_t *hole_array, max_id; 446168404Spjd uint_t c; 447168404Spjd boolean_t isactive; 448168498Spjd uint64_t hostid; 449185029Spjd nvlist_t *nvl; 450185029Spjd boolean_t found_one = B_FALSE; 451219089Spjd boolean_t valid_top_config = B_FALSE; 452168404Spjd 453168404Spjd if (nvlist_alloc(&ret, 0, 0) != 0) 454168404Spjd goto nomem; 455168404Spjd 456168404Spjd for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 457219089Spjd uint64_t id, max_txg = 0; 458168404Spjd 459168404Spjd if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) 460168404Spjd goto nomem; 461168404Spjd config_seen = B_FALSE; 462168404Spjd 463168404Spjd /* 464168404Spjd * Iterate over all toplevel vdevs. Grab the pool configuration 465168404Spjd * from the first one we find, and then go through the rest and 466168404Spjd * add them as necessary to the 'vdevs' member of the config. 467168404Spjd */ 468168404Spjd for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 469168404Spjd 470168404Spjd /* 471168404Spjd * Determine the best configuration for this vdev by 472168404Spjd * selecting the config with the latest transaction 473168404Spjd * group. 474168404Spjd */ 475168404Spjd best_txg = 0; 476168404Spjd for (ce = ve->ve_configs; ce != NULL; 477168404Spjd ce = ce->ce_next) { 478168404Spjd 479168404Spjd if (ce->ce_txg > best_txg) { 480168404Spjd tmp = ce->ce_config; 481168404Spjd best_txg = ce->ce_txg; 482168404Spjd } 483168404Spjd } 484168404Spjd 485219089Spjd /* 486219089Spjd * We rely on the fact that the max txg for the 487219089Spjd * pool will contain the most up-to-date information 488219089Spjd * about the valid top-levels in the vdev namespace. 489219089Spjd */ 490219089Spjd if (best_txg > max_txg) { 491219089Spjd (void) nvlist_remove(config, 492219089Spjd ZPOOL_CONFIG_VDEV_CHILDREN, 493219089Spjd DATA_TYPE_UINT64); 494219089Spjd (void) nvlist_remove(config, 495219089Spjd ZPOOL_CONFIG_HOLE_ARRAY, 496219089Spjd DATA_TYPE_UINT64_ARRAY); 497219089Spjd 498219089Spjd max_txg = best_txg; 499219089Spjd hole_array = NULL; 500219089Spjd holes = 0; 501219089Spjd max_id = 0; 502219089Spjd valid_top_config = B_FALSE; 503219089Spjd 504219089Spjd if (nvlist_lookup_uint64(tmp, 505219089Spjd ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) { 506219089Spjd verify(nvlist_add_uint64(config, 507219089Spjd ZPOOL_CONFIG_VDEV_CHILDREN, 508219089Spjd max_id) == 0); 509219089Spjd valid_top_config = B_TRUE; 510219089Spjd } 511219089Spjd 512219089Spjd if (nvlist_lookup_uint64_array(tmp, 513219089Spjd ZPOOL_CONFIG_HOLE_ARRAY, &hole_array, 514219089Spjd &holes) == 0) { 515219089Spjd verify(nvlist_add_uint64_array(config, 516219089Spjd ZPOOL_CONFIG_HOLE_ARRAY, 517219089Spjd hole_array, holes) == 0); 518219089Spjd } 519219089Spjd } 520219089Spjd 521168404Spjd if (!config_seen) { 522168404Spjd /* 523168404Spjd * Copy the relevant pieces of data to the pool 524168404Spjd * configuration: 525168404Spjd * 526168404Spjd * version 527239620Smm * pool guid 528239620Smm * name 529228103Smm * comment (if available) 530239620Smm * pool state 531168498Spjd * hostid (if available) 532168498Spjd * hostname (if available) 533168404Spjd */ 534246631Smm uint64_t state, version; 535239620Smm char *comment = NULL; 536168404Spjd 537239620Smm version = fnvlist_lookup_uint64(tmp, 538239620Smm ZPOOL_CONFIG_VERSION); 539239620Smm fnvlist_add_uint64(config, 540239620Smm ZPOOL_CONFIG_VERSION, version); 541239620Smm guid = fnvlist_lookup_uint64(tmp, 542239620Smm ZPOOL_CONFIG_POOL_GUID); 543239620Smm fnvlist_add_uint64(config, 544239620Smm ZPOOL_CONFIG_POOL_GUID, guid); 545239620Smm name = fnvlist_lookup_string(tmp, 546239620Smm ZPOOL_CONFIG_POOL_NAME); 547239620Smm fnvlist_add_string(config, 548239620Smm ZPOOL_CONFIG_POOL_NAME, name); 549228103Smm 550228103Smm if (nvlist_lookup_string(tmp, 551239620Smm ZPOOL_CONFIG_COMMENT, &comment) == 0) 552239620Smm fnvlist_add_string(config, 553239620Smm ZPOOL_CONFIG_COMMENT, comment); 554228103Smm 555239620Smm state = fnvlist_lookup_uint64(tmp, 556239620Smm ZPOOL_CONFIG_POOL_STATE); 557239620Smm fnvlist_add_uint64(config, 558239620Smm ZPOOL_CONFIG_POOL_STATE, state); 559228103Smm 560168498Spjd hostid = 0; 561168498Spjd if (nvlist_lookup_uint64(tmp, 562168498Spjd ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 563239620Smm fnvlist_add_uint64(config, 564239620Smm ZPOOL_CONFIG_HOSTID, hostid); 565239620Smm hostname = fnvlist_lookup_string(tmp, 566239620Smm ZPOOL_CONFIG_HOSTNAME); 567239620Smm fnvlist_add_string(config, 568239620Smm ZPOOL_CONFIG_HOSTNAME, hostname); 569168498Spjd } 570168404Spjd 571168404Spjd config_seen = B_TRUE; 572168404Spjd } 573168404Spjd 574168404Spjd /* 575168404Spjd * Add this top-level vdev to the child array. 576168404Spjd */ 577168404Spjd verify(nvlist_lookup_nvlist(tmp, 578168404Spjd ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); 579168404Spjd verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, 580168404Spjd &id) == 0); 581219089Spjd 582168404Spjd if (id >= children) { 583168404Spjd nvlist_t **newchild; 584168404Spjd 585168404Spjd newchild = zfs_alloc(hdl, (id + 1) * 586168404Spjd sizeof (nvlist_t *)); 587168404Spjd if (newchild == NULL) 588168404Spjd goto nomem; 589168404Spjd 590168404Spjd for (c = 0; c < children; c++) 591168404Spjd newchild[c] = child[c]; 592168404Spjd 593168404Spjd free(child); 594168404Spjd child = newchild; 595168404Spjd children = id + 1; 596168404Spjd } 597168404Spjd if (nvlist_dup(nvtop, &child[id], 0) != 0) 598168404Spjd goto nomem; 599168404Spjd 600168404Spjd } 601168404Spjd 602219089Spjd /* 603219089Spjd * If we have information about all the top-levels then 604219089Spjd * clean up the nvlist which we've constructed. This 605219089Spjd * means removing any extraneous devices that are 606219089Spjd * beyond the valid range or adding devices to the end 607219089Spjd * of our array which appear to be missing. 