1b1b8ab34Slling /* 2b1b8ab34Slling * GRUB -- GRand Unified Bootloader 3b1b8ab34Slling * Copyright (C) 1999,2000,2001,2002,2003,2004 Free Software Foundation, Inc. 4b1b8ab34Slling * 5b1b8ab34Slling * This program is free software; you can redistribute it and/or modify 6b1b8ab34Slling * it under the terms of the GNU General Public License as published by 7b1b8ab34Slling * the Free Software Foundation; either version 2 of the License, or 8b1b8ab34Slling * (at your option) any later version. 9b1b8ab34Slling * 10b1b8ab34Slling * This program is distributed in the hope that it will be useful, 11b1b8ab34Slling * but WITHOUT ANY WARRANTY; without even the implied warranty of 12b1b8ab34Slling * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13b1b8ab34Slling * GNU General Public License for more details. 14b1b8ab34Slling * 15b1b8ab34Slling * You should have received a copy of the GNU General Public License 16b1b8ab34Slling * along with this program; if not, write to the Free Software 17b1b8ab34Slling * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. 18b1b8ab34Slling */ 19b1b8ab34Slling /* 202264ca7fSLin Ling * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 21b1b8ab34Slling * Use is subject to license terms. 22b1b8ab34Slling */ 23b1b8ab34Slling 24b1b8ab34Slling /* 25b1b8ab34Slling * The zfs plug-in routines for GRUB are: 26b1b8ab34Slling * 27b1b8ab34Slling * zfs_mount() - locates a valid uberblock of the root pool and reads 28b1b8ab34Slling * in its MOS at the memory address MOS. 29b1b8ab34Slling * 30b1b8ab34Slling * zfs_open() - locates a plain file object by following the MOS 31b1b8ab34Slling * and places its dnode at the memory address DNODE. 32b1b8ab34Slling * 33b1b8ab34Slling * zfs_read() - read in the data blocks pointed by the DNODE. 34b1b8ab34Slling * 35b1b8ab34Slling * ZFS_SCRATCH is used as a working area. 36b1b8ab34Slling * 37b1b8ab34Slling * (memory addr) MOS DNODE ZFS_SCRATCH 38b1b8ab34Slling * | | | 39b1b8ab34Slling * +-------V---------V----------V---------------+ 40b1b8ab34Slling * memory | | dnode | dnode | scratch | 41b1b8ab34Slling * | | 512B | 512B | area | 42b1b8ab34Slling * +--------------------------------------------+ 43b1b8ab34Slling */ 44b1b8ab34Slling 45b1b8ab34Slling #ifdef FSYS_ZFS 46b1b8ab34Slling 47b1b8ab34Slling #include "shared.h" 48b1b8ab34Slling #include "filesys.h" 49b1b8ab34Slling #include "fsys_zfs.h" 50b1b8ab34Slling 51b1b8ab34Slling /* cache for a file block of the currently zfs_open()-ed file */ 52b1b8ab34Slling static void *file_buf = NULL; 53b1b8ab34Slling static uint64_t file_start = 0; 54b1b8ab34Slling static uint64_t file_end = 0; 55b1b8ab34Slling 56b1b8ab34Slling /* cache for a dnode block */ 57b1b8ab34Slling static dnode_phys_t *dnode_buf = NULL; 58b1b8ab34Slling static dnode_phys_t *dnode_mdn = NULL; 59b1b8ab34Slling static uint64_t dnode_start = 0; 60b1b8ab34Slling