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 /* 22*f9af39baSGeorge Wilson * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23fa9e4066Sahrens */ 24fa9e4066Sahrens 25fa9e4066Sahrens /* 26fa9e4066Sahrens * Pool import support functions. 27fa9e4066Sahrens * 28fa9e4066Sahrens * To import a pool, we rely on reading the configuration information from the 29fa9e4066Sahrens * ZFS label of each device. If we successfully read the label, then we 30fa9e4066Sahrens * organize the configuration information in the following hierarchy: 31fa9e4066Sahrens * 32fa9e4066Sahrens * pool guid -> toplevel vdev guid -> label txg 33fa9e4066Sahrens * 34fa9e4066Sahrens * Duplicate entries matching this same tuple will be discarded. Once we have 35fa9e4066Sahrens * examined every device, we pick the best label txg config for each toplevel 36fa9e4066Sahrens * vdev. We then arrange these toplevel vdevs into a complete pool config, and 37fa9e4066Sahrens * update any paths that have changed. Finally, we attempt to import the pool 38fa9e4066Sahrens * using our derived config, and record the results. 39fa9e4066Sahrens */ 40fa9e4066Sahrens 414f67d755SEric Taylor #include <ctype.h> 42fa9e4066Sahrens #include <devid.h> 43fa9e4066Sahrens #include <dirent.h> 44fa9e4066Sahrens #include <errno.h> 45fa9e4066Sahrens #include <libintl.h> 464f67d755SEric Taylor #include <stddef.h> 47fa9e4066Sahrens #include <stdlib.h> 48fa9e4066Sahrens #include <string.h> 49fa9e4066Sahrens #include <sys/stat.h> 50fa9e4066Sahrens #include <unistd.h> 51fa9e4066Sahrens #include <fcntl.h> 524f67d755SEric Taylor #include <sys/vtoc.h> 534f67d755SEric Taylor #include <sys/dktp/fdisk.h> 544f67d755SEric Taylor #include <sys/efi_partition.h> 554f67d755SEric Taylor #include <thread_pool.h> 56fa9e4066Sahrens 57fa9e4066Sahrens #include <sys/vdev_impl.h> 58fa9e4066Sahrens 59fa9e4066Sahrens #include "libzfs.h" 60fa9e4066Sahrens #include "libzfs_impl.h" 61fa9e4066Sahrens 62fa9e4066Sahrens /* 63fa9e4066Sahrens * Intermediate structures used to gather configuration information. 64fa9e4066Sahrens */ 65fa9e4066Sahrens typedef struct config_entry { 66fa9e4066Sahrens uint64_t ce_txg; 67fa9e4066Sahrens nvlist_t *ce_config; 68fa9e4066Sahrens struct config_entry *ce_next; 69fa9e4066Sahrens } config_entry_t; 70fa9e4066Sahrens 71fa9e4066Sahrens typedef struct vdev_entry { 72fa9e4066Sahrens uint64_t ve_guid; 73fa9e4066Sahrens config_entry_t *ve_configs; 74fa9e4066Sahrens struct vdev_entry *ve_next; 75fa9e4066Sahrens } vdev_entry_t; 76fa9e4066Sahrens 77fa9e4066Sahrens typedef struct pool_entry { 78fa9e4066Sahrens uint64_t pe_guid; 79fa9e4066Sahrens vdev_entry_t *pe_vdevs; 80fa9e4066Sahrens struct pool_entry *pe_next; 81fa9e4066Sahrens } pool_entry_t; 82fa9e4066Sahrens 83fa9e4066Sahrens typedef struct name_entry { 8499653d4eSeschrock char *ne_name; 85fa9e4066Sahrens uint64_t ne_guid; 86fa9e4066Sahrens struct name_entry *ne_next; 87fa9e4066Sahrens } name_entry_t; 88fa9e4066Sahrens 89fa9e4066Sahrens typedef struct pool_list { 90fa9e4066Sahrens pool_entry_t *pools; 91fa9e4066Sahrens name_entry_t *names; 92fa9e4066Sahrens } pool_list_t; 93fa9e4066Sahrens 94fa9e4066Sahrens static char * 95fa9e4066Sahrens get_devid(const char *path) 96fa9e4066Sahrens { 97fa9e4066Sahrens int fd; 98fa9e4066Sahrens ddi_devid_t devid; 99fa9e4066Sahrens char *minor, *ret; 100fa9e4066Sahrens 101fa9e4066Sahrens if ((fd = open(path, O_RDONLY)) < 0) 102fa9e4066Sahrens return (NULL); 103fa9e4066Sahrens 104fa9e4066Sahrens minor = NULL; 105fa9e4066Sahrens ret = NULL; 106fa9e4066Sahrens if (devid_get(fd, &devid) == 0) { 107fa9e4066Sahrens if (devid_get_minor_name(fd, &minor) == 0) 108fa9e4066Sahrens ret = devid_str_encode(devid, minor); 109fa9e4066Sahrens if (minor != NULL) 110fa9e4066Sahrens devid_str_free(minor); 111fa9e4066Sahrens devid_free(devid); 112fa9e4066Sahrens } 113c67d9675Seschrock (void) close(fd); 114fa9e4066Sahrens 115fa9e4066Sahrens return (ret); 116fa9e4066Sahrens } 117fa9e4066Sahrens 118fa9e4066Sahrens 119fa9e4066Sahrens /* 120fa9e4066Sahrens * Go through and fix up any path and/or devid information for the given vdev 121fa9e4066Sahrens * configuration. 122fa9e4066Sahrens */ 12399653d4eSeschrock static int 124fa9e4066Sahrens fix_paths(nvlist_t *nv, name_entry_t *names) 125fa9e4066Sahrens { 126fa9e4066Sahrens nvlist_t **child; 127fa9e4066Sahrens uint_t c, children; 128fa9e4066Sahrens uint64_t guid; 129c67d9675Seschrock name_entry_t *ne, *best; 130c67d9675Seschrock char *path, *devid; 131c67d9675Seschrock int matched; 132fa9e4066Sahrens 133fa9e4066Sahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 134fa9e4066Sahrens &child, &children) == 0) { 135fa9e4066Sahrens for (c = 0; c < children; c++) 13699653d4eSeschrock if (fix_paths(child[c], names) != 0) 13799653d4eSeschrock return (-1); 13899653d4eSeschrock return (0); 139fa9e4066Sahrens } 140fa9e4066Sahrens 141fa9e4066Sahrens /* 142fa9e4066Sahrens * This is a leaf (file or disk) vdev. In either case, go through 143fa9e4066Sahrens * the name list and see if we find a matching guid. If so, replace 144fa9e4066Sahrens * the path and see if we can calculate a new devid. 145c67d9675Seschrock * 146c67d9675Seschrock * There may be multiple names associated with a particular guid, in 147c67d9675Seschrock * which case we have overlapping slices or multiple paths to the same 148c67d9675Seschrock * disk. If this is the case, then we want to pick the path that is 149c67d9675Seschrock * the most similar to the original, where "most similar" is the number 150c67d9675Seschrock * of matching characters starting from the end of the path. This will 151c67d9675Seschrock * preserve slice numbers even if the disks have been reorganized, and 152c67d9675Seschrock * will also catch preferred disk names if multiple paths exist. 153fa9e4066Sahrens */ 154fa9e4066Sahrens verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &guid) == 0); 155c67d9675Seschrock if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) 156c67d9675Seschrock path = NULL; 157c67d9675Seschrock 158c67d9675Seschrock matched = 0; 159c67d9675Seschrock best = NULL; 160c67d9675Seschrock for (ne = names; ne != NULL; ne = ne->ne_next) { 161c67d9675Seschrock if (ne->ne_guid == guid) { 162c67d9675Seschrock const char *src, *dst; 163c67d9675Seschrock int count; 164c67d9675Seschrock 165c67d9675Seschrock if (path == NULL) { 166c67d9675Seschrock best = ne; 167c67d9675Seschrock break; 168c67d9675Seschrock } 169c67d9675Seschrock 170c67d9675Seschrock src = ne->ne_name + strlen(ne->ne_name) - 1; 171c67d9675Seschrock dst = path + strlen(path) - 1; 172c67d9675Seschrock for (count = 0; src >= ne->ne_name && dst >= path; 173c67d9675Seschrock src--, dst--, count++) 174c67d9675Seschrock if (*src != *dst) 175c67d9675Seschrock break; 176c67d9675Seschrock 177c67d9675Seschrock /* 178c67d9675Seschrock * At this point, 'count' is the number of characters 179c67d9675Seschrock * matched from the end. 180c67d9675Seschrock */ 181c67d9675Seschrock if (count > matched || best == NULL) { 182c67d9675Seschrock best = ne; 183c67d9675Seschrock matched = count; 184c67d9675Seschrock } 185c67d9675Seschrock } 186c67d9675Seschrock } 187fa9e4066Sahrens 188c67d9675Seschrock if (best == NULL) 18999653d4eSeschrock return (0); 190fa9e4066Sahrens 19199653d4eSeschrock if (nvlist_add_string(nv, ZPOOL_CONFIG_PATH, best->ne_name) != 0) 19299653d4eSeschrock return (-1); 193fa9e4066Sahrens 194c67d9675Seschrock if ((devid = get_devid(best->ne_name)) == NULL) { 195fa9e4066Sahrens (void) nvlist_remove_all(nv, ZPOOL_CONFIG_DEVID); 196fa9e4066Sahrens } else { 19799653d4eSeschrock if (nvlist_add_string(nv, ZPOOL_CONFIG_DEVID, devid) != 0) 19899653d4eSeschrock return (-1); 199fa9e4066Sahrens devid_str_free(devid); 200fa9e4066Sahrens } 20199653d4eSeschrock 20299653d4eSeschrock return (0); 203fa9e4066Sahrens } 204fa9e4066Sahrens 205fa9e4066Sahrens /* 206fa9e4066Sahrens * Add the given configuration to the list of known devices. 