608219089Spjd */ 609219089Spjd if (valid_top_config) { 610219089Spjd if (max_id < children) { 611219089Spjd for (c = max_id; c < children; c++) 612219089Spjd nvlist_free(child[c]); 613219089Spjd children = max_id; 614219089Spjd } else if (max_id > children) { 615219089Spjd nvlist_t **newchild; 616219089Spjd 617219089Spjd newchild = zfs_alloc(hdl, (max_id) * 618219089Spjd sizeof (nvlist_t *)); 619219089Spjd if (newchild == NULL) 620219089Spjd goto nomem; 621219089Spjd 622219089Spjd for (c = 0; c < children; c++) 623219089Spjd newchild[c] = child[c]; 624219089Spjd 625219089Spjd free(child); 626219089Spjd child = newchild; 627219089Spjd children = max_id; 628219089Spjd } 629219089Spjd } 630219089Spjd 631168404Spjd verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 632168404Spjd &guid) == 0); 633168404Spjd 634168404Spjd /* 635219089Spjd * The vdev namespace may contain holes as a result of 636219089Spjd * device removal. We must add them back into the vdev 637219089Spjd * tree before we process any missing devices. 638219089Spjd */ 639219089Spjd if (holes > 0) { 640219089Spjd ASSERT(valid_top_config); 641219089Spjd 642219089Spjd for (c = 0; c < children; c++) { 643219089Spjd nvlist_t *holey; 644219089Spjd 645219089Spjd if (child[c] != NULL || 646219089Spjd !vdev_is_hole(hole_array, holes, c)) 647219089Spjd continue; 648219089Spjd 649219089Spjd if (nvlist_alloc(&holey, NV_UNIQUE_NAME, 650219089Spjd 0) != 0) 651219089Spjd goto nomem; 652219089Spjd 653219089Spjd /* 654219089Spjd * Holes in the namespace are treated as 655219089Spjd * "hole" top-level vdevs and have a 656219089Spjd * special flag set on them. 657219089Spjd */ 658219089Spjd if (nvlist_add_string(holey, 659219089Spjd ZPOOL_CONFIG_TYPE, 660219089Spjd VDEV_TYPE_HOLE) != 0 || 661219089Spjd nvlist_add_uint64(holey, 662219089Spjd ZPOOL_CONFIG_ID, c) != 0 || 663219089Spjd nvlist_add_uint64(holey, 664219089Spjd ZPOOL_CONFIG_GUID, 0ULL) != 0) 665219089Spjd goto nomem; 666219089Spjd child[c] = holey; 667219089Spjd } 668219089Spjd } 669219089Spjd 670219089Spjd /* 671168404Spjd * Look for any missing top-level vdevs. If this is the case, 672168404Spjd * create a faked up 'missing' vdev as a placeholder. We cannot 673168404Spjd * simply compress the child array, because the kernel performs 674168404Spjd * certain checks to make sure the vdev IDs match their location 675168404Spjd * in the configuration. 676168404Spjd */ 677219089Spjd for (c = 0; c < children; c++) { 678168404Spjd if (child[c] == NULL) { 679168404Spjd nvlist_t *missing; 680168404Spjd if (nvlist_alloc(&missing, NV_UNIQUE_NAME, 681168404Spjd 0) != 0) 682168404Spjd goto nomem; 683168404Spjd if (nvlist_add_string(missing, 684168404Spjd ZPOOL_CONFIG_TYPE, 685168404Spjd VDEV_TYPE_MISSING) != 0 || 686168404Spjd nvlist_add_uint64(missing, 687168404Spjd ZPOOL_CONFIG_ID, c) != 0 || 688168404Spjd nvlist_add_uint64(missing, 689168404Spjd ZPOOL_CONFIG_GUID, 0ULL) != 0) { 690168404Spjd nvlist_free(missing); 691168404Spjd goto nomem; 692168404Spjd } 693168404Spjd child[c] = missing; 694168404Spjd } 695219089Spjd } 696168404Spjd 697168404Spjd /* 698168404Spjd * Put all of this pool's top-level vdevs into a root vdev. 699168404Spjd */ 700168404Spjd if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) 701168404Spjd goto nomem; 702168404Spjd if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, 703168404Spjd VDEV_TYPE_ROOT) != 0 || 704168404Spjd nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || 705168404Spjd nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || 706168404Spjd nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 707168404Spjd child, children) != 0) { 708168404Spjd nvlist_free(nvroot); 709168404Spjd goto nomem; 710168404Spjd } 711168404Spjd 712168404Spjd for (c = 0; c < children; c++) 713168404Spjd nvlist_free(child[c]); 714168404Spjd free(child); 715168404Spjd children = 0; 716168404Spjd child = NULL; 717168404Spjd 718168404Spjd /* 719168404Spjd * Go through and fix up any paths and/or devids based on our 720168404Spjd * known list of vdev GUID -> path mappings. 721168404Spjd */ 722168404Spjd if (fix_paths(nvroot, pl->names) != 0) { 723168404Spjd nvlist_free(nvroot); 724168404Spjd goto nomem; 725168404Spjd } 726168404Spjd 727168404Spjd /* 728168404Spjd * Add the root vdev to this pool's configuration. 729168404Spjd */ 730168404Spjd if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 731168404Spjd nvroot) != 0) { 732168404Spjd nvlist_free(nvroot); 733168404Spjd goto nomem; 734168404Spjd } 735168404Spjd nvlist_free(nvroot); 736168404Spjd 737168404Spjd /* 738185029Spjd * zdb uses this path to report on active pools that were 739185029Spjd * imported or created using -R. 740185029Spjd */ 741185029Spjd if (active_ok) 742185029Spjd goto add_pool; 743185029Spjd 744185029Spjd /* 745168404Spjd * Determine if this pool is currently active, in which case we 746168404Spjd * can't actually import it. 747168404Spjd */ 748168404Spjd verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 749168404Spjd &name) == 0); 750168404Spjd verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 751168404Spjd &guid) == 0); 752168404Spjd 753168404Spjd if (pool_active(hdl, name, guid, &isactive) != 0) 754168404Spjd goto error; 755168404Spjd 756168404Spjd if (isactive) { 757168404Spjd nvlist_free(config); 758168404Spjd config = NULL; 759168404Spjd continue; 760168404Spjd } 761168404Spjd 762219089Spjd if ((nvl = refresh_config(hdl, config)) == NULL) { 763219089Spjd nvlist_free(config); 764219089Spjd config = NULL; 765219089Spjd continue; 766219089Spjd } 767168404Spjd 768168404Spjd nvlist_free(config); 769185029Spjd config = nvl; 770168404Spjd 771168404Spjd /* 772168404Spjd * Go through and update the paths for spares, now that we have 773168404Spjd * them. 774168404Spjd */ 775168404Spjd verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 776168404Spjd &nvroot) == 0); 777168404Spjd if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 778168404Spjd &spares, &nspares) == 0) { 779168404Spjd for (i = 0; i < nspares; i++) { 780168404Spjd if (fix_paths(spares[i], pl->names) != 0) 781168404Spjd goto nomem; 782168404Spjd } 783168404Spjd } 784168404Spjd 785168404Spjd /* 786185029Spjd * Update the paths for l2cache devices. 