static uint64_t dnode_end = 0; 61b1b8ab34Slling 62*e23347b1SEric Taylor static uint64_t pool_guid = 0; 63051aabe6Staylor static uberblock_t current_uberblock; 64b1b8ab34Slling static char *stackbase; 65b1b8ab34Slling 66b1b8ab34Slling decomp_entry_t decomp_table[ZIO_COMPRESS_FUNCTIONS] = 67b1b8ab34Slling { 6815e6edf1Sgw {"inherit", 0}, /* ZIO_COMPRESS_INHERIT */ 69b1b8ab34Slling {"on", lzjb_decompress}, /* ZIO_COMPRESS_ON */ 7015e6edf1Sgw {"off", 0}, /* ZIO_COMPRESS_OFF */ 7115e6edf1Sgw {"lzjb", lzjb_decompress}, /* ZIO_COMPRESS_LZJB */ 7215e6edf1Sgw {"empty", 0} /* ZIO_COMPRESS_EMPTY */ 73b1b8ab34Slling }; 74b1b8ab34Slling 75cd9c78d9SLin Ling static int zio_read_data(blkptr_t *bp, void *buf, char *stack); 76cd9c78d9SLin Ling 77b1b8ab34Slling /* 78b1b8ab34Slling * Our own version of bcmp(). 79b1b8ab34Slling */ 80b1b8ab34Slling static int 81b1b8ab34Slling zfs_bcmp(const void *s1, const void *s2, size_t n) 82b1b8ab34Slling { 83b1b8ab34Slling const uchar_t *ps1 = s1; 84b1b8ab34Slling const uchar_t *ps2 = s2; 85b1b8ab34Slling 86b1b8ab34Slling if (s1 != s2 && n != 0) { 87b1b8ab34Slling do { 88b1b8ab34Slling if (*ps1++ != *ps2++) 89b1b8ab34Slling return (1); 90b1b8ab34Slling } while (--n != 0); 91b1b8ab34Slling } 92b1b8ab34Slling 93b1b8ab34Slling return (0); 94b1b8ab34Slling } 95b1b8ab34Slling 96b1b8ab34Slling /* 97b1b8ab34Slling * Our own version of log2(). Same thing as highbit()-1. 98b1b8ab34Slling */ 99b1b8ab34Slling static int 100b1b8ab34Slling zfs_log2(uint64_t num) 101b1b8ab34Slling { 102b1b8ab34Slling int i = 0; 103b1b8ab34Slling 104b1b8ab34Slling while (num > 1) { 105b1b8ab34Slling i++; 106b1b8ab34Slling num = num >> 1; 107b1b8ab34Slling } 108b1b8ab34Slling 109b1b8ab34Slling return (i); 110b1b8ab34Slling } 111b1b8ab34Slling 112b1b8ab34Slling /* Checksum Functions */ 113b1b8ab34Slling static void 114b1b8ab34Slling zio_checksum_off(const void *buf, uint64_t size, zio_cksum_t *zcp) 115b1b8ab34Slling { 116b1b8ab34Slling ZIO_SET_CHECKSUM(zcp, 0, 0, 0, 0); 117b1b8ab34Slling } 118b1b8ab34Slling 119b1b8ab34Slling /* Checksum Table and Values */ 120b1b8ab34Slling zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS] = { 121b1b8ab34Slling NULL, NULL, 0, 0, "inherit", 122b1b8ab34Slling NULL, NULL, 0, 0, "on", 123b1b8ab34Slling zio_checksum_off, zio_checksum_off, 0, 0, "off", 124b1b8ab34Slling zio_checksum_SHA256, zio_checksum_SHA256, 1, 1, "label", 125b1b8ab34Slling zio_checksum_SHA256, zio_checksum_SHA256, 1, 1, "gang_header", 126b1b8ab34Slling fletcher_2_native, fletcher_2_byteswap, 0, 1, "zilog", 127b1b8ab34Slling fletcher_2_native, fletcher_2_byteswap, 0, 0, "fletcher2", 128b1b8ab34Slling fletcher_4_native, fletcher_4_byteswap, 1, 0, "fletcher4", 129b1b8ab34Slling zio_checksum_SHA256, zio_checksum_SHA256, 1, 0, "SHA256", 130b1b8ab34Slling }; 131b1b8ab34Slling 132b1b8ab34Slling /* 133b1b8ab34Slling * zio_checksum_verify: Provides support for checksum verification. 