207fa9e4066Sahrens */ 20899653d4eSeschrock static int 20999653d4eSeschrock add_config(libzfs_handle_t *hdl, pool_list_t *pl, const char *path, 21099653d4eSeschrock nvlist_t *config) 211fa9e4066Sahrens { 21299653d4eSeschrock uint64_t pool_guid, vdev_guid, top_guid, txg, state; 213fa9e4066Sahrens pool_entry_t *pe; 214fa9e4066Sahrens vdev_entry_t *ve; 215fa9e4066Sahrens config_entry_t *ce; 216fa9e4066Sahrens name_entry_t *ne; 217fa9e4066Sahrens 21899653d4eSeschrock /* 219fa94a07fSbrendan * If this is a hot spare not currently in use or level 2 cache 220fa94a07fSbrendan * device, add it to the list of names to translate, but don't do 221fa94a07fSbrendan * anything else. 22299653d4eSeschrock */ 22399653d4eSeschrock if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 224fa94a07fSbrendan &state) == 0 && 225fa94a07fSbrendan (state == POOL_STATE_SPARE || state == POOL_STATE_L2CACHE) && 22699653d4eSeschrock nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, &vdev_guid) == 0) { 22799653d4eSeschrock if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 228ccae0b50Seschrock return (-1); 22999653d4eSeschrock 23099653d4eSeschrock if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 23199653d4eSeschrock free(ne); 23299653d4eSeschrock return (-1); 23399653d4eSeschrock } 23499653d4eSeschrock ne->ne_guid = vdev_guid; 23599653d4eSeschrock ne->ne_next = pl->names; 23699653d4eSeschrock pl->names = ne; 23799653d4eSeschrock return (0); 23899653d4eSeschrock } 23999653d4eSeschrock 240fa9e4066Sahrens /* 241fa9e4066Sahrens * If we have a valid config but cannot read any of these fields, then 242fa9e4066Sahrens * it means we have a half-initialized label. In vdev_label_init() 243fa9e4066Sahrens * we write a label with txg == 0 so that we can identify the device 244fa9e4066Sahrens * in case the user refers to the same disk later on. If we fail to 245fa9e4066Sahrens * create the pool, we'll be left with a label in this state 246fa9e4066Sahrens * which should not be considered part of a valid pool. 247fa9e4066Sahrens */ 248fa9e4066Sahrens if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 249fa9e4066Sahrens &pool_guid) != 0 || 250fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 251fa9e4066Sahrens &vdev_guid) != 0 || 252fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_TOP_GUID, 253fa9e4066Sahrens &top_guid) != 0 || 254fa9e4066Sahrens nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_TXG, 255fa9e4066Sahrens &txg) != 0 || txg == 0) { 256fa9e4066Sahrens nvlist_free(config); 25799653d4eSeschrock return (0); 258fa9e4066Sahrens } 259fa9e4066Sahrens 260fa9e4066Sahrens /* 261fa9e4066Sahrens * First, see if we know about this pool. If not, then add it to the 262fa9e4066Sahrens * list of known pools. 263fa9e4066Sahrens */ 264fa9e4066Sahrens for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 265fa9e4066Sahrens if (pe->pe_guid == pool_guid) 266fa9e4066Sahrens break; 267fa9e4066Sahrens } 268fa9e4066Sahrens 269fa9e4066Sahrens if (pe == NULL) { 27099653d4eSeschrock if ((pe = zfs_alloc(hdl, sizeof (pool_entry_t))) == NULL) { 27199653d4eSeschrock nvlist_free(config); 27299653d4eSeschrock return (-1); 27399653d4eSeschrock } 274fa9e4066Sahrens pe->pe_guid = pool_guid; 275fa9e4066Sahrens pe->pe_next = pl->pools; 276fa9e4066Sahrens pl->pools = pe; 277fa9e4066Sahrens } 278fa9e4066Sahrens 279fa9e4066Sahrens /* 280fa9e4066Sahrens * Second, see if we know about this toplevel vdev. Add it if its 281fa9e4066Sahrens * missing. 282fa9e4066Sahrens */ 283fa9e4066Sahrens for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 284fa9e4066Sahrens if (ve->ve_guid == top_guid) 285fa9e4066Sahrens break; 286fa9e4066Sahrens } 287fa9e4066Sahrens 288fa9e4066Sahrens if (ve == NULL) { 28999653d4eSeschrock if ((ve = zfs_alloc(hdl, sizeof (vdev_entry_t))) == NULL) { 29099653d4eSeschrock nvlist_free(config); 29199653d4eSeschrock return (-1); 29299653d4eSeschrock } 293fa9e4066Sahrens ve->ve_guid = top_guid; 294fa9e4066Sahrens ve->ve_next = pe->pe_vdevs; 295fa9e4066Sahrens pe->pe_vdevs = ve; 296fa9e4066Sahrens } 297fa9e4066Sahrens 298fa9e4066Sahrens /* 299fa9e4066Sahrens * Third, see if we have a config with a matching transaction group. If 300fa9e4066Sahrens * so, then we do nothing. Otherwise, add it to the list of known 301fa9e4066Sahrens * configs. 302fa9e4066Sahrens */ 303fa9e4066Sahrens for (ce = ve->ve_configs; ce != NULL; ce = ce->ce_next) { 304fa9e4066Sahrens if (ce->ce_txg == txg) 305fa9e4066Sahrens break; 306fa9e4066Sahrens } 307fa9e4066Sahrens 308fa9e4066Sahrens if (ce == NULL) { 30999653d4eSeschrock if ((ce = zfs_alloc(hdl, sizeof (config_entry_t))) == NULL) { 31099653d4eSeschrock nvlist_free(config); 31199653d4eSeschrock return (-1); 31299653d4eSeschrock } 313fa9e4066Sahrens ce->ce_txg = txg; 314fa9e4066Sahrens ce->ce_config = config; 315fa9e4066Sahrens ce->ce_next = ve->ve_configs; 316fa9e4066Sahrens ve->ve_configs = ce; 317fa9e4066Sahrens } else { 318fa9e4066Sahrens nvlist_free(config); 319fa9e4066Sahrens } 320fa9e4066Sahrens 321fa9e4066Sahrens /* 322fa9e4066Sahrens * At this point we've successfully added our config to the list of 323fa9e4066Sahrens * known configs. The last thing to do is add the vdev guid -> path 324fa9e4066Sahrens * mappings so that we can fix up the configuration as necessary before 325fa9e4066Sahrens * doing the import. 326fa9e4066Sahrens */ 32799653d4eSeschrock if ((ne = zfs_alloc(hdl, sizeof (name_entry_t))) == NULL) 32899653d4eSeschrock return (-1); 32999653d4eSeschrock 33099653d4eSeschrock if ((ne->ne_name = zfs_strdup(hdl, path)) == NULL) { 33199653d4eSeschrock free(ne); 33299653d4eSeschrock return (-1); 33399653d4eSeschrock } 334fa9e4066Sahrens 335fa9e4066Sahrens ne->ne_guid = vdev_guid; 336fa9e4066Sahrens ne->ne_next = pl->names; 337fa9e4066Sahrens pl->names = ne; 33899653d4eSeschrock 33999653d4eSeschrock return (0); 340fa9e4066Sahrens } 341fa9e4066Sahrens 342eaca9bbdSeschrock /* 343eaca9bbdSeschrock * Returns true if the named pool matches the given GUID. 344eaca9bbdSeschrock */ 34594de1d4cSeschrock static int 34694de1d4cSeschrock pool_active(libzfs_handle_t *hdl, const char *name, uint64_t guid, 34794de1d4cSeschrock boolean_t *isactive) 348eaca9bbdSeschrock { 349eaca9bbdSeschrock zpool_handle_t *zhp; 350eaca9bbdSeschrock uint64_t theguid; 351eaca9bbdSeschrock 35294de1d4cSeschrock if (zpool_open_silent(hdl, name, &zhp) != 0) 35394de1d4cSeschrock return (-1); 35494de1d4cSeschrock 35594de1d4cSeschrock if (zhp == NULL) { 35694de1d4cSeschrock *isactive = B_FALSE; 35794de1d4cSeschrock return (0); 35894de1d4cSeschrock } 359eaca9bbdSeschrock 360eaca9bbdSeschrock verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_POOL_GUID, 361eaca9bbdSeschrock &theguid) == 0); 362eaca9bbdSeschrock 363eaca9bbdSeschrock zpool_close(zhp); 364eaca9bbdSeschrock 36594de1d4cSeschrock *isactive = (theguid == guid); 36694de1d4cSeschrock return (0); 367eaca9bbdSeschrock } 368eaca9bbdSeschrock 3692f8aaab3Seschrock static nvlist_t * 3702f8aaab3Seschrock refresh_config(libzfs_handle_t *hdl, nvlist_t *config) 3712f8aaab3Seschrock { 3722f8aaab3Seschrock nvlist_t *nvl; 3732f8aaab3Seschrock zfs_cmd_t zc = { 0 }; 3742f8aaab3Seschrock int err; 3752f8aaab3Seschrock 3762f8aaab3Seschrock if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) 3772f8aaab3Seschrock return (NULL); 3782f8aaab3Seschrock 3792f8aaab3Seschrock if (zcmd_alloc_dst_nvlist(hdl, &zc, 3802f8aaab3Seschrock zc.zc_nvlist_conf_size * 2) != 0) { 3812f8aaab3Seschrock zcmd_free_nvlists(&zc); 3822f8aaab3Seschrock return (NULL); 3832f8aaab3Seschrock } 3842f8aaab3Seschrock 3852f8aaab3Seschrock while ((err = ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_TRYIMPORT, 3862f8aaab3Seschrock &zc)) != 0 && errno == ENOMEM) { 3872f8aaab3Seschrock if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { 3882f8aaab3Seschrock zcmd_free_nvlists(&zc); 3892f8aaab3Seschrock return (NULL); 3902f8aaab3Seschrock } 3912f8aaab3Seschrock } 3922f8aaab3Seschrock 3932f8aaab3Seschrock if (err) { 3942f8aaab3Seschrock zcmd_free_nvlists(&zc); 3952f8aaab3Seschrock return (NULL); 3962f8aaab3Seschrock } 3972f8aaab3Seschrock 3982f8aaab3Seschrock if (zcmd_read_dst_nvlist(hdl, &zc, &nvl) != 0) { 3992f8aaab3Seschrock zcmd_free_nvlists(&zc); 4002f8aaab3Seschrock return (NULL); 4012f8aaab3Seschrock } 4022f8aaab3Seschrock 4032f8aaab3Seschrock zcmd_free_nvlists(&zc); 4042f8aaab3Seschrock return (nvl); 4052f8aaab3Seschrock } 4062f8aaab3Seschrock 40788ecc943SGeorge Wilson /* 40888ecc943SGeorge Wilson * Determine if the vdev id is a hole in the namespace. 40988ecc943SGeorge Wilson */ 41088ecc943SGeorge Wilson boolean_t 41188ecc943SGeorge Wilson vdev_is_hole(uint64_t *hole_array, uint_t holes, uint_t id) 41288ecc943SGeorge Wilson { 41388ecc943SGeorge Wilson for (int c = 0; c < holes; c++) { 41488ecc943SGeorge Wilson 41588ecc943SGeorge Wilson /* Top-level is a hole */ 41688ecc943SGeorge Wilson if (hole_array[c] == id) 41788ecc943SGeorge Wilson return (B_TRUE); 41888ecc943SGeorge Wilson } 41988ecc943SGeorge Wilson return (B_FALSE); 42088ecc943SGeorge Wilson } 42188ecc943SGeorge Wilson 422fa9e4066Sahrens /* 423fa9e4066Sahrens * Convert our list of pools into the definitive set of configurations. We 424fa9e4066Sahrens * start by picking the best config for each toplevel vdev. Once that's done, 425fa9e4066Sahrens * we assemble the toplevel vdevs into a full config for the pool. We make a 426fa9e4066Sahrens * pass to fix up any incorrect paths, and then add it to the main list to 427fa9e4066Sahrens * return to the user. 428fa9e4066Sahrens */ 429fa9e4066Sahrens static nvlist_t * 4303a57275aSck get_configs(libzfs_handle_t *hdl, pool_list_t *pl, boolean_t active_ok) 431fa9e4066Sahrens { 43299653d4eSeschrock pool_entry_t *pe; 43399653d4eSeschrock vdev_entry_t *ve; 43499653d4eSeschrock config_entry_t *ce; 43599653d4eSeschrock nvlist_t *ret = NULL, *config = NULL, *tmp, *nvtop, *nvroot; 436fa94a07fSbrendan nvlist_t **spares, **l2cache; 437fa94a07fSbrendan uint_t i, nspares, nl2cache; 43899653d4eSeschrock boolean_t config_seen; 439fa9e4066Sahrens uint64_t best_txg; 44095173954Sek char *name, *hostname; 44199653d4eSeschrock uint64_t version, guid; 44299653d4eSeschrock uint_t children = 0; 44399653d4eSeschrock nvlist_t **child = NULL; 44488ecc943SGeorge Wilson uint_t holes; 44588ecc943SGeorge Wilson uint64_t *hole_array, max_id; 44699653d4eSeschrock uint_t c; 44794de1d4cSeschrock boolean_t isactive; 44895173954Sek uint64_t hostid; 4492f8aaab3Seschrock nvlist_t *nvl; 45024e697d4Sck boolean_t found_one = B_FALSE; 45188ecc943SGeorge Wilson boolean_t valid_top_config = B_FALSE; 452fa9e4066Sahrens 45399653d4eSeschrock if (nvlist_alloc(&ret, 0, 0) != 0) 45499653d4eSeschrock goto nomem; 455fa9e4066Sahrens 45699653d4eSeschrock for (pe = pl->pools; pe != NULL; pe = pe->pe_next) { 45788ecc943SGeorge Wilson uint64_t id, max_txg = 0; 458fa9e4066Sahrens 45999653d4eSeschrock if (nvlist_alloc(&config, NV_UNIQUE_NAME, 0) != 0) 46099653d4eSeschrock goto nomem; 46199653d4eSeschrock config_seen = B_FALSE; 462fa9e4066Sahrens 463fa9e4066Sahrens /* 464fa9e4066Sahrens * Iterate over all toplevel vdevs. Grab the pool configuration 465fa9e4066Sahrens * from the first one we find, and then go through the rest and 466fa9e4066Sahrens * add them as necessary to the 'vdevs' member of the config. 