787185029Spjd */ 788185029Spjd if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 789185029Spjd &l2cache, &nl2cache) == 0) { 790185029Spjd for (i = 0; i < nl2cache; i++) { 791185029Spjd if (fix_paths(l2cache[i], pl->names) != 0) 792185029Spjd goto nomem; 793185029Spjd } 794185029Spjd } 795185029Spjd 796185029Spjd /* 797168498Spjd * Restore the original information read from the actual label. 798168498Spjd */ 799168498Spjd (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, 800168498Spjd DATA_TYPE_UINT64); 801168498Spjd (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, 802168498Spjd DATA_TYPE_STRING); 803168498Spjd if (hostid != 0) { 804168498Spjd verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 805168498Spjd hostid) == 0); 806168498Spjd verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 807168498Spjd hostname) == 0); 808168498Spjd } 809168498Spjd 810185029Spjdadd_pool: 811168498Spjd /* 812168404Spjd * Add this pool to the list of configs. 813168404Spjd */ 814168404Spjd verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 815168404Spjd &name) == 0); 816168404Spjd if (nvlist_add_nvlist(ret, name, config) != 0) 817168404Spjd goto nomem; 818168404Spjd 819185029Spjd found_one = B_TRUE; 820168404Spjd nvlist_free(config); 821168404Spjd config = NULL; 822168404Spjd } 823168404Spjd 824185029Spjd if (!found_one) { 825185029Spjd nvlist_free(ret); 826185029Spjd ret = NULL; 827185029Spjd } 828185029Spjd 829168404Spjd return (ret); 830168404Spjd 831168404Spjdnomem: 832168404Spjd (void) no_memory(hdl); 833168404Spjderror: 834168404Spjd nvlist_free(config); 835168404Spjd nvlist_free(ret); 836168404Spjd for (c = 0; c < children; c++) 837168404Spjd nvlist_free(child[c]); 838168404Spjd free(child); 839168404Spjd 840168404Spjd return (NULL); 841168404Spjd} 842168404Spjd 843168404Spjd/* 844168404Spjd * Return the offset of the given label. 845168404Spjd */ 846168404Spjdstatic uint64_t 847185029Spjdlabel_offset(uint64_t size, int l) 848168404Spjd{ 849185029Spjd ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0); 850168404Spjd return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 851168404Spjd 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); 852168404Spjd} 853168404Spjd 854168404Spjd/* 855168404Spjd * Given a file descriptor, read the label information and return an nvlist 856168404Spjd * describing the configuration, if there is one. 857168404Spjd */ 858168404Spjdint 859168404Spjdzpool_read_label(int fd, nvlist_t **config) 860168404Spjd{ 861168404Spjd struct stat64 statbuf; 862168404Spjd int l; 863168404Spjd vdev_label_t *label; 864185029Spjd uint64_t state, txg, size; 865168404Spjd 866168404Spjd *config = NULL; 867168404Spjd 868168404Spjd if (fstat64(fd, &statbuf) == -1) 869168404Spjd return (0); 870185029Spjd size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 871168404Spjd 872168404Spjd if ((label = malloc(sizeof (vdev_label_t))) == NULL) 873168404Spjd return (-1); 874168404Spjd 875168404Spjd for (l = 0; l < VDEV_LABELS; l++) { 876185029Spjd if (pread64(fd, label, sizeof (vdev_label_t), 877185029Spjd label_offset(size, l)) != sizeof (vdev_label_t)) 878168404Spjd continue; 879168404Spjd 880168404Spjd if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 881168404Spjd sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) 882168404Spjd continue; 883168404Spjd 884168404Spjd if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 885185029Spjd &state) != 0 || state > POOL_STATE_L2CACHE) { 886168404Spjd nvlist_free(*config); 887168404Spjd continue; 888168404Spjd } 889168404Spjd 890185029Spjd if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 891168404Spjd (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 892168404Spjd &txg) != 0 || txg == 0)) { 893168404Spjd nvlist_free(*config); 894168404Spjd continue; 895168404Spjd } 896168404Spjd 897168404Spjd free(label); 898168404Spjd return (0); 899168404Spjd } 900168404Spjd 901168404Spjd free(label); 902168404Spjd *config = NULL; 903168404Spjd return (0); 904168404Spjd} 905168404Spjd 906219089Spjdtypedef struct rdsk_node { 907219089Spjd char *rn_name; 908219089Spjd int rn_dfd; 909219089Spjd libzfs_handle_t *rn_hdl; 910219089Spjd nvlist_t *rn_config; 911219089Spjd avl_tree_t *rn_avl; 912219089Spjd avl_node_t rn_node; 913219089Spjd boolean_t rn_nozpool; 914219089Spjd} rdsk_node_t; 915219089Spjd 916185029Spjdstatic int 917219089Spjdslice_cache_compare(const void *arg1, const void *arg2) 918168404Spjd{ 919219089Spjd const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; 920219089Spjd const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; 921219089Spjd char *nm1slice, *nm2slice; 922219089Spjd int rv; 923219089Spjd 924219089Spjd /* 925219089Spjd * slices zero and two are the most likely to provide results, 926219089Spjd * so put those first 927219089Spjd */ 928219089Spjd nm1slice = strstr(nm1, "s0"); 929219089Spjd nm2slice = strstr(nm2, "s0"); 930219089Spjd if (nm1slice && !nm2slice) { 931219089Spjd return (-1); 932219089Spjd } 933219089Spjd if (!nm1slice && nm2slice) { 934219089Spjd return (1); 935219089Spjd } 936219089Spjd nm1slice = strstr(nm1, "s2"); 937219089Spjd nm2slice = strstr(nm2, "s2"); 938219089Spjd if (nm1slice && !nm2slice) { 939219089Spjd return (-1); 940219089Spjd } 941219089Spjd if (!nm1slice && nm2slice) { 942219089Spjd return (1); 943219089Spjd } 944219089Spjd 945219089Spjd rv = strcmp(nm1, nm2); 946219089Spjd if (rv == 0) 947219089Spjd return (0); 948219089Spjd return (rv > 0 ? 