134b1b8ab34Slling * 135b1b8ab34Slling * Fletcher2, Fletcher4, and SHA256 are supported. 136b1b8ab34Slling * 137b1b8ab34Slling * Return: 138b1b8ab34Slling * -1 = Failure 139b1b8ab34Slling * 0 = Success 140b1b8ab34Slling */ 141b1b8ab34Slling static int 142b1b8ab34Slling zio_checksum_verify(blkptr_t *bp, char *data, int size) 143b1b8ab34Slling { 144b1b8ab34Slling zio_cksum_t zc = bp->blk_cksum; 145cd9c78d9SLin Ling uint32_t checksum = BP_GET_CHECKSUM(bp); 146b1b8ab34Slling int byteswap = BP_SHOULD_BYTESWAP(bp); 147b1b8ab34Slling zio_block_tail_t *zbt = (zio_block_tail_t *)(data + size) - 1; 148b1b8ab34Slling zio_checksum_info_t *ci = &zio_checksum_table[checksum]; 149b1b8ab34Slling zio_cksum_t actual_cksum, expected_cksum; 150b1b8ab34Slling 151b1b8ab34Slling /* byteswap is not supported */ 152b1b8ab34Slling if (byteswap) 153b1b8ab34Slling return (-1); 154b1b8ab34Slling 155b1b8ab34Slling if (checksum >= ZIO_CHECKSUM_FUNCTIONS || ci->ci_func[0] == NULL) 156b1b8ab34Slling return (-1); 157b1b8ab34Slling 158b1b8ab34Slling if (ci->ci_zbt) { 159cd9c78d9SLin Ling expected_cksum = zbt->zbt_cksum; 160cd9c78d9SLin Ling zbt->zbt_cksum = zc; 161cd9c78d9SLin Ling ci->ci_func[0](data, size, &actual_cksum); 162cd9c78d9SLin Ling zbt->zbt_cksum = expected_cksum; 163b1b8ab34Slling zc = expected_cksum; 164b1b8ab34Slling 165b1b8ab34Slling } else { 166b1b8ab34Slling ci->ci_func[byteswap](data, size, &actual_cksum); 167b1b8ab34Slling } 168b1b8ab34Slling 169b1b8ab34Slling if ((actual_cksum.zc_word[0] - zc.zc_word[0]) | 170b1b8ab34Slling (actual_cksum.zc_word[1] - zc.zc_word[1]) | 171b1b8ab34Slling (actual_cksum.zc_word[2] - zc.zc_word[2]) | 172b1b8ab34Slling (actual_cksum.zc_word[3] - zc.zc_word[3])) 173b1b8ab34Slling return (-1); 174b1b8ab34Slling 175b1b8ab34Slling return (0); 176b1b8ab34Slling } 177b1b8ab34Slling 178b1b8ab34Slling /* 179*e23347b1SEric Taylor * vdev_label_start returns the physical disk offset (in bytes) of 180*e23347b1SEric Taylor * label "l". 181b1b8ab34Slling */ 182e7cbe64fSgw static uint64_t 183*e23347b1SEric Taylor vdev_label_start(uint64_t psize, int l) 184b1b8ab34Slling { 185*e23347b1SEric Taylor return (l * sizeof (vdev_label_t) + (l < VDEV_LABELS / 2 ? 186b1b8ab34Slling 0 : psize - VDEV_LABELS * sizeof (vdev_label_t))); 187b1b8ab34Slling } 188b1b8ab34Slling 189b1b8ab34Slling /* 190b1b8ab34Slling * vdev_uberblock_compare takes two uberblock structures and returns an integer 191b1b8ab34Slling * indicating the more recent of the two. 192b1b8ab34Slling * Return Value = 1 if ub2 is more recent 193b1b8ab34Slling * Return Value = -1 if ub1 is more recent 194b1b8ab34Slling * The most recent uberblock is determined using its transaction number and 195b1b8ab34Slling * timestamp. The uberblock with the highest transaction number is 196b1b8ab34Slling * considered "newer". If the transaction numbers of the two blocks match, the 197b1b8ab34Slling * timestamps are compared to determine the "newer" of the two. 198b1b8ab34Slling */ 199b1b8ab34Slling static int 200b1b8ab34Slling vdev_uberblock_compare(uberblock_t *ub1, uberblock_t *ub2) 201b1b8ab34Slling { 202b1b8ab34Slling if (ub1->ub_txg < ub2->ub_txg) 203b1b8ab34Slling return (-1); 204b1b8ab34Slling if (ub1->ub_txg > ub2->ub_txg) 205b1b8ab34Slling return (1); 206b1b8ab34Slling 207b1b8ab34Slling if (ub1->ub_timestamp < ub2->ub_timestamp) 208b1b8ab34Slling return (-1); 209b1b8ab34Slling if (ub1->ub_timestamp > ub2->ub_timestamp) 210b1b8ab34Slling return (1); 211b1b8ab34Slling 212b1b8ab34Slling return (0); 213b1b8ab34Slling } 214b1b8ab34Slling 215b1b8ab34Slling /* 216b1b8ab34Slling * Three pieces of information are needed to verify an uberblock: the magic 217b1b8ab34Slling * number, the version number, and the checksum. 218b1b8ab34Slling * 219b1b8ab34Slling * Currently Implemented: version number, magic number 220b1b8ab34Slling * Need to Implement: checksum 221b1b8ab34Slling * 222b1b8ab34Slling * Return: 223b1b8ab34Slling * 0 - Success 224b1b8ab34Slling * -1 - Failure 225b1b8ab34Slling */ 226b1b8ab34Slling static int 227*e23347b1SEric Taylor uberblock_verify(uberblock_phys_t *ub, uint64_t offset) 228b1b8ab34Slling { 229b1b8ab34Slling 230b1b8ab34Slling uberblock_t *uber = &ub->ubp_uberblock; 231b1b8ab34Slling blkptr_t bp; 232b1b8ab34Slling 233b1b8ab34Slling BP_ZERO(&bp); 234b1b8ab34Slling BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); 235b1b8ab34Slling BP_SET_BYTEORDER(&bp, ZFS_HOST_BYTEORDER); 236b1b8ab34Slling ZIO_SET_CHECKSUM(&bp.blk_cksum, offset, 0, 0, 0); 237b1b8ab34Slling 238b1b8ab34Slling if (zio_checksum_verify(&bp, (char *)ub, UBERBLOCK_SIZE) != 0) 239b1b8ab34Slling return (-1); 240b1b8ab34Slling 241b1b8ab34Slling if (uber->ub_magic == UBERBLOCK_MAGIC && 242bb0ade09Sahrens uber->ub_version > 0 && uber->ub_version <= SPA_VERSION) 243b1b8ab34Slling return (0); 244b1b8ab34Slling 245b1b8ab34Slling return (-1); 246b1b8ab34Slling } 247b1b8ab34Slling 248b1b8ab34Slling /* 249b1b8ab34Slling * Find the best uberblock. 250b1b8ab34Slling * Return: 251b1b8ab34Slling * Success - Pointer to the best uberblock. 252b1b8ab34Slling * Failure - NULL 253b1b8ab34Slling */ 254b1b8ab34Slling static uberblock_phys_t * 255*e23347b1SEric Taylor find_bestub(uberblock_phys_t *ub_array, uint64_t sector) 256b1b8ab34Slling { 257b1b8ab34Slling uberblock_phys_t *ubbest = NULL; 258*e23347b1SEric Taylor uint64_t offset; 259*e23347b1SEric Taylor int i; 260b1b8ab34Slling 261b1b8ab34Slling for (i = 0; i < (VDEV_UBERBLOCK_RING >> VDEV_UBERBLOCK_SHIFT); i++) { 262*e23347b1SEric Taylor offset = (sector << SPA_MINBLOCKSHIFT) + 263*e23347b1SEric Taylor VDEV_UBERBLOCK_OFFSET(i); 264b1b8ab34Slling if (uberblock_verify(&ub_array[i], offset) == 0) { 265b1b8ab34Slling