467fa9e4066Sahrens */ 46899653d4eSeschrock for (ve = pe->pe_vdevs; ve != NULL; ve = ve->ve_next) { 469fa9e4066Sahrens 470fa9e4066Sahrens /* 471fa9e4066Sahrens * Determine the best configuration for this vdev by 472fa9e4066Sahrens * selecting the config with the latest transaction 473fa9e4066Sahrens * group. 474fa9e4066Sahrens */ 475fa9e4066Sahrens best_txg = 0; 476fa9e4066Sahrens for (ce = ve->ve_configs; ce != NULL; 477fa9e4066Sahrens ce = ce->ce_next) { 478fa9e4066Sahrens 47999653d4eSeschrock if (ce->ce_txg > best_txg) { 480fa9e4066Sahrens tmp = ce->ce_config; 48199653d4eSeschrock best_txg = ce->ce_txg; 48299653d4eSeschrock } 483fa9e4066Sahrens } 484fa9e4066Sahrens 48588ecc943SGeorge Wilson /* 48688ecc943SGeorge Wilson * We rely on the fact that the max txg for the 48788ecc943SGeorge Wilson * pool will contain the most up-to-date information 48888ecc943SGeorge Wilson * about the valid top-levels in the vdev namespace. 48988ecc943SGeorge Wilson */ 49088ecc943SGeorge Wilson if (best_txg > max_txg) { 49188ecc943SGeorge Wilson (void) nvlist_remove(config, 49288ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, 49388ecc943SGeorge Wilson DATA_TYPE_UINT64); 49488ecc943SGeorge Wilson (void) nvlist_remove(config, 49588ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, 49688ecc943SGeorge Wilson DATA_TYPE_UINT64_ARRAY); 49788ecc943SGeorge Wilson 49888ecc943SGeorge Wilson max_txg = best_txg; 49988ecc943SGeorge Wilson hole_array = NULL; 50088ecc943SGeorge Wilson holes = 0; 50188ecc943SGeorge Wilson max_id = 0; 50288ecc943SGeorge Wilson valid_top_config = B_FALSE; 50388ecc943SGeorge Wilson 50488ecc943SGeorge Wilson if (nvlist_lookup_uint64(tmp, 50588ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, &max_id) == 0) { 50688ecc943SGeorge Wilson verify(nvlist_add_uint64(config, 50788ecc943SGeorge Wilson ZPOOL_CONFIG_VDEV_CHILDREN, 50888ecc943SGeorge Wilson max_id) == 0); 50988ecc943SGeorge Wilson valid_top_config = B_TRUE; 51088ecc943SGeorge Wilson } 51188ecc943SGeorge Wilson 51288ecc943SGeorge Wilson if (nvlist_lookup_uint64_array(tmp, 51388ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, &hole_array, 51488ecc943SGeorge Wilson &holes) == 0) { 51588ecc943SGeorge Wilson verify(nvlist_add_uint64_array(config, 51688ecc943SGeorge Wilson ZPOOL_CONFIG_HOLE_ARRAY, 51788ecc943SGeorge Wilson hole_array, holes) == 0); 51888ecc943SGeorge Wilson } 51988ecc943SGeorge Wilson } 52088ecc943SGeorge Wilson 521fa9e4066Sahrens if (!config_seen) { 522fa9e4066Sahrens /* 523fa9e4066Sahrens * Copy the relevant pieces of data to the pool 524fa9e4066Sahrens * configuration: 525fa9e4066Sahrens * 52699653d4eSeschrock * version 527fa9e4066Sahrens * pool guid 528fa9e4066Sahrens * name 529fa9e4066Sahrens * pool state 53095173954Sek * hostid (if available) 53195173954Sek * hostname (if available) 532fa9e4066Sahrens */ 533fa9e4066Sahrens uint64_t state; 534fa9e4066Sahrens 53599653d4eSeschrock verify(nvlist_lookup_uint64(tmp, 53699653d4eSeschrock ZPOOL_CONFIG_VERSION, &version) == 0); 53799653d4eSeschrock if (nvlist_add_uint64(config, 53899653d4eSeschrock ZPOOL_CONFIG_VERSION, version) != 0) 53999653d4eSeschrock goto nomem; 540fa9e4066Sahrens verify(nvlist_lookup_uint64(tmp, 541fa9e4066Sahrens ZPOOL_CONFIG_POOL_GUID, &guid) == 0); 54299653d4eSeschrock if (nvlist_add_uint64(config, 54399653d4eSeschrock ZPOOL_CONFIG_POOL_GUID, guid) != 0) 54499653d4eSeschrock goto nomem; 545fa9e4066Sahrens verify(nvlist_lookup_string(tmp, 546fa9e4066Sahrens ZPOOL_CONFIG_POOL_NAME, &name) == 0); 54799653d4eSeschrock if (nvlist_add_string(config, 54899653d4eSeschrock ZPOOL_CONFIG_POOL_NAME, name) != 0) 54999653d4eSeschrock goto nomem; 550fa9e4066Sahrens verify(nvlist_lookup_uint64(tmp, 551fa9e4066Sahrens ZPOOL_CONFIG_POOL_STATE, &state) == 0); 55299653d4eSeschrock if (nvlist_add_uint64(config, 55399653d4eSeschrock ZPOOL_CONFIG_POOL_STATE, state) != 0) 55499653d4eSeschrock goto nomem; 55595173954Sek hostid = 0; 55695173954Sek if (nvlist_lookup_uint64(tmp, 55795173954Sek ZPOOL_CONFIG_HOSTID, &hostid) == 0) { 55895173954Sek if (nvlist_add_uint64(config, 55995173954Sek ZPOOL_CONFIG_HOSTID, hostid) != 0) 56095173954Sek goto nomem; 56195173954Sek verify(nvlist_lookup_string(tmp, 56295173954Sek ZPOOL_CONFIG_HOSTNAME, 56395173954Sek &hostname) == 0); 56495173954Sek if (nvlist_add_string(config, 56595173954Sek ZPOOL_CONFIG_HOSTNAME, 56695173954Sek hostname) != 0) 56795173954Sek goto nomem; 56895173954Sek } 569fa9e4066Sahrens 57099653d4eSeschrock config_seen = B_TRUE; 571fa9e4066Sahrens } 572fa9e4066Sahrens 573fa9e4066Sahrens /* 574fa9e4066Sahrens * Add this top-level vdev to the child array. 575fa9e4066Sahrens */ 576fa9e4066Sahrens verify(nvlist_lookup_nvlist(tmp, 577fa9e4066Sahrens ZPOOL_CONFIG_VDEV_TREE, &nvtop) == 0); 578fa9e4066Sahrens verify(nvlist_lookup_uint64(nvtop, ZPOOL_CONFIG_ID, 579fa9e4066Sahrens &id) == 0); 58088ecc943SGeorge Wilson 581fa9e4066Sahrens if (id >= children) { 582fa9e4066Sahrens nvlist_t **newchild; 583fa9e4066Sahrens 58499653d4eSeschrock newchild = zfs_alloc(hdl, (id + 1) * 585fa9e4066Sahrens sizeof (nvlist_t *)); 58699653d4eSeschrock if (newchild == NULL) 58799653d4eSeschrock goto nomem; 588fa9e4066Sahrens 589fa9e4066Sahrens for (c = 0; c < children; c++) 590fa9e4066Sahrens newchild[c] = child[c]; 591fa9e4066Sahrens 592fa9e4066Sahrens free(child); 593fa9e4066Sahrens child = newchild; 594fa9e4066Sahrens children = id + 1; 595fa9e4066Sahrens } 59699653d4eSeschrock if (nvlist_dup(nvtop, &child[id], 0) != 0) 59799653d4eSeschrock goto nomem; 598fa9e4066Sahrens 599fa9e4066Sahrens } 600fa9e4066Sahrens 60188ecc943SGeorge Wilson /* 60288ecc943SGeorge Wilson * If we have information about all the top-levels then 60388ecc943SGeorge Wilson * clean up the nvlist which we've constructed. This 60488ecc943SGeorge Wilson * means removing any extraneous devices that are 60588ecc943SGeorge Wilson * beyond the valid range or adding devices to the end 60688ecc943SGeorge Wilson * of our array which appear to be missing. 60788ecc943SGeorge Wilson */ 60888ecc943SGeorge Wilson if (valid_top_config) { 60988ecc943SGeorge Wilson if (max_id < children) { 61088ecc943SGeorge Wilson for (c = max_id; c < children; c++) 61188ecc943SGeorge Wilson nvlist_free(child[c]); 61288ecc943SGeorge Wilson children = max_id; 61388ecc943SGeorge Wilson } else if (max_id > children) { 61488ecc943SGeorge Wilson nvlist_t **newchild; 61588ecc943SGeorge Wilson 61688ecc943SGeorge Wilson newchild = zfs_alloc(hdl, (max_id) * 61788ecc943SGeorge Wilson sizeof (nvlist_t *)); 61888ecc943SGeorge Wilson if (newchild == NULL) 61988ecc943SGeorge Wilson goto nomem; 62088ecc943SGeorge Wilson 62188ecc943SGeorge Wilson for (c = 0; c < children; c++) 62288ecc943SGeorge Wilson newchild[c] = child[c]; 62388ecc943SGeorge Wilson 62488ecc943SGeorge Wilson free(child); 62588ecc943SGeorge Wilson child = newchild; 62688ecc943SGeorge Wilson children = max_id; 62788ecc943SGeorge Wilson } 62888ecc943SGeorge Wilson } 62988ecc943SGeorge Wilson 630fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 631fa9e4066Sahrens &guid) == 0); 632fa9e4066Sahrens 63388ecc943SGeorge Wilson /* 63488ecc943SGeorge Wilson * The vdev namespace may contain holes as a result of 63588ecc943SGeorge Wilson * device removal. We must add them back into the vdev 63688ecc943SGeorge Wilson * tree before we process any missing devices. 