1 : -1); 949219089Spjd} 950219089Spjd 951219089Spjd#ifdef sun 952219089Spjdstatic void 953219089Spjdcheck_one_slice(avl_tree_t *r, char *diskname, uint_t partno, 954219089Spjd diskaddr_t size, uint_t blksz) 955219089Spjd{ 956219089Spjd rdsk_node_t tmpnode; 957219089Spjd rdsk_node_t *node; 958219089Spjd char sname[MAXNAMELEN]; 959219089Spjd 960219089Spjd tmpnode.rn_name = &sname[0]; 961219089Spjd (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u", 962219089Spjd diskname, partno); 963219089Spjd /* 964219089Spjd * protect against division by zero for disk labels that 965219089Spjd * contain a bogus sector size 966219089Spjd */ 967219089Spjd if (blksz == 0) 968219089Spjd blksz = DEV_BSIZE; 969219089Spjd /* too small to contain a zpool? */ 970219089Spjd if ((size < (SPA_MINDEVSIZE / blksz)) && 971219089Spjd (node = avl_find(r, &tmpnode, NULL))) 972219089Spjd node->rn_nozpool = B_TRUE; 973219089Spjd} 974219089Spjd#endif /* sun */ 975219089Spjd 976219089Spjdstatic void 977219089Spjdnozpool_all_slices(avl_tree_t *r, const char *sname) 978219089Spjd{ 979219089Spjd#ifdef sun 980219089Spjd char diskname[MAXNAMELEN]; 981219089Spjd char *ptr; 982219089Spjd int i; 983219089Spjd 984219089Spjd (void) strncpy(diskname, sname, MAXNAMELEN); 985219089Spjd if (((ptr = strrchr(diskname, 's')) == NULL) && 986219089Spjd ((ptr = strrchr(diskname, 'p')) == NULL)) 987219089Spjd return; 988219089Spjd ptr[0] = 's'; 989219089Spjd ptr[1] = '\0'; 990219089Spjd for (i = 0; i < NDKMAP; i++) 991219089Spjd check_one_slice(r, diskname, i, 0, 1); 992219089Spjd ptr[0] = 'p'; 993219089Spjd for (i = 0; i <= FD_NUMPART; i++) 994219089Spjd check_one_slice(r, diskname, i, 0, 1); 995219089Spjd#endif /* sun */ 996219089Spjd} 997219089Spjd 998219089Spjdstatic void 999219089Spjdcheck_slices(avl_tree_t *r, int fd, const char *sname) 1000219089Spjd{ 1001219089Spjd#ifdef sun 1002219089Spjd struct extvtoc vtoc; 1003219089Spjd struct dk_gpt *gpt; 1004219089Spjd char diskname[MAXNAMELEN]; 1005219089Spjd char *ptr; 1006219089Spjd int i; 1007219089Spjd 1008219089Spjd (void) strncpy(diskname, sname, MAXNAMELEN); 1009219089Spjd if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1])) 1010219089Spjd return; 1011219089Spjd ptr[1] = '\0'; 1012219089Spjd 1013219089Spjd if (read_extvtoc(fd, &vtoc) >= 0) { 1014219089Spjd for (i = 0; i < NDKMAP; i++) 1015219089Spjd check_one_slice(r, diskname, i, 1016219089Spjd vtoc.v_part[i].p_size, vtoc.v_sectorsz); 1017219089Spjd } else if (efi_alloc_and_read(fd, &gpt) >= 0) { 1018219089Spjd /* 1019219089Spjd * on x86 we'll still have leftover links that point 1020219089Spjd * to slices s[9-15], so use NDKMAP instead 1021219089Spjd */ 1022219089Spjd for (i = 0; i < NDKMAP; i++) 1023219089Spjd check_one_slice(r, diskname, i, 1024219089Spjd gpt->efi_parts[i].p_size, gpt->efi_lbasize); 1025219089Spjd /* nodes p[1-4] are never used with EFI labels */ 1026219089Spjd ptr[0] = 'p'; 1027219089Spjd for (i = 1; i <= FD_NUMPART; i++) 1028219089Spjd check_one_slice(r, diskname, i, 0, 1); 1029219089Spjd efi_free(gpt); 1030219089Spjd } 1031219089Spjd#endif /* sun */ 1032219089Spjd} 1033219089Spjd 1034219089Spjdstatic void 1035219089Spjdzpool_open_func(void *arg) 1036219089Spjd{ 1037219089Spjd rdsk_node_t *rn = arg; 1038219089Spjd struct stat64 statbuf; 1039185029Spjd nvlist_t *config; 1040219089Spjd int fd; 1041168404Spjd 1042219089Spjd if (rn->rn_nozpool) 1043219089Spjd return; 1044219089Spjd if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { 1045219089Spjd /* symlink to a device that's no longer there */ 1046219089Spjd if (errno == ENOENT) 1047219089Spjd nozpool_all_slices(rn->rn_avl, rn->rn_name); 1048219089Spjd return; 1049219089Spjd } 1050168404Spjd /* 1051219089Spjd * Ignore failed stats. We only want regular 1052219089Spjd * files, character devs and block devs. 1053168404Spjd */ 1054219089Spjd if (fstat64(fd, &statbuf) != 0 || 1055219089Spjd (!S_ISREG(statbuf.st_mode) && 1056219089Spjd !S_ISCHR(statbuf.st_mode) && 1057219089Spjd !S_ISBLK(statbuf.st_mode))) { 1058219089Spjd (void) close(fd); 1059219089Spjd return; 1060219089Spjd } 1061219089Spjd /* this file is too small to hold a zpool */ 1062219089Spjd if (S_ISREG(statbuf.st_mode) && 1063219089Spjd statbuf.st_size < SPA_MINDEVSIZE) { 1064219089Spjd (void) close(fd); 1065219089Spjd return; 1066219089Spjd } else if (!S_ISREG(statbuf.st_mode)) { 1067219089Spjd /* 1068219089Spjd * Try to read the disk label first so we don't have to 1069219089Spjd * open a bunch of minor nodes that can't have a zpool. 1070219089Spjd */ 1071219089Spjd check_slices(rn->rn_avl, fd, rn->rn_name); 1072219089Spjd } 1073168404Spjd 1074219089Spjd if ((zpool_read_label(fd, &config)) != 0) { 1075219089Spjd (void) close(fd); 1076219089Spjd (void) no_memory(rn->rn_hdl); 1077219089Spjd return; 1078219089Spjd } 1079219089Spjd (void) close(fd); 1080168404Spjd 1081168404Spjd 1082219089Spjd rn->rn_config = config; 1083219089Spjd if (config != NULL) { 1084219089Spjd assert(rn->rn_nozpool == B_FALSE); 1085219089Spjd } 1086219089Spjd} 1087168404Spjd 1088219089Spjd/* 1089219089Spjd * Given a file descriptor, clear (zero) the label information. This function 1090224171Sgibbs * is used in the appliance stack as part of the ZFS sysevent module and 1091224171Sgibbs * to implement the "zpool labelclear" command. 