if (ubbest == NULL) { 266b1b8ab34Slling ubbest = &ub_array[i]; 267e7437265Sahrens } else if (vdev_uberblock_compare( 268e7437265Sahrens &(ub_array[i].ubp_uberblock), 269e7437265Sahrens &(ubbest->ubp_uberblock)) > 0) { 270e7437265Sahrens ubbest = &ub_array[i]; 271b1b8ab34Slling } 272b1b8ab34Slling } 273b1b8ab34Slling } 274b1b8ab34Slling 275b1b8ab34Slling return (ubbest); 276b1b8ab34Slling } 277b1b8ab34Slling 278b1b8ab34Slling /* 279cd9c78d9SLin Ling * Read a block of data based on the gang block address dva, 280cd9c78d9SLin Ling * and put its data in buf. 281b1b8ab34Slling * 282b1b8ab34Slling * Return: 283b1b8ab34Slling * 0 - success 284cd9c78d9SLin Ling * 1 - failure 285b1b8ab34Slling */ 286b1b8ab34Slling static int 287cd9c78d9SLin Ling zio_read_gang(blkptr_t *bp, dva_t *dva, void *buf, char *stack) 288b1b8ab34Slling { 289cd9c78d9SLin Ling zio_gbh_phys_t *zio_gb; 290b1b8ab34Slling uint64_t offset, sector; 291cd9c78d9SLin Ling blkptr_t tmpbp; 292cd9c78d9SLin Ling int i; 293b1b8ab34Slling 294cd9c78d9SLin Ling zio_gb = (zio_gbh_phys_t *)stack; 295cd9c78d9SLin Ling stack += SPA_GANGBLOCKSIZE; 296cd9c78d9SLin Ling offset = DVA_GET_OFFSET(dva); 297cd9c78d9SLin Ling sector = DVA_OFFSET_TO_PHYS_SECTOR(offset); 298b1b8ab34Slling 299cd9c78d9SLin Ling /* read in the gang block header */ 300cd9c78d9SLin Ling if (devread(sector, 0, SPA_GANGBLOCKSIZE, (char *)zio_gb) == 0) { 301cd9c78d9SLin Ling grub_printf("failed to read in a gang block header\n"); 302cd9c78d9SLin Ling return (1); 303cd9c78d9SLin Ling } 304cd9c78d9SLin Ling 305cd9c78d9SLin Ling /* self checksuming the gang block header */ 306cd9c78d9SLin Ling BP_ZERO(&tmpbp); 307cd9c78d9SLin Ling BP_SET_CHECKSUM(&tmpbp, ZIO_CHECKSUM_GANG_HEADER); 308cd9c78d9SLin Ling BP_SET_BYTEORDER(&tmpbp, ZFS_HOST_BYTEORDER); 309cd9c78d9SLin Ling ZIO_SET_CHECKSUM(&tmpbp.blk_cksum, DVA_GET_VDEV(dva), 310cd9c78d9SLin Ling DVA_GET_OFFSET(dva), bp->blk_birth, 0); 311cd9c78d9SLin Ling if (zio_checksum_verify(&tmpbp, (char *)zio_gb, SPA_GANGBLOCKSIZE)) { 312cd9c78d9SLin Ling grub_printf("failed to checksum a gang block header\n"); 313cd9c78d9SLin Ling return (1); 314cd9c78d9SLin Ling } 315cd9c78d9SLin Ling 316cd9c78d9SLin Ling for (i = 0; i < SPA_GBH_NBLKPTRS; i++) { 317cd9c78d9SLin Ling if (zio_gb->zg_blkptr[i].blk_birth == 0) 318cd9c78d9SLin Ling continue; 319cd9c78d9SLin Ling 320cd9c78d9SLin Ling if (zio_read_data(&zio_gb->zg_blkptr[i], buf, stack)) 321cd9c78d9SLin Ling return (1); 322cd9c78d9SLin Ling buf += BP_GET_PSIZE(&zio_gb->zg_blkptr[i]); 323cd9c78d9SLin Ling } 324cd9c78d9SLin Ling 325cd9c78d9SLin Ling return (0); 326cd9c78d9SLin Ling } 327cd9c78d9SLin Ling 328cd9c78d9SLin Ling /* 329cd9c78d9SLin Ling * Read in a block of raw data to buf. 330cd9c78d9SLin Ling * 331cd9c78d9SLin Ling * Return: 332cd9c78d9SLin Ling * 0 - success 333cd9c78d9SLin Ling * 1 - failure 334cd9c78d9SLin