63788ecc943SGeorge Wilson */ 63888ecc943SGeorge Wilson if (holes > 0) { 63988ecc943SGeorge Wilson ASSERT(valid_top_config); 64088ecc943SGeorge Wilson 64188ecc943SGeorge Wilson for (c = 0; c < children; c++) { 64288ecc943SGeorge Wilson nvlist_t *holey; 64388ecc943SGeorge Wilson 64488ecc943SGeorge Wilson if (child[c] != NULL || 64588ecc943SGeorge Wilson !vdev_is_hole(hole_array, holes, c)) 64688ecc943SGeorge Wilson continue; 64788ecc943SGeorge Wilson 64888ecc943SGeorge Wilson if (nvlist_alloc(&holey, NV_UNIQUE_NAME, 64988ecc943SGeorge Wilson 0) != 0) 65088ecc943SGeorge Wilson goto nomem; 65188ecc943SGeorge Wilson 65288ecc943SGeorge Wilson /* 65388ecc943SGeorge Wilson * Holes in the namespace are treated as 65488ecc943SGeorge Wilson * "hole" top-level vdevs and have a 65588ecc943SGeorge Wilson * special flag set on them. 65688ecc943SGeorge Wilson */ 65788ecc943SGeorge Wilson if (nvlist_add_string(holey, 65888ecc943SGeorge Wilson ZPOOL_CONFIG_TYPE, 65988ecc943SGeorge Wilson VDEV_TYPE_HOLE) != 0 || 66088ecc943SGeorge Wilson nvlist_add_uint64(holey, 66188ecc943SGeorge Wilson ZPOOL_CONFIG_ID, c) != 0 || 66288ecc943SGeorge Wilson nvlist_add_uint64(holey, 66388ecc943SGeorge Wilson ZPOOL_CONFIG_GUID, 0ULL) != 0) 66488ecc943SGeorge Wilson goto nomem; 66588ecc943SGeorge Wilson child[c] = holey; 66688ecc943SGeorge Wilson } 66788ecc943SGeorge Wilson } 66888ecc943SGeorge Wilson 669fa9e4066Sahrens /* 670fa9e4066Sahrens * Look for any missing top-level vdevs. If this is the case, 671fa9e4066Sahrens * create a faked up 'missing' vdev as a placeholder. We cannot 672fa9e4066Sahrens * simply compress the child array, because the kernel performs 673fa9e4066Sahrens * certain checks to make sure the vdev IDs match their location 674fa9e4066Sahrens * in the configuration. 675fa9e4066Sahrens */ 67688ecc943SGeorge Wilson for (c = 0; c < children; c++) { 677fa9e4066Sahrens if (child[c] == NULL) { 678fa9e4066Sahrens nvlist_t *missing; 67999653d4eSeschrock if (nvlist_alloc(&missing, NV_UNIQUE_NAME, 68099653d4eSeschrock 0) != 0) 68199653d4eSeschrock goto nomem; 68299653d4eSeschrock if (nvlist_add_string(missing, 68399653d4eSeschrock ZPOOL_CONFIG_TYPE, 68499653d4eSeschrock VDEV_TYPE_MISSING) != 0 || 68599653d4eSeschrock nvlist_add_uint64(missing, 68699653d4eSeschrock ZPOOL_CONFIG_ID, c) != 0 || 68799653d4eSeschrock nvlist_add_uint64(missing, 68899653d4eSeschrock ZPOOL_CONFIG_GUID, 0ULL) != 0) { 68999653d4eSeschrock nvlist_free(missing); 69099653d4eSeschrock goto nomem; 69199653d4eSeschrock } 692fa9e4066Sahrens child[c] = missing; 693fa9e4066Sahrens } 69488ecc943SGeorge Wilson } 695fa9e4066Sahrens 696fa9e4066Sahrens /* 697fa9e4066Sahrens * Put all of this pool's top-level vdevs into a root vdev. 698fa9e4066Sahrens */ 69999653d4eSeschrock if (nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) != 0) 70099653d4eSeschrock goto nomem; 70199653d4eSeschrock if (nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, 70299653d4eSeschrock VDEV_TYPE_ROOT) != 0 || 70399653d4eSeschrock nvlist_add_uint64(nvroot, ZPOOL_CONFIG_ID, 0ULL) != 0 || 70499653d4eSeschrock nvlist_add_uint64(nvroot, ZPOOL_CONFIG_GUID, guid) != 0 || 70599653d4eSeschrock nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, 70699653d4eSeschrock child, children) != 0) { 70799653d4eSeschrock nvlist_free(nvroot); 70899653d4eSeschrock goto nomem; 70999653d4eSeschrock } 710fa9e4066Sahrens 711fa9e4066Sahrens for (c = 0; c < children; c++) 712fa9e4066Sahrens nvlist_free(child[c]); 713fa9e4066Sahrens free(child); 71499653d4eSeschrock children = 0; 71599653d4eSeschrock child = NULL; 716fa9e4066Sahrens 717fa9e4066Sahrens /* 718fa9e4066Sahrens * Go through and fix up any paths and/or devids based on our 719fa9e4066Sahrens * known list of vdev GUID -> path mappings. 720fa9e4066Sahrens */ 72199653d4eSeschrock if (fix_paths(nvroot, pl->names) != 0) { 72299653d4eSeschrock nvlist_free(nvroot); 72399653d4eSeschrock goto nomem; 72499653d4eSeschrock } 725fa9e4066Sahrens 726fa9e4066Sahrens /* 727fa9e4066Sahrens * Add the root vdev to this pool's configuration. 728fa9e4066Sahrens */ 72999653d4eSeschrock if (nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 73099653d4eSeschrock nvroot) != 0) { 73199653d4eSeschrock nvlist_free(nvroot); 73299653d4eSeschrock goto nomem; 73399653d4eSeschrock } 734fa9e4066Sahrens nvlist_free(nvroot); 735fa9e4066Sahrens 7363a57275aSck /* 7373a57275aSck * zdb uses this path to report on active pools that were 7383a57275aSck * imported or created using -R. 7393a57275aSck */ 7403a57275aSck if (active_ok) 7413a57275aSck goto add_pool; 7423a57275aSck 743fa9e4066Sahrens /* 744fa9e4066Sahrens * Determine if this pool is currently active, in which case we 745fa9e4066Sahrens * can't actually import it. 746fa9e4066Sahrens */ 747fa9e4066Sahrens verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 748fa9e4066Sahrens &name) == 0); 749fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 750fa9e4066Sahrens &guid) == 0); 751fa9e4066Sahrens 75294de1d4cSeschrock if (pool_active(hdl, name, guid, &isactive) != 0) 75394de1d4cSeschrock goto error; 75494de1d4cSeschrock 7550192a278Seschrock if (isactive) { 756fa9e4066Sahrens nvlist_free(config); 75799653d4eSeschrock config = NULL; 758fa9e4066Sahrens continue; 759fa9e4066Sahrens } 760fa9e4066Sahrens 76188ecc943SGeorge Wilson if ((nvl = refresh_config(hdl, config)) == NULL) { 76288ecc943SGeorge Wilson nvlist_free(config); 76388ecc943SGeorge Wilson config = NULL; 76488ecc943SGeorge Wilson continue; 76588ecc943SGeorge Wilson } 766fa9e4066Sahrens 767fa9e4066Sahrens nvlist_free(config); 7682f8aaab3Seschrock config = nvl; 769fa9e4066Sahrens 77099653d4eSeschrock /* 77199653d4eSeschrock * Go through and update the paths for spares, now that we have 77299653d4eSeschrock * them. 77399653d4eSeschrock */ 77499653d4eSeschrock verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, 77599653d4eSeschrock &nvroot) == 0); 77699653d4eSeschrock if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, 77799653d4eSeschrock &spares, &nspares) == 0) { 77899653d4eSeschrock for (i = 0; i < nspares; i++) { 77999653d4eSeschrock if (fix_paths(spares[i], pl->names) != 0) 78099653d4eSeschrock goto nomem; 78199653d4eSeschrock } 78299653d4eSeschrock } 78399653d4eSeschrock 784fa94a07fSbrendan /* 785fa94a07fSbrendan * Update the paths for l2cache devices. 786fa94a07fSbrendan */ 787fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, 788fa94a07fSbrendan &l2cache, &nl2cache) == 0) { 789fa94a07fSbrendan for (i = 0; i < nl2cache; i++) { 790fa94a07fSbrendan if (fix_paths(l2cache[i], pl->names) != 0) 791fa94a07fSbrendan goto nomem; 792fa94a07fSbrendan } 793fa94a07fSbrendan } 794fa94a07fSbrendan 79595173954Sek /* 79695173954Sek * Restore the original information read from the actual label. 79795173954Sek */ 79895173954Sek (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTID, 79995173954Sek DATA_TYPE_UINT64); 80095173954Sek (void) nvlist_remove(config, ZPOOL_CONFIG_HOSTNAME, 80195173954Sek DATA_TYPE_STRING); 80295173954Sek if (hostid != 0) { 80395173954Sek verify(nvlist_add_uint64(config, ZPOOL_CONFIG_HOSTID, 80495173954Sek hostid) == 0); 80595173954Sek verify(nvlist_add_string(config, ZPOOL_CONFIG_HOSTNAME, 80695173954Sek hostname) == 0); 80795173954Sek } 80895173954Sek 8093a57275aSck add_pool: 810fa9e4066Sahrens /* 811fa9e4066Sahrens * Add this pool to the list of configs. 812fa9e4066Sahrens */ 813e9dbad6fSeschrock verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 814e9dbad6fSeschrock &name) == 0); 81599653d4eSeschrock if (nvlist_add_nvlist(ret, name, config) != 0) 81699653d4eSeschrock goto nomem; 817fa9e4066Sahrens 81824e697d4Sck found_one = B_TRUE; 819fa9e4066Sahrens nvlist_free(config); 82099653d4eSeschrock config = NULL; 821fa9e4066Sahrens } 822fa9e4066Sahrens 82324e697d4Sck if (!found_one) { 82424e697d4Sck nvlist_free(ret); 82524e697d4Sck ret = NULL; 82624e697d4Sck } 82724e697d4Sck 828fa9e4066Sahrens return (ret); 82999653d4eSeschrock 83099653d4eSeschrock nomem: 83199653d4eSeschrock (void) no_memory(hdl); 83299653d4eSeschrock error: 83394de1d4cSeschrock nvlist_free(config); 83494de1d4cSeschrock nvlist_free(ret); 83599653d4eSeschrock for (c = 0; c < children; c++) 83699653d4eSeschrock nvlist_free(child[c]); 83794de1d4cSeschrock free(child); 83899653d4eSeschrock 83999653d4eSeschrock return (NULL); 840fa9e4066Sahrens } 841fa9e4066Sahrens 842fa9e4066Sahrens /* 843fa9e4066Sahrens * Return the offset of the given label. 