1092219089Spjd */ 1093219089Spjdint 1094219089Spjdzpool_clear_label(int fd) 1095219089Spjd{ 1096219089Spjd struct stat64 statbuf; 1097219089Spjd int l; 1098219089Spjd vdev_label_t *label; 1099219089Spjd uint64_t size; 1100168404Spjd 1101219089Spjd if (fstat64(fd, &statbuf) == -1) 1102219089Spjd return (0); 1103219089Spjd size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 1104168404Spjd 1105219089Spjd if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) 1106219089Spjd return (-1); 1107168404Spjd 1108219089Spjd for (l = 0; l < VDEV_LABELS; l++) { 1109219089Spjd if (pwrite64(fd, label, sizeof (vdev_label_t), 1110219089Spjd label_offset(size, l)) != sizeof (vdev_label_t)) 1111219089Spjd return (-1); 1112185029Spjd } 1113219089Spjd 1114219089Spjd free(label); 1115219089Spjd return (0); 1116185029Spjd} 1117185029Spjd 1118185029Spjd/* 1119185029Spjd * Given a list of directories to search, find all pools stored on disk. This 1120185029Spjd * includes partial pools which are not available to import. If no args are 1121185029Spjd * given (argc is 0), then the default directory (/dev/dsk) is searched. 1122185029Spjd * poolname or guid (but not both) are provided by the caller when trying 1123185029Spjd * to import a specific pool. 1124185029Spjd */ 1125185029Spjdstatic nvlist_t * 1126219089Spjdzpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg) 1127185029Spjd{ 1128219089Spjd int i, dirs = iarg->paths; 1129185029Spjd DIR *dirp = NULL; 1130185029Spjd struct dirent64 *dp; 1131185029Spjd char path[MAXPATHLEN]; 1132219089Spjd char *end, **dir = iarg->path; 1133185029Spjd size_t pathleft; 1134219089Spjd nvlist_t *ret = NULL; 1135235479Savg static char *default_dir = "/dev"; 1136185029Spjd pool_list_t pools = { 0 }; 1137185029Spjd pool_entry_t *pe, *penext; 1138185029Spjd vdev_entry_t *ve, *venext; 1139185029Spjd config_entry_t *ce, *cenext; 1140185029Spjd name_entry_t *ne, *nenext; 1141219089Spjd avl_tree_t slice_cache; 1142219089Spjd rdsk_node_t *slice; 1143219089Spjd void *cookie; 1144185029Spjd 1145219089Spjd if (dirs == 0) { 1146219089Spjd dirs = 1; 1147219089Spjd dir = &default_dir; 1148185029Spjd } 1149185029Spjd 1150185029Spjd /* 1151185029Spjd * Go through and read the label configuration information from every 1152185029Spjd * possible device, organizing the information according to pool GUID 1153185029Spjd * and toplevel GUID. 1154185029Spjd */ 1155219089Spjd for (i = 0; i < dirs; i++) { 1156219089Spjd tpool_t *t; 1157185029Spjd char *rdsk; 1158185029Spjd int dfd; 1159185029Spjd 1160185029Spjd /* use realpath to normalize the path */ 1161219089Spjd if (realpath(dir[i], path) == 0) { 1162185029Spjd (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1163219089Spjd dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]); 1164185029Spjd goto error; 1165185029Spjd } 1166185029Spjd end = &path[strlen(path)]; 1167185029Spjd *end++ = '/'; 1168185029Spjd *end = 0; 1169185029Spjd pathleft = &path[sizeof (path)] - end; 1170185029Spjd 1171185029Spjd /* 1172185029Spjd * Using raw devices instead of block devices when we're 1173185029Spjd * reading the labels skips a bunch of slow operations during 1174185029Spjd * close(2) processing, so we replace /dev/dsk with /dev/rdsk. 1175185029Spjd */ 1176185029Spjd if (strcmp(path, "/dev/dsk/") == 0) 1177219089Spjd rdsk = "/dev/"; 1178185029Spjd else 1179185029Spjd rdsk = path; 1180185029Spjd 1181219089Spjd if ((dfd = open64(rdsk, O_RDONLY)) < 0 || 1182219089Spjd (dirp = fdopendir(dfd)) == NULL) { 1183185029Spjd zfs_error_aux(hdl, strerror(errno)); 1184185029Spjd (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1185185029Spjd dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1186185029Spjd rdsk); 1187185029Spjd goto error; 1188185029Spjd } 1189185029Spjd 1190219089Spjd avl_create(&slice_cache, slice_cache_compare, 1191219089Spjd sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node)); 1192219089Spjd 1193219089Spjd if (strcmp(rdsk, "/dev/") == 0) { 1194219089Spjd struct gmesh mesh; 1195219089Spjd struct gclass *mp; 1196219089Spjd struct ggeom *gp; 1197219089Spjd struct gprovider *pp; 1198219089Spjd 1199219089Spjd errno = geom_gettree(&mesh); 1200219089Spjd if (errno != 0) { 1201219089Spjd zfs_error_aux(hdl, strerror(errno)); 1202219089Spjd (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1203219089Spjd dgettext(TEXT_DOMAIN, "cannot get GEOM tree")); 1204219089Spjd goto error; 1205219089Spjd } 1206219089Spjd 1207219089Spjd LIST_FOREACH(mp, &mesh.lg_class, lg_class) { 1208219089Spjd LIST_FOREACH(gp, &mp->lg_geom, lg_geom) { 1209219089Spjd LIST_FOREACH(pp, &gp->lg_provider, lg_provider) { 1210219089Spjd slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 1211219089Spjd slice->rn_name = zfs_strdup(hdl, pp->lg_name); 1212219089Spjd slice->rn_avl = &slice_cache; 1213219089Spjd slice->rn_dfd = dfd; 1214219089Spjd slice->rn_hdl = hdl; 1215219089Spjd slice->rn_nozpool = B_FALSE; 1216219089Spjd avl_add(&slice_cache, slice); 1217219089Spjd } 1218219089Spjd } 1219219089Spjd } 1220219089Spjd 1221219089Spjd geom_deletetree(&mesh); 1222219089Spjd goto skipdir; 1223219089Spjd } 1224219089Spjd 1225185029Spjd /* 1226185029Spjd * This is not MT-safe, but we have no MT consumers of libzfs 1227185029Spjd */ 1228185029Spjd while ((dp = readdir64(dirp)) != NULL) { 1229185029Spjd const char *name = dp->d_name; 1230185029Spjd if (name[0] == '.' && 1231185029Spjd (name[1] == 0 || (name[1] == '.' && name[2] == 0))) 1232185029Spjd continue; 1233185029Spjd 1234219089Spjd slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 1235219089Spjd slice->rn_name = zfs_strdup(hdl, name); 1236219089Spjd slice->rn_avl = &slice_cache; 1237219089Spjd slice->rn_dfd = dfd; 1238219089Spjd slice->rn_hdl = hdl; 1239219089Spjd slice->rn_nozpool = B_FALSE; 1240219089Spjd avl_add(&slice_cache, slice); 1241219089Spjd } 1242219089Spjdskipdir: 1243219089Spjd /* 1244219089Spjd * create a thread pool to do all of this in parallel; 1245219089Spjd * rn_nozpool is not protected, so this is racy in that 1246219089Spjd * multiple tasks could decide that the same slice can 1247219089Spjd * not hold a zpool, which is benign. Also choose 1248219089Spjd * double the number of processors; we hold a lot of 1249219089Spjd * locks in the kernel, so