Ling */ 335cd9c78d9SLin Ling static int 336cd9c78d9SLin Ling zio_read_data(blkptr_t *bp, void *buf, char *stack) 337cd9c78d9SLin Ling { 338cd9c78d9SLin Ling int i, psize; 339cd9c78d9SLin Ling 340cd9c78d9SLin Ling psize = BP_GET_PSIZE(bp); 341ae8180dbSlling 342b1b8ab34Slling /* pick a good dva from the block pointer */ 343b1b8ab34Slling for (i = 0; i < SPA_DVAS_PER_BP; i++) { 344cd9c78d9SLin Ling uint64_t offset, sector; 345b1b8ab34Slling 346b1b8ab34Slling if (bp->blk_dva[i].dva_word[0] == 0 && 347b1b8ab34Slling bp->blk_dva[i].dva_word[1] == 0) 348b1b8ab34Slling continue; 349b1b8ab34Slling 350cd9c78d9SLin Ling if (DVA_GET_GANG(&bp->blk_dva[i])) { 351cd9c78d9SLin Ling if (zio_read_gang(bp, &bp->blk_dva[i], buf, stack) == 0) 352cd9c78d9SLin Ling return (0); 353b1b8ab34Slling } else { 354cd9c78d9SLin Ling /* read in a data block */ 355cd9c78d9SLin Ling offset = DVA_GET_OFFSET(&bp->blk_dva[i]); 356cd9c78d9SLin Ling sector = DVA_OFFSET_TO_PHYS_SECTOR(offset); 357cd9c78d9SLin Ling if (devread(sector, 0, psize, buf)) 358cd9c78d9SLin Ling return (0); 359b1b8ab34Slling } 360b1b8ab34Slling } 361b1b8ab34Slling 362cd9c78d9SLin Ling return (1); 363cd9c78d9SLin Ling } 364cd9c78d9SLin Ling 365cd9c78d9SLin Ling /* 366cd9c78d9SLin Ling * Read in a block of data, verify its checksum, decompress if needed, 367cd9c78d9SLin Ling * and put the uncompressed data in buf. 368cd9c78d9SLin Ling * 369cd9c78d9SLin Ling * Return: 370cd9c78d9SLin Ling * 0 - success 371cd9c78d9SLin Ling * errnum - failure 372cd9c78d9SLin Ling */ 373cd9c78d9SLin Ling static int 374cd9c78d9SLin Ling zio_read(blkptr_t *bp, void *buf, char *stack) 375cd9c78d9SLin Ling { 376cd9c78d9SLin Ling int lsize, psize, comp; 377cd9c78d9SLin Ling char *retbuf; 378cd9c78d9SLin Ling 379cd9c78d9SLin Ling comp = BP_GET_COMPRESS(bp); 380cd9c78d9SLin Ling lsize = BP_GET_LSIZE(bp); 381cd9c78d9SLin Ling psize = BP_GET_PSIZE(bp); 382cd9c78d9SLin Ling 383cd9c78d9SLin Ling if ((unsigned int)comp >= ZIO_COMPRESS_FUNCTIONS || 384cd9c78d9SLin Ling (comp != ZIO_COMPRESS_OFF && 385cd9c78d9SLin Ling decomp_table[comp].decomp_func == NULL)) { 386cd9c78d9SLin Ling grub_printf("compression algorithm not supported\n"); 387cd9c78d9SLin Ling return (ERR_FSYS_CORRUPT); 388cd9c78d9SLin Ling } 389cd9c78d9SLin Ling 390cd9c78d9SLin Ling if ((char *)buf < stack && ((char *)buf) + lsize > stack) { 391cd9c78d9SLin Ling grub_printf("not enough memory allocated\n"); 392cd9c78d9SLin Ling return (ERR_WONT_FIT); 393cd9c78d9SLin Ling } 394cd9c78d9SLin Ling 395cd9c78d9SLin Ling retbuf = buf; 396cd9c78d9SLin Ling if (comp != ZIO_COMPRESS_OFF) { 397cd9c78d9SLin Ling buf = stack; 398cd9c78d9SLin Ling stack += psize; 399cd9c78d9SLin Ling } 400cd9c78d9SLin Ling 401cd9c78d9SLin Ling if (zio_read_data(bp, buf, stack)) { 402cd9c78d9SLin Ling grub_printf("zio_read_data failed\n"); 403cd9c78d9SLin Ling return (ERR_FSYS_CORRUPT); 