844fa9e4066Sahrens */ 845fa9e4066Sahrens static uint64_t 846e7437265Sahrens label_offset(uint64_t size, int l) 847fa9e4066Sahrens { 848e7437265Sahrens ASSERT(P2PHASE_TYPED(size, sizeof (vdev_label_t), uint64_t) == 0); 849fa9e4066Sahrens return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 850fa9e4066Sahrens 0 : size - VDEV_LABELS * sizeof (vdev_label_t))); 851fa9e4066Sahrens } 852fa9e4066Sahrens 853fa9e4066Sahrens /* 854fa9e4066Sahrens * Given a file descriptor, read the label information and return an nvlist 855fa9e4066Sahrens * describing the configuration, if there is one. 856fa9e4066Sahrens */ 85799653d4eSeschrock int 85899653d4eSeschrock zpool_read_label(int fd, nvlist_t **config) 859fa9e4066Sahrens { 860fa9e4066Sahrens struct stat64 statbuf; 861fa9e4066Sahrens int l; 862fa9e4066Sahrens vdev_label_t *label; 863e7437265Sahrens uint64_t state, txg, size; 864fa9e4066Sahrens 86599653d4eSeschrock *config = NULL; 86699653d4eSeschrock 867fa9e4066Sahrens if (fstat64(fd, &statbuf) == -1) 86899653d4eSeschrock return (0); 869e7437265Sahrens size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 870fa9e4066Sahrens 87199653d4eSeschrock if ((label = malloc(sizeof (vdev_label_t))) == NULL) 87299653d4eSeschrock return (-1); 873fa9e4066Sahrens 874fa9e4066Sahrens for (l = 0; l < VDEV_LABELS; l++) { 875c5904d13Seschrock if (pread64(fd, label, sizeof (vdev_label_t), 876e7437265Sahrens label_offset(size, l)) != sizeof (vdev_label_t)) 877fa9e4066Sahrens continue; 878fa9e4066Sahrens 879fa9e4066Sahrens if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 88099653d4eSeschrock sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) 881fa9e4066Sahrens continue; 882fa9e4066Sahrens 88399653d4eSeschrock if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 884fa94a07fSbrendan &state) != 0 || state > POOL_STATE_L2CACHE) { 88599653d4eSeschrock nvlist_free(*config); 886fa9e4066Sahrens continue; 887fa9e4066Sahrens } 888fa9e4066Sahrens 889fa94a07fSbrendan if (state != POOL_STATE_SPARE && state != POOL_STATE_L2CACHE && 89099653d4eSeschrock (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 89199653d4eSeschrock &txg) != 0 || txg == 0)) { 89299653d4eSeschrock nvlist_free(*config); 893fa9e4066Sahrens continue; 894fa9e4066Sahrens } 895fa9e4066Sahrens 896fa9e4066Sahrens free(label); 89799653d4eSeschrock return (0); 898fa9e4066Sahrens } 899fa9e4066Sahrens 900fa9e4066Sahrens free(label); 90199653d4eSeschrock *config = NULL; 90299653d4eSeschrock return (0); 903fa9e4066Sahrens } 904fa9e4066Sahrens 9054f67d755SEric Taylor typedef struct rdsk_node { 9064f67d755SEric Taylor char *rn_name; 9074f67d755SEric Taylor int rn_dfd; 9084f67d755SEric Taylor libzfs_handle_t *rn_hdl; 9094f67d755SEric Taylor nvlist_t *rn_config; 9104f67d755SEric Taylor avl_tree_t *rn_avl; 9114f67d755SEric Taylor avl_node_t rn_node; 9124f67d755SEric Taylor boolean_t rn_nozpool; 9134f67d755SEric Taylor } rdsk_node_t; 9144f67d755SEric Taylor 9154f67d755SEric Taylor static int 9164f67d755SEric Taylor slice_cache_compare(const void *arg1, const void *arg2) 9174f67d755SEric Taylor { 9184f67d755SEric Taylor const char *nm1 = ((rdsk_node_t *)arg1)->rn_name; 9194f67d755SEric Taylor const char *nm2 = ((rdsk_node_t *)arg2)->rn_name; 9204f67d755SEric Taylor char *nm1slice, *nm2slice; 9214f67d755SEric Taylor int rv; 9224f67d755SEric Taylor 9234f67d755SEric Taylor /* 9244f67d755SEric Taylor * slices zero and two are the most likely to provide results, 9254f67d755SEric Taylor * so put those first 9264f67d755SEric Taylor */ 9274f67d755SEric Taylor nm1slice = strstr(nm1, "s0"); 9284f67d755SEric Taylor nm2slice = strstr(nm2, "s0"); 9294f67d755SEric Taylor if (nm1slice && !nm2slice) { 9304f67d755SEric Taylor return (-1); 9314f67d755SEric Taylor } 9324f67d755SEric Taylor if (!nm1slice && nm2slice) { 9334f67d755SEric Taylor return (1); 9344f67d755SEric Taylor } 9354f67d755SEric Taylor nm1slice = strstr(nm1, "s2"); 9364f67d755SEric Taylor nm2slice = strstr(nm2, "s2"); 9374f67d755SEric Taylor if (nm1slice && !nm2slice) { 9384f67d755SEric Taylor return (-1); 9394f67d755SEric Taylor } 9404f67d755SEric Taylor if (!nm1slice && nm2slice) { 9414f67d755SEric Taylor return (1); 9424f67d755SEric Taylor } 9434f67d755SEric Taylor 9444f67d755SEric Taylor rv = strcmp(nm1, nm2); 9454f67d755SEric Taylor if (rv == 0) 9464f67d755SEric Taylor return (0); 9474f67d755SEric Taylor return (rv > 0 ? 1 : -1); 9484f67d755SEric Taylor } 9494f67d755SEric Taylor 9504f67d755SEric Taylor static void 9514f67d755SEric Taylor check_one_slice(avl_tree_t *r, char *diskname, uint_t partno, 9524f67d755SEric Taylor diskaddr_t size, uint_t blksz) 9534f67d755SEric Taylor { 9544f67d755SEric Taylor rdsk_node_t tmpnode; 9554f67d755SEric Taylor rdsk_node_t *node; 9564f67d755SEric Taylor char sname[MAXNAMELEN]; 9574f67d755SEric Taylor 9584f67d755SEric Taylor tmpnode.rn_name = &sname[0]; 9594f67d755SEric Taylor (void) snprintf(tmpnode.rn_name, MAXNAMELEN, "%s%u", 9604f67d755SEric Taylor diskname, partno); 96116d2251eSEric Taylor /* 96216d2251eSEric Taylor * protect against division by zero for disk labels that 96316d2251eSEric Taylor * contain a bogus sector size 96416d2251eSEric Taylor */ 96516d2251eSEric Taylor if (blksz == 0) 96616d2251eSEric Taylor blksz = DEV_BSIZE; 9674f67d755SEric Taylor /* too small to contain a zpool? */ 9684f67d755SEric Taylor if ((size < (SPA_MINDEVSIZE / blksz)) && 9694f67d755SEric Taylor (node = avl_find(r, &tmpnode, NULL))) 9704f67d755SEric Taylor node->rn_nozpool = B_TRUE; 9714f67d755SEric Taylor } 9724f67d755SEric Taylor 9734f67d755SEric Taylor static void 9744f67d755SEric Taylor nozpool_all_slices(avl_tree_t *r, const char *sname) 9754f67d755SEric Taylor { 9764f67d755SEric Taylor char diskname[MAXNAMELEN]; 9774f67d755SEric Taylor char *ptr; 9784f67d755SEric Taylor int i; 9794f67d755SEric Taylor 9804f67d755SEric Taylor (void) strncpy(diskname, sname, MAXNAMELEN); 9814f67d755SEric Taylor if (((ptr = strrchr(diskname, 's')) == NULL) && 9824f67d755SEric Taylor ((ptr = strrchr(diskname, 'p')) == NULL)) 9834f67d755SEric Taylor return; 9844f67d755SEric Taylor ptr[0] = 's'; 9854f67d755SEric Taylor ptr[1] = '\0'; 9864f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 9874f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 9884f67d755SEric Taylor ptr[0] = 'p'; 9894f67d755SEric Taylor for (i = 0; i <= FD_NUMPART; i++) 9904f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 9914f67d755SEric Taylor } 9924f67d755SEric Taylor 9934f67d755SEric Taylor static void 9944f67d755SEric Taylor check_slices(avl_tree_t *r, int fd, const char *sname) 9954f67d755SEric Taylor { 9964f67d755SEric Taylor struct extvtoc vtoc; 9974f67d755SEric Taylor struct dk_gpt *gpt; 9984f67d755SEric Taylor char diskname[MAXNAMELEN]; 9994f67d755SEric Taylor char *ptr; 10004f67d755SEric Taylor int i; 10014f67d755SEric Taylor 10024f67d755SEric Taylor (void) strncpy(diskname, sname, MAXNAMELEN); 10034f67d755SEric Taylor if ((ptr = strrchr(diskname, 's')) == NULL || !isdigit(ptr[1])) 10044f67d755SEric Taylor return; 10054f67d755SEric Taylor ptr[1] = '\0'; 10064f67d755SEric Taylor 10074f67d755SEric Taylor if (read_extvtoc(fd, &vtoc) >= 0) { 10084f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 10094f67d755SEric Taylor check_one_slice(r, diskname, i, 10104f67d755SEric Taylor vtoc.v_part[i].p_size, vtoc.v_sectorsz); 10114f67d755SEric Taylor } else if (efi_alloc_and_read(fd, &gpt) >= 0) { 10124f67d755SEric Taylor /* 10134f67d755SEric Taylor * on x86 we'll still have leftover links that point 10144f67d755SEric Taylor * to slices s[9-15], so use NDKMAP instead 10154f67d755SEric Taylor */ 10164f67d755SEric Taylor for (i = 0; i < NDKMAP; i++) 10174f67d755SEric Taylor check_one_slice(r, diskname, i, 10184f67d755SEric Taylor gpt->efi_parts[i].p_size, gpt->efi_lbasize); 10194f67d755SEric Taylor /* nodes p[1-4] are never used with EFI labels */ 10204f67d755SEric Taylor ptr[0] = 'p'; 10214f67d755SEric Taylor for (i = 1; i <= FD_NUMPART; i++) 10224f67d755SEric Taylor check_one_slice(r, diskname, i, 0, 1); 10234f67d755SEric Taylor efi_free(gpt); 10244f67d755SEric Taylor } 10254f67d755SEric Taylor } 10264f67d755SEric Taylor 10274f67d755SEric Taylor static void 10284f67d755SEric Taylor zpool_open_func(void *arg) 10294f67d755SEric Taylor { 10304f67d755SEric Taylor rdsk_node_t *rn = arg; 10314f67d755SEric Taylor struct stat64 statbuf; 10324f67d755SEric Taylor nvlist_t *config; 10334f67d755SEric Taylor int fd; 10344f67d755SEric Taylor 10354f67d755SEric Taylor if (rn->rn_nozpool) 10364f67d755SEric Taylor return; 10374f67d755SEric Taylor if ((fd = openat64(rn->rn_dfd, rn->rn_name, O_RDONLY)) < 0) { 10384f67d755SEric Taylor /* symlink to a device that's no longer there */ 10394f67d755SEric Taylor if (errno == ENOENT) 10404f67d755SEric Taylor nozpool_all_slices(rn->rn_avl, rn->rn_name); 10414f67d755SEric Taylor return; 10424f67d755SEric Taylor } 10434f67d755SEric Taylor /* 10444f67d755SEric Taylor * Ignore failed stats. We only want regular 10454f67d755SEric Taylor * files, character devs and block devs. 10464f67d755SEric Taylor */ 10474f67d755SEric Taylor if (fstat64(fd, &statbuf) != 0 || 10484f67d755SEric Taylor (!S_ISREG(statbuf.st_mode) && 10494f67d755SEric Taylor !S_ISCHR(statbuf.st_mode) && 10504f67d755SEric Taylor !S_ISBLK(statbuf.st_mode))) { 10514f67d755SEric Taylor (void) close(fd); 10524f67d755SEric Taylor return; 10534f67d755SEric Taylor } 10544f67d755SEric Taylor /* this file is too small to hold a zpool */ 10554f67d755SEric Taylor if (S_ISREG(statbuf.st_mode) && 10564f67d755SEric Taylor statbuf.st_size < SPA_MINDEVSIZE) { 10574f67d755SEric Taylor (void) close(fd); 10584f67d755SEric Taylor return; 10594f67d755SEric Taylor } else if (!S_ISREG(statbuf.st_mode)) { 10604f67d755SEric Taylor /* 10614f67d755SEric Taylor * Try to read the disk label first so we don't have to 10624f67d755SEric Taylor * open a bunch of minor nodes that can't have a zpool. 10634f67d755SEric Taylor */ 10644f67d755SEric Taylor check_slices(rn->rn_avl, fd, rn->rn_name); 10654f67d755SEric Taylor } 10664f67d755SEric Taylor 10674f67d755SEric Taylor if ((zpool_read_label(fd, &config)) != 0) { 10684f67d755SEric Taylor (void) close(fd); 10694f67d755SEric Taylor (void) no_memory(rn->rn_hdl); 10704f67d755SEric Taylor return; 10714f67d755SEric Taylor } 10724f67d755SEric Taylor (void) close(fd); 10734f67d755SEric Taylor 10744f67d755SEric Taylor 10754f67d755SEric Taylor rn->rn_config = config; 10764f67d755SEric Taylor if (config != NULL) { 10774f67d755SEric Taylor assert(rn->rn_nozpool == B_FALSE); 10784f67d755SEric Taylor } 10794f67d755SEric Taylor } 10804f67d755SEric Taylor 1081096d22d4SEric Schrock /* 1082096d22d4SEric Schrock * Given a file descriptor, clear (zero) the label information. This function 1083096d22d4SEric Schrock * is currently only used in the appliance stack as part of the ZFS sysevent 1084096d22d4SEric Schrock * module. 