going beyond this doesn't 1250219089Spjd * buy us much. 1251219089Spjd */ 1252219089Spjd t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN), 1253219089Spjd 0, NULL); 1254219089Spjd for (slice = avl_first(&slice_cache); slice; 1255219089Spjd (slice = avl_walk(&slice_cache, slice, 1256219089Spjd AVL_AFTER))) 1257219089Spjd (void) tpool_dispatch(t, zpool_open_func, slice); 1258219089Spjd tpool_wait(t); 1259219089Spjd tpool_destroy(t); 1260185029Spjd 1261219089Spjd cookie = NULL; 1262219089Spjd while ((slice = avl_destroy_nodes(&slice_cache, 1263219089Spjd &cookie)) != NULL) { 1264219089Spjd if (slice->rn_config != NULL) { 1265219089Spjd nvlist_t *config = slice->rn_config; 1266185029Spjd boolean_t matched = B_TRUE; 1267185029Spjd 1268219089Spjd if (iarg->poolname != NULL) { 1269185029Spjd char *pname; 1270185029Spjd 1271185029Spjd matched = nvlist_lookup_string(config, 1272185029Spjd ZPOOL_CONFIG_POOL_NAME, 1273185029Spjd &pname) == 0 && 1274219089Spjd strcmp(iarg->poolname, pname) == 0; 1275219089Spjd } else if (iarg->guid != 0) { 1276185029Spjd uint64_t this_guid; 1277185029Spjd 1278185029Spjd matched = nvlist_lookup_uint64(config, 1279185029Spjd ZPOOL_CONFIG_POOL_GUID, 1280185029Spjd &this_guid) == 0 && 1281219089Spjd iarg->guid == this_guid; 1282185029Spjd } 1283185029Spjd if (!matched) { 1284185029Spjd nvlist_free(config); 1285185029Spjd config = NULL; 1286185029Spjd continue; 1287185029Spjd } 1288185029Spjd /* use the non-raw path for the config */ 1289219089Spjd (void) strlcpy(end, slice->rn_name, pathleft); 1290168404Spjd if (add_config(hdl, &pools, path, config) != 0) 1291168404Spjd goto error; 1292168404Spjd } 1293219089Spjd free(slice->rn_name); 1294219089Spjd free(slice); 1295168404Spjd } 1296219089Spjd avl_destroy(&slice_cache); 1297185029Spjd 1298185029Spjd (void) closedir(dirp); 1299185029Spjd dirp = NULL; 1300168404Spjd } 1301168404Spjd 1302219089Spjd ret = get_configs(hdl, &pools, iarg->can_be_active); 1303168404Spjd 1304168404Spjderror: 1305168404Spjd for (pe = pools.pools; pe != NULL; pe = penext) { 1306168404Spjd penext = pe->pe_next; 1307168404Spjd for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { 1308168404Spjd venext = ve->ve_next; 1309168404Spjd for (ce = ve->ve_configs; ce != NULL; ce = cenext) { 1310168404Spjd cenext = ce->ce_next; 1311168404Spjd if (ce->ce_config) 1312168404Spjd nvlist_free(ce->ce_config); 1313168404Spjd free(ce); 1314168404Spjd } 1315168404Spjd free(ve); 1316168404Spjd } 1317168404Spjd free(pe); 1318168404Spjd } 1319168404Spjd 1320168404Spjd for (ne = pools.names; ne != NULL; ne = nenext) { 1321168404Spjd nenext = ne->ne_next; 1322168404Spjd if (ne->ne_name) 1323168404Spjd free(ne->ne_name); 1324168404Spjd free(ne); 1325168404Spjd } 1326168404Spjd 1327185029Spjd if (dirp) 1328185029Spjd (void) closedir(dirp); 1329185029Spjd 1330168404Spjd return (ret); 1331168404Spjd} 1332168404Spjd 1333185029Spjdnvlist_t * 1334185029Spjdzpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) 1335185029Spjd{ 1336219089Spjd importargs_t iarg = { 0 }; 1337185029Spjd 1338219089Spjd iarg.paths = argc; 1339219089Spjd iarg.path = argv; 1340185029Spjd 1341219089Spjd return (zpool_find_import_impl(hdl, &iarg)); 1342185029Spjd} 1343185029Spjd 1344185029Spjd/* 1345185029Spjd * Given a cache file, return the contents as a list of importable pools. 1346185029Spjd * poolname or guid (but not both) are provided by the caller when trying 1347185029Spjd * to import a specific pool. 1348185029Spjd */ 1349185029Spjdnvlist_t * 1350185029Spjdzpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, 1351185029Spjd char *poolname, uint64_t guid) 1352185029Spjd{ 1353185029Spjd char *buf; 1354185029Spjd int fd; 1355185029Spjd struct stat64 statbuf; 1356185029Spjd nvlist_t *raw, *src, *dst; 1357185029Spjd nvlist_t *pools; 1358185029Spjd nvpair_t *elem; 1359185029Spjd char *name; 1360185029Spjd uint64_t this_guid; 1361185029Spjd boolean_t active; 1362185029Spjd 1363185029Spjd verify(poolname == NULL || guid == 0); 1364185029Spjd 1365185029Spjd if ((fd = open(cachefile, O_RDONLY)) < 0) { 1366185029Spjd zfs_error_aux(hdl, "%s", strerror(errno)); 1367185029Spjd (void) zfs_error(hdl, EZFS_BADCACHE, 1368185029Spjd dgettext(TEXT_DOMAIN, "failed to open cache file")); 1369185029Spjd return (NULL); 1370185029Spjd } 1371185029Spjd 1372185029Spjd if (fstat64(fd, &statbuf) != 0) { 1373185029Spjd zfs_error_aux(hdl, "%s", strerror(errno)); 1374185029Spjd (void) close(fd); 1375185029Spjd (void) zfs_error(hdl, EZFS_BADCACHE, 1376185029Spjd dgettext(TEXT_DOMAIN, "failed to get size of cache file")); 1377185029Spjd return (NULL); 1378185029Spjd } 1379185029Spjd 1380185029Spjd if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { 1381185029Spjd (void) close(fd); 1382185029Spjd return (NULL); 1383185029Spjd } 1384185029Spjd 1385185029Spjd if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 1386185029Spjd (void) close(fd); 1387185029Spjd free(buf); 1388185029Spjd (void) zfs_error(hdl, EZFS_BADCACHE, 1389185029Spjd dgettext(TEXT_DOMAIN, 1390185029Spjd "failed to read cache file contents")); 1391185029Spjd return (NULL); 1392185029Spjd } 1393185029Spjd 1394185029Spjd (void) close(fd); 1395185029Spjd 1396185029Spjd if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { 1397185029Spjd free(buf); 1398185029Spjd (void) zfs_error(hdl, EZFS_BADCACHE, 1399185029Spjd dgettext(TEXT_DOMAIN, 1400185029Spjd "invalid or corrupt cache file contents")); 1401185029Spjd return (NULL); 1402185029Spjd } 1403185029Spjd 1404185029Spjd free(buf); 1405185029Spjd 1406185029Spjd /* 1407185029Spjd * Go through and get the current state of the pools and refresh their 1408185029Spjd * state. 1409185029Spjd */ 1410185029Spjd if (nvlist_alloc(&pools, 0, 0) != 0) { 1411185029Spjd (void) no_memory(hdl); 1412185029Spjd nvlist_free(raw); 1413185029Spjd return (NULL); 1414185029Spjd } 1415185029Spjd 1416185029Spjd elem = NULL; 1417185029Spjd while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { 1418185029Spjd verify(nvpair_value_nvlist(elem, &src) == 0); 1419185029Spjd 1420185029Spjd verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME, 1421185029Spjd &name) == 0); 1422185029Spjd if (poolname != NULL && strcmp(poolname, name) != 0) 1423185029Spjd continue; 1424185029Spjd 1425185029Spjd verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 1426185029Spjd &this_guid) == 0); 1427185029Spjd if (guid != 0) { 1428185029Spjd verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 1429185029Spjd &this_guid) == 0); 1430185029Spjd if (guid != this_guid) 1431185029Spjd continue; 1432185029Spjd } 1433185029Spjd 1434185029Spjd if (pool_active(hdl, name, this_guid, &active) != 0) { 1435185029Spjd nvlist_free(raw); 1436185029Spjd nvlist_free(pools); 1437185029Spjd return (NULL); 1438185029Spjd } 1439185029Spjd 1440185029Spjd if (active) 1441185029Spjd continue; 1442185029Spjd 1443185029Spjd if ((dst = refresh_config(hdl, src)) == NULL) { 1444185029Spjd nvlist_free(raw); 1445185029Spjd nvlist_free(pools); 1446185029Spjd return (NULL); 1447185029Spjd } 1448185029Spjd 1449185029Spjd if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { 1450185029Spjd (void) no_memory(hdl); 1451185029Spjd nvlist_free(dst); 1452185029Spjd nvlist_free(raw); 1453185029Spjd nvlist_free(pools); 1454185029Spjd return (NULL); 1455185029Spjd } 1456185029Spjd nvlist_free(dst); 1457185029Spjd } 1458185029Spjd 1459185029Spjd nvlist_free(raw); 1460185029Spjd return (pools); 1461185029Spjd} 1462185029Spjd 1463219089Spjdstatic int 1464219089Spjdname_or_guid_exists(zpool_handle_t *zhp, void *data) 1465219089Spjd{ 1466219089Spjd importargs_t *import = data; 1467219089Spjd int found = 0; 1468185029Spjd 1469219089Spjd if (import->poolname != NULL) { 1470219089Spjd char *pool_name; 1471219089Spjd 1472219089Spjd verify(nvlist_lookup_string(zhp->zpool_config, 1473219089Spjd ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); 1474219089Spjd if (strcmp(pool_name, import->poolname) == 0) 1475219089Spjd found = 1; 1476219089Spjd } else { 1477219089Spjd uint64_t pool_guid; 1478219089Spjd 1479219089Spjd verify(nvlist_lookup_uint64(zhp->zpool_config, 1480219089Spjd ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); 1481219089Spjd if (pool_guid == import->guid) 1482219089Spjd found = 1; 1483219089Spjd } 1484219089Spjd 1485219089Spjd zpool_close(zhp); 1486219089Spjd return (found); 1487219089Spjd} 1488219089Spjd 1489219089Spjdnvlist_t * 1490219089Spjdzpool_search_import(libzfs_handle_t *hdl, importargs_t *import) 1491219089Spjd{ 1492219089Spjd verify(import->poolname == NULL || import->guid == 0); 1493219089Spjd 1494219089Spjd if (import->unique) 1495219089Spjd import->exists = zpool_iter(hdl, name_or_guid_exists, import); 1496219089Spjd 1497219089Spjd if (import->cachefile != NULL) 1498219089Spjd return (zpool_find_import_cached(hdl, import->cachefile, 1499219089Spjd import->poolname, import->guid)); 1500219089Spjd 1501219089Spjd return (zpool_find_import_impl(hdl, import)); 1502219089Spjd} 1503219089Spjd 1504168404Spjdboolean_t 1505168404Spjdfind_guid(nvlist_t *nv, uint64_t guid) 1506168404Spjd{ 1507168404Spjd uint64_t tmp; 1508168404Spjd nvlist_t **child; 1509168404Spjd uint_t c, children; 1510168404Spjd 1511168404Spjd verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); 1512168404Spjd if (tmp == guid) 1513168404Spjd return (B_TRUE); 1514168404Spjd 1515168404Spjd if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1516168404Spjd &child, &children) == 0) { 1517168404Spjd for (c = 0; c < children; c++) 1518168404Spjd if (find_guid(child[c], guid)) 1519168404Spjd return (B_TRUE); 1520168404Spjd } 1521168404Spjd 1522168404Spjd return (B_FALSE); 1523168404Spjd} 1524168404Spjd 1525185029Spjdtypedef struct aux_cbdata { 1526185029Spjd const char *cb_type; 1527168404Spjd uint64_t cb_guid; 1528168404Spjd zpool_handle_t *cb_zhp; 1529185029Spjd} aux_cbdata_t; 1530168404Spjd 1531168404Spjdstatic int 1532185029Spjdfind_aux(zpool_handle_t *zhp, void *data) 1533168404Spjd{ 1534185029Spjd aux_cbdata_t *cbp = data; 1535185029Spjd nvlist_t **list; 1536185029Spjd uint_t i, count; 1537168404Spjd uint64_t guid; 1538168404Spjd nvlist_t *nvroot; 1539168404Spjd 1540168404Spjd verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 1541168404Spjd &nvroot) == 0); 1542168404Spjd 1543185029Spjd if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, 1544185029Spjd &list, &count) == 0) { 1545185029Spjd for (i = 0; i < count; i++) { 1546185029Spjd verify(nvlist_lookup_uint64(list[i], 1547168404Spjd ZPOOL_CONFIG_GUID, &guid) == 0); 1548168404Spjd if (guid == cbp->cb_guid) { 1549168404Spjd cbp->cb_zhp = zhp; 1550168404Spjd return (1); 1551168404Spjd } 1552168404Spjd } 1553168404Spjd } 1554168404Spjd 1555168404Spjd zpool_close(zhp); 1556168404Spjd return (0); 1557168404Spjd} 1558168404Spjd 1559168404Spjd/* 1560168404Spjd * Determines if the pool is in use. If so, it returns true and the state of 1561168404Spjd * the pool as well as the name of the pool. Both strings are allocated and 1562168404Spjd * must be freed by the caller. 