1085096d22d4SEric Schrock */ 1086096d22d4SEric Schrock int 1087096d22d4SEric Schrock zpool_clear_label(int fd) 1088096d22d4SEric Schrock { 1089096d22d4SEric Schrock struct stat64 statbuf; 1090096d22d4SEric Schrock int l; 1091096d22d4SEric Schrock vdev_label_t *label; 1092096d22d4SEric Schrock uint64_t size; 1093096d22d4SEric Schrock 1094096d22d4SEric Schrock if (fstat64(fd, &statbuf) == -1) 1095096d22d4SEric Schrock return (0); 1096096d22d4SEric Schrock size = P2ALIGN_TYPED(statbuf.st_size, sizeof (vdev_label_t), uint64_t); 1097096d22d4SEric Schrock 1098096d22d4SEric Schrock if ((label = calloc(sizeof (vdev_label_t), 1)) == NULL) 1099096d22d4SEric Schrock return (-1); 1100096d22d4SEric Schrock 1101096d22d4SEric Schrock for (l = 0; l < VDEV_LABELS; l++) { 1102096d22d4SEric Schrock if (pwrite64(fd, label, sizeof (vdev_label_t), 1103096d22d4SEric Schrock label_offset(size, l)) != sizeof (vdev_label_t)) 1104096d22d4SEric Schrock return (-1); 1105096d22d4SEric Schrock } 1106096d22d4SEric Schrock 1107096d22d4SEric Schrock free(label); 1108096d22d4SEric Schrock return (0); 1109096d22d4SEric Schrock } 1110096d22d4SEric Schrock 1111fa9e4066Sahrens /* 1112fa9e4066Sahrens * Given a list of directories to search, find all pools stored on disk. This 1113fa9e4066Sahrens * includes partial pools which are not available to import. If no args are 1114fa9e4066Sahrens * given (argc is 0), then the default directory (/dev/dsk) is searched. 111524e697d4Sck * poolname or guid (but not both) are provided by the caller when trying 111624e697d4Sck * to import a specific pool. 1117fa9e4066Sahrens */ 111824e697d4Sck static nvlist_t * 1119d41c4376SMark J Musante zpool_find_import_impl(libzfs_handle_t *hdl, importargs_t *iarg) 1120fa9e4066Sahrens { 1121d41c4376SMark J Musante int i, dirs = iarg->paths; 1122ccae0b50Seschrock DIR *dirp = NULL; 1123fa9e4066Sahrens struct dirent64 *dp; 1124fa9e4066Sahrens char path[MAXPATHLEN]; 1125d41c4376SMark J Musante char *end, **dir = iarg->path; 1126842c5645Sjwadams size_t pathleft; 11274f67d755SEric Taylor nvlist_t *ret = NULL; 1128fa9e4066Sahrens static char *default_dir = "/dev/dsk"; 1129fa9e4066Sahrens pool_list_t pools = { 0 }; 113099653d4eSeschrock pool_entry_t *pe, *penext; 113199653d4eSeschrock vdev_entry_t *ve, *venext; 113299653d4eSeschrock config_entry_t *ce, *cenext; 113399653d4eSeschrock name_entry_t *ne, *nenext; 11344f67d755SEric Taylor avl_tree_t slice_cache; 11354f67d755SEric Taylor rdsk_node_t *slice; 11364f67d755SEric Taylor void *cookie; 113799653d4eSeschrock 1138d41c4376SMark J Musante if (dirs == 0) { 1139d41c4376SMark J Musante dirs = 1; 1140d41c4376SMark J Musante dir = &default_dir; 1141fa9e4066Sahrens } 1142fa9e4066Sahrens 1143fa9e4066Sahrens /* 1144fa9e4066Sahrens * Go through and read the label configuration information from every 1145fa9e4066Sahrens * possible device, organizing the information according to pool GUID 1146fa9e4066Sahrens * and toplevel GUID. 1147fa9e4066Sahrens */ 1148d41c4376SMark J Musante for (i = 0; i < dirs; i++) { 11494f67d755SEric Taylor tpool_t *t; 1150842c5645Sjwadams char *rdsk; 1151842c5645Sjwadams int dfd; 1152842c5645Sjwadams 1153842c5645Sjwadams /* use realpath to normalize the path */ 1154d41c4376SMark J Musante if (realpath(dir[i], path) == 0) { 1155ece3d9b3Slling (void) zfs_error_fmt(hdl, EZFS_BADPATH, 1156d41c4376SMark J Musante dgettext(TEXT_DOMAIN, "cannot open '%s'"), dir[i]); 115799653d4eSeschrock goto error; 1158fa9e4066Sahrens } 1159842c5645Sjwadams end = &path[strlen(path)]; 1160842c5645Sjwadams *end++ = '/'; 1161842c5645Sjwadams *end = 0; 1162842c5645Sjwadams pathleft = &path[sizeof (path)] - end; 1163842c5645Sjwadams 1164842c5645Sjwadams /* 1165842c5645Sjwadams * Using raw devices instead of block devices when we're 1166842c5645Sjwadams * reading the labels skips a bunch of slow operations during 1167842c5645Sjwadams * close(2) processing, so we replace /dev/dsk with /dev/rdsk. 1168842c5645Sjwadams */ 1169842c5645Sjwadams if (strcmp(path, "/dev/dsk/") == 0) 1170842c5645Sjwadams rdsk = "/dev/rdsk/"; 1171842c5645Sjwadams else 1172842c5645Sjwadams rdsk = path; 1173fa9e4066Sahrens 1174842c5645Sjwadams if ((dfd = open64(rdsk, O_RDONLY)) < 0 || 1175842c5645Sjwadams (dirp = fdopendir(dfd)) == NULL) { 117699653d4eSeschrock zfs_error_aux(hdl, strerror(errno)); 1177ece3d9b3Slling (void) zfs_error_fmt(hdl, EZFS_BADPATH, 117899653d4eSeschrock dgettext(TEXT_DOMAIN, "cannot open '%s'"), 1179842c5645Sjwadams rdsk); 118099653d4eSeschrock goto error; 1181fa9e4066Sahrens } 1182fa9e4066Sahrens 11834f67d755SEric Taylor avl_create(&slice_cache, slice_cache_compare, 11844f67d755SEric Taylor sizeof (rdsk_node_t), offsetof(rdsk_node_t, rn_node)); 1185fa9e4066Sahrens /* 1186fa9e4066Sahrens * This is not MT-safe, but we have no MT consumers of libzfs 1187fa9e4066Sahrens */ 1188fa9e4066Sahrens while ((dp = readdir64(dirp)) != NULL) { 1189842c5645Sjwadams const char *name = dp->d_name; 1190842c5645Sjwadams if (name[0] == '.' && 1191842c5645Sjwadams (name[1] == 0 || (name[1] == '.' && name[2] == 0))) 1192fa9e4066Sahrens continue; 1193fa9e4066Sahrens 11944f67d755SEric Taylor slice = zfs_alloc(hdl, sizeof (rdsk_node_t)); 11954f67d755SEric Taylor slice->rn_name = zfs_strdup(hdl, name); 11964f67d755SEric Taylor slice->rn_avl = &slice_cache; 11974f67d755SEric Taylor slice->rn_dfd = dfd; 11984f67d755SEric Taylor slice->rn_hdl = hdl; 11994f67d755SEric Taylor slice->rn_nozpool = B_FALSE; 12004f67d755SEric Taylor avl_add(&slice_cache, slice); 12014f67d755SEric Taylor } 12024f67d755SEric Taylor /* 12034f67d755SEric Taylor * create a thread pool to do all of this in parallel; 12044f67d755SEric Taylor * rn_nozpool is not protected, so this is racy in that 12054f67d755SEric Taylor * multiple tasks could decide that the same slice can 12064f67d755SEric Taylor * not hold a zpool, which is benign. Also choose 12074f67d755SEric Taylor * double the number of processors; we hold a lot of 12084f67d755SEric Taylor * locks in the kernel, so going beyond this doesn't 12094f67d755SEric Taylor * buy us much. 12104f67d755SEric Taylor */ 12114f67d755SEric Taylor t = tpool_create(1, 2 * sysconf(_SC_NPROCESSORS_ONLN), 12124f67d755SEric Taylor 0, NULL); 12134f67d755SEric Taylor for (slice = avl_first(&slice_cache); slice; 12144f67d755SEric Taylor (slice = avl_walk(&slice_cache, slice, 12154f67d755SEric Taylor AVL_AFTER))) 12164f67d755SEric Taylor (void) tpool_dispatch(t, zpool_open_func, slice); 12174f67d755SEric Taylor tpool_wait(t); 12184f67d755SEric Taylor tpool_destroy(t); 12194f67d755SEric Taylor 12204f67d755SEric Taylor cookie = NULL; 12214f67d755SEric Taylor while ((slice = avl_destroy_nodes(&slice_cache, 12224f67d755SEric Taylor &cookie)) != NULL) { 12234f67d755SEric Taylor if (slice->rn_config != NULL) { 12244f67d755SEric Taylor nvlist_t *config = slice->rn_config; 122524e697d4Sck boolean_t matched = B_TRUE; 122624e697d4Sck 1227d41c4376SMark J Musante if (iarg->poolname != NULL) { 122824e697d4Sck char *pname; 12297b0e68eaSck 12307b0e68eaSck matched = nvlist_lookup_string(config, 123124e697d4Sck ZPOOL_CONFIG_POOL_NAME, 12327b0e68eaSck &pname) == 0 && 1233d41c4376SMark J Musante strcmp(iarg->poolname, pname) == 0; 1234d41c4376SMark J Musante } else if (iarg->guid != 0) { 123524e697d4Sck uint64_t this_guid; 1236913f2d80Sck 1237913f2d80Sck matched = nvlist_lookup_uint64(config, 123824e697d4Sck ZPOOL_CONFIG_POOL_GUID, 1239913f2d80Sck &this_guid) == 0 && 1240d41c4376SMark J Musante iarg->guid == this_guid; 124124e697d4Sck } 124224e697d4Sck if (!matched) { 124324e697d4Sck nvlist_free(config); 124424e697d4Sck config = NULL; 124524e697d4Sck continue; 124624e697d4Sck } 1247842c5645Sjwadams /* use the non-raw path for the config */ 12484f67d755SEric Taylor (void) strlcpy(end, slice->rn_name, pathleft); 124999653d4eSeschrock if (add_config(hdl, &pools, path, config) != 0) 125099653d4eSeschrock goto error; 1251842c5645Sjwadams } 12524f67d755SEric Taylor free(slice->rn_name); 12534f67d755SEric Taylor free(slice); 1254fa9e4066Sahrens } 12554f67d755SEric Taylor avl_destroy(&slice_cache); 1256ccae0b50Seschrock 1257ccae0b50Seschrock (void) closedir(dirp); 1258ccae0b50Seschrock dirp = NULL; 1259fa9e4066Sahrens } 1260fa9e4066Sahrens 1261d41c4376SMark J Musante ret = get_configs(hdl, &pools, iarg->can_be_active); 126299653d4eSeschrock 126399653d4eSeschrock error: 126499653d4eSeschrock for (pe = pools.pools; pe != NULL; pe = penext) { 126599653d4eSeschrock penext = pe->pe_next; 126699653d4eSeschrock for (ve = pe->pe_vdevs; ve != NULL; ve = venext) { 126799653d4eSeschrock venext = ve->ve_next; 126899653d4eSeschrock for (ce = ve->ve_configs; ce != NULL; ce = cenext) { 126999653d4eSeschrock cenext = ce->ce_next; 127099653d4eSeschrock if (ce->ce_config) 127199653d4eSeschrock nvlist_free(ce->ce_config); 127299653d4eSeschrock free(ce); 127399653d4eSeschrock } 127499653d4eSeschrock free(ve); 127599653d4eSeschrock } 127699653d4eSeschrock free(pe); 127799653d4eSeschrock } 127899653d4eSeschrock 127999653d4eSeschrock for (ne = pools.names; ne != NULL; ne = nenext) { 128099653d4eSeschrock nenext = ne->ne_next; 128199653d4eSeschrock if (ne->ne_name) 128299653d4eSeschrock free(ne->ne_name); 128399653d4eSeschrock free(ne); 128499653d4eSeschrock } 128599653d4eSeschrock 1286ccae0b50Seschrock if (dirp) 1287ccae0b50Seschrock (void) closedir(dirp); 1288fa9e4066Sahrens 1289fa9e4066Sahrens return (ret); 1290fa9e4066Sahrens } 1291fa9e4066Sahrens 129224e697d4Sck nvlist_t * 129324e697d4Sck zpool_find_import(libzfs_handle_t *hdl, int argc, char **argv) 129424e697d4Sck { 1295d41c4376SMark J Musante importargs_t iarg = { 0 }; 129624e697d4Sck 1297d41c4376SMark J Musante iarg.paths = argc; 1298d41c4376SMark J Musante iarg.path = argv; 129924e697d4Sck 1300d41c4376SMark J Musante return (zpool_find_import_impl(hdl, &iarg)); 130124e697d4Sck } 130224e697d4Sck 13032f8aaab3Seschrock /* 13042f8aaab3Seschrock * Given a cache file, return the contents as a list of importable pools. 130524e697d4Sck * poolname or guid (but not both) are provided by the caller when trying 130624e697d4Sck * to import a specific pool. 13072f8aaab3Seschrock */ 13082f8aaab3Seschrock nvlist_t * 13093a57275aSck zpool_find_import_cached(libzfs_handle_t *hdl, const char *cachefile, 1310e829d913Sck char *poolname, uint64_t guid) 13112f8aaab3Seschrock { 13122f8aaab3Seschrock char *buf; 13132f8aaab3Seschrock int fd; 13142f8aaab3Seschrock struct stat64 statbuf; 13152f8aaab3Seschrock nvlist_t *raw, *src, *dst; 13162f8aaab3Seschrock nvlist_t *pools; 13172f8aaab3Seschrock nvpair_t *elem; 13182f8aaab3Seschrock char *name; 131924e697d4Sck uint64_t this_guid; 13202f8aaab3Seschrock boolean_t active; 13212f8aaab3Seschrock 132224e697d4Sck verify(poolname == NULL || guid == 0); 132324e697d4Sck 13242f8aaab3Seschrock if ((fd = open(cachefile, O_RDONLY)) < 0) { 13252f8aaab3Seschrock zfs_error_aux(hdl, "%s", strerror(errno)); 13262f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13272f8aaab3Seschrock dgettext(TEXT_DOMAIN, "failed to open cache file")); 13282f8aaab3Seschrock return (NULL); 13292f8aaab3Seschrock } 13302f8aaab3Seschrock 13312f8aaab3Seschrock if (fstat64(fd, &statbuf) != 0) { 13322f8aaab3Seschrock zfs_error_aux(hdl, "%s", strerror(errno)); 13332f8aaab3Seschrock (void) close(fd); 13342f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13352f8aaab3Seschrock dgettext(TEXT_DOMAIN, "failed to get size of cache file")); 13362f8aaab3Seschrock return (NULL); 13372f8aaab3Seschrock } 13382f8aaab3Seschrock 13392f8aaab3Seschrock if ((buf = zfs_alloc(hdl, statbuf.st_size)) == NULL) { 13402f8aaab3Seschrock (void) close(fd); 13412f8aaab3Seschrock return (NULL); 13422f8aaab3Seschrock } 13432f8aaab3Seschrock 13442f8aaab3Seschrock if (read(fd, buf, statbuf.st_size) != statbuf.st_size) { 13452f8aaab3Seschrock (void) close(fd); 13462f8aaab3Seschrock free(buf); 13472f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13482f8aaab3Seschrock dgettext(TEXT_DOMAIN, 13492f8aaab3Seschrock "failed to read cache file contents")); 13502f8aaab3Seschrock return (NULL); 13512f8aaab3Seschrock } 13522f8aaab3Seschrock 13532f8aaab3Seschrock (void) close(fd); 13542f8aaab3Seschrock 13552f8aaab3Seschrock if (nvlist_unpack(buf, statbuf.st_size, &raw, 0) != 0) { 13562f8aaab3Seschrock free(buf); 13572f8aaab3Seschrock (void) zfs_error(hdl, EZFS_BADCACHE, 13582f8aaab3Seschrock dgettext(TEXT_DOMAIN, 13592f8aaab3Seschrock "invalid or corrupt cache file contents")); 13602f8aaab3Seschrock return (NULL); 13612f8aaab3Seschrock } 13622f8aaab3Seschrock 13632f8aaab3Seschrock free(buf); 13642f8aaab3Seschrock 13652f8aaab3Seschrock /* 13662f8aaab3Seschrock * Go through and get the current state of the pools and refresh their 13672f8aaab3Seschrock * state. 13682f8aaab3Seschrock */ 13692f8aaab3Seschrock if (nvlist_alloc(&pools, 0, 0) != 0) { 13702f8aaab3Seschrock (void) no_memory(hdl); 13712f8aaab3Seschrock nvlist_free(raw); 13722f8aaab3Seschrock return (NULL); 13732f8aaab3Seschrock } 13742f8aaab3Seschrock 13752f8aaab3Seschrock elem = NULL; 13762f8aaab3Seschrock while ((elem = nvlist_next_nvpair(raw, elem)) != NULL) { 13772f8aaab3Seschrock verify(nvpair_value_nvlist(elem, &src) == 0); 13782f8aaab3Seschrock 13792f8aaab3Seschrock verify(nvlist_lookup_string(src, ZPOOL_CONFIG_POOL_NAME, 13802f8aaab3Seschrock &name) == 0); 138124e697d4Sck if (poolname != NULL && strcmp(poolname, name) != 0) 138224e697d4Sck continue; 138324e697d4Sck 13842f8aaab3Seschrock verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 138524e697d4Sck &this_guid) == 0); 138624e697d4Sck if (guid != 0) { 138724e697d4Sck verify(nvlist_lookup_uint64(src, ZPOOL_CONFIG_POOL_GUID, 138824e697d4Sck &this_guid) == 0); 138924e697d4Sck if (guid != this_guid) 139024e697d4Sck continue; 139124e697d4Sck } 13922f8aaab3Seschrock 1393e829d913Sck if (pool_active(hdl, name, this_guid, &active) != 0) { 1394e829d913Sck nvlist_free(raw); 1395e829d913Sck nvlist_free(pools); 1396e829d913Sck return (NULL); 1397e829d913Sck } 13982f8aaab3Seschrock 1399e829d913Sck if (active) 1400e829d913Sck continue; 14012f8aaab3Seschrock 1402e829d913Sck if ((dst = refresh_config(hdl, src)) == NULL) { 1403e829d913Sck nvlist_free(raw); 1404e829d913Sck nvlist_free(pools); 1405e829d913Sck return (NULL); 1406e829d913Sck } 14072f8aaab3Seschrock 1408e829d913Sck if (nvlist_add_nvlist(pools, nvpair_name(elem), dst) != 0) { 1409e829d913Sck (void) no_memory(hdl); 14102f8aaab3Seschrock nvlist_free(dst); 1411e829d913Sck nvlist_free(raw); 1412e829d913Sck nvlist_free(pools); 1413e829d913Sck return (NULL); 14142f8aaab3Seschrock } 1415e829d913Sck nvlist_free(dst); 14162f8aaab3Seschrock } 14172f8aaab3Seschrock 14182f8aaab3Seschrock nvlist_free(raw); 14192f8aaab3Seschrock return (pools); 14202f8aaab3Seschrock } 14212f8aaab3Seschrock 1422d41c4376SMark J Musante static int 1423d41c4376SMark J Musante name_or_guid_exists(zpool_handle_t *zhp, void *data) 1424d41c4376SMark J Musante { 1425d41c4376SMark J Musante importargs_t *import = data; 1426d41c4376SMark J Musante int found = 0; 1427d41c4376SMark J Musante 1428d41c4376SMark J Musante if (import->poolname != NULL) { 1429d41c4376SMark J Musante char *pool_name; 1430d41c4376SMark J Musante 1431d41c4376SMark J Musante verify(nvlist_lookup_string(zhp->zpool_config, 1432d41c4376SMark J Musante ZPOOL_CONFIG_POOL_NAME, &pool_name) == 0); 1433d41c4376SMark J Musante if (strcmp(pool_name, import->poolname) == 0) 1434d41c4376SMark J Musante found = 1; 1435d41c4376SMark J Musante } else { 1436d41c4376SMark J Musante uint64_t pool_guid; 1437d41c4376SMark J Musante 1438d41c4376SMark J Musante verify(nvlist_lookup_uint64(zhp->zpool_config, 1439d41c4376SMark J Musante ZPOOL_CONFIG_POOL_GUID, &pool_guid) == 0); 1440d41c4376SMark J Musante if (pool_guid == import->guid) 1441d41c4376SMark J Musante found = 1; 1442d41c4376SMark J Musante } 1443d41c4376SMark J Musante 1444d41c4376SMark J Musante zpool_close(zhp); 1445d41c4376SMark J Musante return (found); 1446d41c4376SMark J Musante } 1447d41c4376SMark J Musante 1448d41c4376SMark J Musante nvlist_t * 1449d41c4376SMark J Musante zpool_search_import(libzfs_handle_t *hdl, importargs_t *import) 1450d41c4376SMark J Musante { 1451d41c4376SMark J Musante verify(import->poolname == NULL || import->guid == 0); 1452d41c4376SMark J Musante 1453d41c4376SMark J Musante if (import->unique) 1454d41c4376SMark J Musante import->exists = zpool_iter(hdl, name_or_guid_exists, import); 1455d41c4376SMark J Musante 1456d41c4376SMark J Musante if (import->cachefile != NULL) 1457d41c4376SMark J Musante return (zpool_find_import_cached(hdl, import->cachefile, 1458d41c4376SMark J Musante import->poolname, import->guid)); 1459d41c4376SMark J Musante 1460d41c4376SMark J Musante return (zpool_find_import_impl(hdl, import)); 1461d41c4376SMark J Musante } 14622f8aaab3Seschrock 146399653d4eSeschrock boolean_t 1464fa9e4066Sahrens find_guid(nvlist_t *nv, uint64_t guid) 1465fa9e4066Sahrens { 1466fa9e4066Sahrens uint64_t tmp; 1467fa9e4066Sahrens nvlist_t **child; 1468fa9e4066Sahrens uint_t c, children; 1469fa9e4066Sahrens 1470fa9e4066Sahrens verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &tmp) == 0); 1471fa9e4066Sahrens if (tmp == guid) 147299653d4eSeschrock return (B_TRUE); 1473fa9e4066Sahrens 1474fa9e4066Sahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, 1475fa9e4066Sahrens &child, &children) == 0) { 1476fa9e4066Sahrens for (c = 0; c < children; c++) 1477fa9e4066Sahrens if (find_guid(child[c], guid)) 147899653d4eSeschrock return (B_TRUE); 147999653d4eSeschrock } 148099653d4eSeschrock 148199653d4eSeschrock return (B_FALSE); 148299653d4eSeschrock } 148399653d4eSeschrock 1484fa94a07fSbrendan typedef struct aux_cbdata { 1485fa94a07fSbrendan const char *cb_type; 148699653d4eSeschrock uint64_t