1563168404Spjd */ 1564168404Spjdint 1565168404Spjdzpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, 1566168404Spjd boolean_t *inuse) 1567168404Spjd{ 1568168404Spjd nvlist_t *config; 1569168404Spjd char *name; 1570168404Spjd boolean_t ret; 1571168404Spjd uint64_t guid, vdev_guid; 1572168404Spjd zpool_handle_t *zhp; 1573168404Spjd nvlist_t *pool_config; 1574168404Spjd uint64_t stateval, isspare; 1575185029Spjd aux_cbdata_t cb = { 0 }; 1576168404Spjd boolean_t isactive; 1577168404Spjd 1578168404Spjd *inuse = B_FALSE; 1579168404Spjd 1580168404Spjd if (zpool_read_label(fd, &config) != 0) { 1581168404Spjd (void) no_memory(hdl); 1582168404Spjd return (-1); 1583168404Spjd } 1584168404Spjd 1585168404Spjd if (config == NULL) 1586168404Spjd return (0); 1587168404Spjd 1588168404Spjd verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 1589168404Spjd &stateval) == 0); 1590168404Spjd verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 1591168404Spjd &vdev_guid) == 0); 1592168404Spjd 1593185029Spjd if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { 1594168404Spjd verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 1595168404Spjd &name) == 0); 1596168404Spjd verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 1597168404Spjd &guid) == 0); 1598168404Spjd } 1599168404Spjd 1600168404Spjd switch (stateval) { 1601168404Spjd case POOL_STATE_EXPORTED: 1602219089Spjd /* 1603219089Spjd * A pool with an exported state may in fact be imported 1604219089Spjd * read-only, so check the in-core state to see if it's 1605219089Spjd * active and imported read-only. If it is, set 1606219089Spjd * its state to active. 1607219089Spjd */ 1608219089Spjd if (pool_active(hdl, name, guid, &isactive) == 0 && isactive && 1609219089Spjd (zhp = zpool_open_canfail(hdl, name)) != NULL && 1610219089Spjd zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL)) 1611219089Spjd stateval = POOL_STATE_ACTIVE; 1612219089Spjd 1613168404Spjd ret = B_TRUE; 1614168404Spjd break; 1615168404Spjd 1616168404Spjd case POOL_STATE_ACTIVE: 1617168404Spjd /* 1618168404Spjd * For an active pool, we have to determine if it's really part 1619168404Spjd * of a currently active pool (in which case the pool will exist 1620168404Spjd * and the guid will be the same), or whether it's part of an 1621168404Spjd * active pool that was disconnected without being explicitly 1622168404Spjd * exported. 1623168404Spjd */ 1624168404Spjd if (pool_active(hdl, name, guid, &isactive) != 0) { 1625168404Spjd nvlist_free(config); 1626168404Spjd return (-1); 1627168404Spjd } 1628168404Spjd 1629168404Spjd if (isactive) { 1630168404Spjd /* 1631168404Spjd * Because the device may have been removed while 1632168404Spjd * offlined, we only report it as active if the vdev is 1633168404Spjd * still present in the config. Otherwise, pretend like 1634168404Spjd * it's not in use. 1635168404Spjd */ 1636168404Spjd if ((zhp = zpool_open_canfail(hdl, name)) != NULL && 1637168404Spjd (pool_config = zpool_get_config(zhp, NULL)) 1638168404Spjd != NULL) { 1639168404Spjd nvlist_t *nvroot; 1640168404Spjd 1641168404Spjd verify(nvlist_lookup_nvlist(pool_config, 1642168404Spjd ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 1643168404Spjd ret = find_guid(nvroot, vdev_guid); 1644168404Spjd } else { 1645168404Spjd ret = B_FALSE; 1646168404Spjd } 1647168404Spjd 1648168404Spjd /* 1649168404Spjd * If this is an active spare within another pool, we 1650168404Spjd * treat it like an unused hot spare. This allows the 1651168404Spjd * user to create a pool with a hot spare that currently 1652168404Spjd * in use within another pool. Since we return B_TRUE, 1653168404Spjd * libdiskmgt will continue to prevent generic consumers 1654168404Spjd * from using the device. 1655168404Spjd */ 1656168404Spjd if (ret && nvlist_lookup_uint64(config, 1657168404Spjd ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) 1658168404Spjd stateval = POOL_STATE_SPARE; 1659168404Spjd 1660168404Spjd if (zhp != NULL) 1661168404Spjd zpool_close(zhp); 1662168404Spjd } else { 1663168404Spjd stateval = POOL_STATE_POTENTIALLY_ACTIVE; 1664168404Spjd ret = B_TRUE; 1665168404Spjd } 1666168404Spjd break; 1667168404Spjd 1668168404Spjd case POOL_STATE_SPARE: 1669168404Spjd /* 1670168404Spjd * For a hot spare, it can be either definitively in use, or 1671168404Spjd * potentially active. To determine if it's in use, we iterate 1672168404Spjd * over all pools in the system and search for one with a spare 1673168404Spjd * with a matching guid. 1674168404Spjd * 1675168404Spjd * Due to the shared nature of spares, we don't actually report 1676168404Spjd * the potentially active case as in use. This means the user 1677168404Spjd * can freely create pools on the hot spares of exported pools, 1678168404Spjd * but to do otherwise makes the resulting code complicated, and 1679168404Spjd * we end up having to deal with this case anyway. 1680168404Spjd */ 1681168404Spjd cb.cb_zhp = NULL; 1682168404Spjd cb.cb_guid = vdev_guid; 1683185029Spjd cb.cb_type = ZPOOL_CONFIG_SPARES; 1684185029Spjd if (zpool_iter(hdl, find_aux, &cb) == 1) { 1685168404Spjd name = (char *)zpool_get_name(cb.cb_zhp); 1686168404Spjd ret = TRUE; 1687168404Spjd } else { 1688168404Spjd ret = FALSE; 1689168404Spjd } 1690168404Spjd break; 1691168404Spjd 1692185029Spjd case POOL_STATE_L2CACHE: 1693185029Spjd 1694185029Spjd /* 1695185029Spjd * Check if any pool is currently using this l2cache device. 1696185029Spjd */ 1697185029Spjd cb.cb_zhp = NULL; 1698185029Spjd cb.cb_guid = vdev_guid; 1699185029Spjd cb.cb_type = ZPOOL_CONFIG_L2CACHE; 1700185029Spjd if (zpool_iter(hdl, find_aux, &cb) == 1) { 1701185029Spjd name = (char *)zpool_get_name(cb.cb_zhp); 1702185029Spjd ret = TRUE; 1703185029Spjd } else { 1704185029Spjd ret = FALSE; 1705185029Spjd } 1706185029Spjd break; 1707185029Spjd 1708168404Spjd default: 1709168404Spjd ret = B_FALSE; 1710168404Spjd } 1711168404Spjd 1712168404Spjd 1713168404Spjd if (ret) { 1714168404Spjd if ((*namestr = zfs_strdup(hdl, name)) == NULL) { 1715185029Spjd if (cb.cb_zhp) 1716185029Spjd zpool_close(cb.cb_zhp); 1717168404Spjd nvlist_free(config); 1718168404Spjd return (-1); 1719168404Spjd } 1720168404Spjd *state = (pool_state_t)stateval; 1721168404Spjd } 1722168404Spjd 1723168404Spjd if (cb.cb_zhp) 1724168404Spjd zpool_close(cb.cb_zhp); 1725168404Spjd 1726168404Spjd nvlist_free(config); 1727168404Spjd *inuse = ret; 1728168404Spjd return (0); 1729168404Spjd} 1730