cb_guid; 148799653d4eSeschrock zpool_handle_t *cb_zhp; 1488fa94a07fSbrendan } aux_cbdata_t; 148999653d4eSeschrock 149099653d4eSeschrock static int 1491fa94a07fSbrendan find_aux(zpool_handle_t *zhp, void *data) 149299653d4eSeschrock { 1493fa94a07fSbrendan aux_cbdata_t *cbp = data; 1494fa94a07fSbrendan nvlist_t **list; 1495fa94a07fSbrendan uint_t i, count; 149699653d4eSeschrock uint64_t guid; 149799653d4eSeschrock nvlist_t *nvroot; 149899653d4eSeschrock 149999653d4eSeschrock verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, 150099653d4eSeschrock &nvroot) == 0); 150199653d4eSeschrock 1502fa94a07fSbrendan if (nvlist_lookup_nvlist_array(nvroot, cbp->cb_type, 1503fa94a07fSbrendan &list, &count) == 0) { 1504fa94a07fSbrendan for (i = 0; i < count; i++) { 1505fa94a07fSbrendan verify(nvlist_lookup_uint64(list[i], 150699653d4eSeschrock ZPOOL_CONFIG_GUID, &guid) == 0); 150799653d4eSeschrock if (guid == cbp->cb_guid) { 150899653d4eSeschrock cbp->cb_zhp = zhp; 150999653d4eSeschrock return (1); 151099653d4eSeschrock } 151199653d4eSeschrock } 1512fa9e4066Sahrens } 1513fa9e4066Sahrens 151499653d4eSeschrock zpool_close(zhp); 151599653d4eSeschrock return (0); 1516fa9e4066Sahrens } 1517fa9e4066Sahrens 1518fa9e4066Sahrens /* 151999653d4eSeschrock * Determines if the pool is in use. If so, it returns true and the state of 1520fa9e4066Sahrens * the pool as well as the name of the pool. Both strings are allocated and 1521fa9e4066Sahrens * must be freed by the caller. 1522fa9e4066Sahrens */ 1523fa9e4066Sahrens int 152499653d4eSeschrock zpool_in_use(libzfs_handle_t *hdl, int fd, pool_state_t *state, char **namestr, 152599653d4eSeschrock boolean_t *inuse) 1526fa9e4066Sahrens { 1527fa9e4066Sahrens nvlist_t *config; 1528fa9e4066Sahrens char *name; 152999653d4eSeschrock boolean_t ret; 1530fa9e4066Sahrens uint64_t guid, vdev_guid; 1531fa9e4066Sahrens zpool_handle_t *zhp; 1532fa9e4066Sahrens nvlist_t *pool_config; 153339c23413Seschrock uint64_t stateval, isspare; 1534fa94a07fSbrendan aux_cbdata_t cb = { 0 }; 153594de1d4cSeschrock boolean_t isactive; 153699653d4eSeschrock 153799653d4eSeschrock *inuse = B_FALSE; 1538fa9e4066Sahrens 153999653d4eSeschrock if (zpool_read_label(fd, &config) != 0) { 154099653d4eSeschrock (void) no_memory(hdl); 154199653d4eSeschrock return (-1); 154299653d4eSeschrock } 154399653d4eSeschrock 154499653d4eSeschrock if (config == NULL) 154599653d4eSeschrock return (0); 1546fa9e4066Sahrens 1547fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_STATE, 154846a2abf2Seschrock &stateval) == 0); 1549fa9e4066Sahrens verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_GUID, 1550fa9e4066Sahrens &vdev_guid) == 0); 1551fa9e4066Sahrens 1552fa94a07fSbrendan if (stateval != POOL_STATE_SPARE && stateval != POOL_STATE_L2CACHE) { 155399653d4eSeschrock verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, 155499653d4eSeschrock &name) == 0); 155599653d4eSeschrock verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, 155699653d4eSeschrock &guid) == 0); 155799653d4eSeschrock } 155899653d4eSeschrock 155946a2abf2Seschrock switch (stateval) { 1560fa9e4066Sahrens case POOL_STATE_EXPORTED: 1561*f9af39baSGeorge Wilson /* 1562*f9af39baSGeorge Wilson * A pool with an exported state may in fact be imported 1563*f9af39baSGeorge Wilson * read-only, so check the in-core state to see if it's 1564*f9af39baSGeorge Wilson * active and imported read-only. If it is, set 1565*f9af39baSGeorge Wilson * its state to active. 1566*f9af39baSGeorge Wilson */ 1567*f9af39baSGeorge Wilson if (pool_active(hdl, name, guid, &isactive) == 0 && isactive && 1568*f9af39baSGeorge Wilson (zhp = zpool_open_canfail(hdl, name)) != NULL && 1569*f9af39baSGeorge Wilson zpool_get_prop_int(zhp, ZPOOL_PROP_READONLY, NULL)) 1570*f9af39baSGeorge Wilson stateval = POOL_STATE_ACTIVE; 1571*f9af39baSGeorge Wilson 157299653d4eSeschrock ret = B_TRUE; 1573fa9e4066Sahrens break; 1574fa9e4066Sahrens 1575fa9e4066Sahrens case POOL_STATE_ACTIVE: 1576fa9e4066Sahrens /* 1577fa9e4066Sahrens * For an active pool, we have to determine if it's really part 1578eaca9bbdSeschrock * of a currently active pool (in which case the pool will exist 1579eaca9bbdSeschrock * and the guid will be the same), or whether it's part of an 1580eaca9bbdSeschrock * active pool that was disconnected without being explicitly 1581eaca9bbdSeschrock * exported. 1582fa9e4066Sahrens */ 158394de1d4cSeschrock if (pool_active(hdl, name, guid, &isactive) != 0) { 158494de1d4cSeschrock nvlist_free(config); 158594de1d4cSeschrock return (-1); 158694de1d4cSeschrock } 158794de1d4cSeschrock 158894de1d4cSeschrock if (isactive) { 1589fa9e4066Sahrens /* 1590fa9e4066Sahrens * Because the device may have been removed while 1591fa9e4066Sahrens * offlined, we only report it as active if the vdev is 1592fa9e4066Sahrens * still present in the config. Otherwise, pretend like 1593fa9e4066Sahrens * it's not in use. 1594fa9e4066Sahrens */ 159599653d4eSeschrock if ((zhp = zpool_open_canfail(hdl, name)) != NULL && 1596088e9d47Seschrock (pool_config = zpool_get_config(zhp, NULL)) 1597088e9d47Seschrock != NULL) { 1598fa9e4066Sahrens nvlist_t *nvroot; 1599fa9e4066Sahrens 1600fa9e4066Sahrens verify(nvlist_lookup_nvlist(pool_config, 1601fa9e4066Sahrens ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); 160246a2abf2Seschrock ret = find_guid(nvroot, vdev_guid); 1603fa9e4066Sahrens } else { 160499653d4eSeschrock ret = B_FALSE; 1605fa9e4066Sahrens } 160699653d4eSeschrock 160739c23413Seschrock /* 160839c23413Seschrock * If this is an active spare within another pool, we 160939c23413Seschrock * treat it like an unused hot spare. This allows the 161039c23413Seschrock * user to create a pool with a hot spare that currently 161139c23413Seschrock * in use within another pool. Since we return B_TRUE, 161239c23413Seschrock * libdiskmgt will continue to prevent generic consumers 161339c23413Seschrock * from using the device. 161439c23413Seschrock */ 161539c23413Seschrock if (ret && nvlist_lookup_uint64(config, 161639c23413Seschrock ZPOOL_CONFIG_IS_SPARE, &isspare) == 0 && isspare) 161739c23413Seschrock stateval = POOL_STATE_SPARE; 161839c23413Seschrock 161999653d4eSeschrock if (zhp != NULL) 162099653d4eSeschrock zpool_close(zhp); 1621fa9e4066Sahrens } else { 162246a2abf2Seschrock stateval = POOL_STATE_POTENTIALLY_ACTIVE; 162399653d4eSeschrock ret = B_TRUE; 162499653d4eSeschrock } 162599653d4eSeschrock break; 162699653d4eSeschrock 162799653d4eSeschrock case POOL_STATE_SPARE: 162899653d4eSeschrock /* 162999653d4eSeschrock * For a hot spare, it can be either definitively in use, or 163099653d4eSeschrock * potentially active. To determine if it's in use, we iterate 163199653d4eSeschrock * over all pools in the system and search for one with a spare 163299653d4eSeschrock * with a matching guid. 163399653d4eSeschrock * 163499653d4eSeschrock * Due to the shared nature of spares, we don't actually report 163599653d4eSeschrock * the potentially active case as in use. This means the user 163699653d4eSeschrock * can freely create pools on the hot spares of exported pools, 163799653d4eSeschrock * but to do otherwise makes the resulting code complicated, and 163899653d4eSeschrock * we end up having to deal with this case anyway. 163999653d4eSeschrock */ 164099653d4eSeschrock cb.cb_zhp = NULL; 164199653d4eSeschrock cb.cb_guid = vdev_guid; 1642fa94a07fSbrendan cb.cb_type = ZPOOL_CONFIG_SPARES; 1643fa94a07fSbrendan if (zpool_iter(hdl, find_aux, &cb) == 1) { 1644fa94a07fSbrendan name = (char *)zpool_get_name(cb.cb_zhp); 1645fa94a07fSbrendan ret = TRUE; 1646fa94a07fSbrendan } else { 1647fa94a07fSbrendan ret = FALSE; 1648fa94a07fSbrendan } 1649fa94a07fSbrendan break; 1650fa94a07fSbrendan 1651fa94a07fSbrendan case POOL_STATE_L2CACHE: 1652fa94a07fSbrendan 1653fa94a07fSbrendan /* 1654fa94a07fSbrendan * Check if any pool is currently using this l2cache device. 1655fa94a07fSbrendan */ 1656fa94a07fSbrendan cb.cb_zhp = NULL; 1657fa94a07fSbrendan cb.cb_guid = vdev_guid; 1658fa94a07fSbrendan cb.cb_type = ZPOOL_CONFIG_L2CACHE; 1659fa94a07fSbrendan if (zpool_iter(hdl, find_aux, &cb) == 1) { 166099653d4eSeschrock name = (char *)zpool_get_name(cb.cb_zhp); 1661fa9e4066Sahrens ret = TRUE; 166299653d4eSeschrock } else { 166399653d4eSeschrock ret = FALSE; 1664fa9e4066Sahrens } 1665fa9e4066Sahrens break; 1666fa9e4066Sahrens 1667fa9e4066Sahrens default: 166899653d4eSeschrock ret = B_FALSE; 1669fa9e4066Sahrens } 1670fa9e4066Sahrens 167146a2abf2Seschrock 167246a2abf2Seschrock if (ret) { 167399653d4eSeschrock if ((*namestr = zfs_strdup(hdl, name)) == NULL) { 1674c6765aabSeschrock if (cb.cb_zhp) 1675c6765aabSeschrock zpool_close(cb.cb_zhp); 167699653d4eSeschrock nvlist_free(config); 167799653d4eSeschrock return (-1); 167899653d4eSeschrock } 167946a2abf2Seschrock *state = (pool_state_t)stateval; 168046a2abf2Seschrock } 168146a2abf2Seschrock 168299653d4eSeschrock if (cb.cb_zhp) 168399653d4eSeschrock zpool_close(cb.cb_zhp); 168499653d4eSeschrock 1685fa9e4066Sahrens nvlist_free(config); 168699653d4eSeschrock *inuse = ret; 168799653d4eSeschrock return (0); 1688fa9e4066Sahrens } 1689