404cd9c78d9SLin Ling } 405cd9c78d9SLin Ling 406cd9c78d9SLin Ling if (zio_checksum_verify(bp, buf, psize) != 0) { 407cd9c78d9SLin Ling grub_printf("checksum verification failed\n"); 408cd9c78d9SLin Ling return (ERR_FSYS_CORRUPT); 409cd9c78d9SLin Ling } 410cd9c78d9SLin Ling 411cd9c78d9SLin Ling if (comp != ZIO_COMPRESS_OFF) 412cd9c78d9SLin Ling decomp_table[comp].decomp_func(buf, retbuf, psize, lsize); 413cd9c78d9SLin Ling 414cd9c78d9SLin Ling return (0); 415b1b8ab34Slling } 416b1b8ab34Slling 417b1b8ab34Slling /* 418b1b8ab34Slling * Get the block from a block id. 419b1b8ab34Slling * push the block onto the stack. 420b1b8ab34Slling * 421b1b8ab34Slling * Return: 422b1b8ab34Slling * 0 - success 423b1b8ab34Slling * errnum - failure 424b1b8ab34Slling */ 425b1b8ab34Slling static int 426b1b8ab34Slling dmu_read(dnode_phys_t *dn, uint64_t blkid, void *buf, char *stack) 427b1b8ab34Slling { 428b1b8ab34Slling int idx, level; 429b1b8ab34Slling blkptr_t *bp_array = dn->dn_blkptr; 430b1b8ab34Slling int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; 431b1b8ab34Slling blkptr_t *bp, *tmpbuf; 432b1b8ab34Slling 433b1b8ab34Slling bp = (blkptr_t *)stack; 434b1b8ab34Slling stack += sizeof (blkptr_t); 435b1b8ab34Slling 436b1b8ab34Slling tmpbuf = (blkptr_t *)stack; 437b1b8ab34Slling stack += 1<<dn->dn_indblkshift; 438b1b8ab34Slling 439b1b8ab34Slling for (level = dn->dn_nlevels - 1; level >= 0; level--) { 440b1b8ab34Slling idx = (blkid >> (epbs * level)) & ((1<<epbs)-1); 441b1b8ab34Slling *bp = bp_array[idx]; 442b1b8ab34Slling if (level == 0) 443b1b8ab34Slling tmpbuf = buf; 444ae8180dbSlling if (BP_IS_HOLE(bp)) { 445ae8180dbSlling grub_memset(buf, 0, 446ae8180dbSlling dn->dn_datablkszsec << SPA_MINBLOCKSHIFT); 447ae8180dbSlling break; 448ae8180dbSlling } else if (errnum = zio_read(bp, tmpbuf, stack)) { 449b1b8ab34Slling return (errnum); 450ae8180dbSlling } 451b1b8ab34Slling 452b1b8ab34Slling bp_array = tmpbuf; 453b1b8ab34Slling } 454b1b8ab34Slling 455b1b8ab34Slling return (0); 456b1b8ab34Slling } 457b1b8ab34Slling 458b1b8ab34Slling /* 459b1b8ab34Slling * mzap_lookup: Looks up property described by "name" and returns the value 460b1b8ab34Slling * in "value". 461b1b8ab34Slling * 462b1b8ab34Slling * Return: 463b1b8ab34Slling * 0 - success 464b1b8ab34Slling * errnum - failure 465b1b8ab34Slling */ 466b1b8ab34Slling static int 467b1b8ab34Slling mzap_lookup(mzap_phys_t *zapobj, int objsize, char *name, 468b1b8ab34Slling uint64_t *value) 469b1b8ab34Slling { 470b1b8ab34Slling int i, chunks; 471b1b8ab34Slling mzap_ent_phys_t *mzap_ent = zapobj->mz_chunk; 472b1b8ab34Slling 473b1b8ab34Slling chunks = objsize/MZAP_ENT_LEN - 1; 474b1b8ab34Slling for (i = 0; i < chunks; i++) { 475b1b8ab34Slling if (grub_strcmp(mzap_ent[i].mze_name, name) == 0) { 476b1b8ab34Slling *value = mzap_ent[i].mze_value; 477b1b8ab34Slling return (0); 478b1b8ab34Slling } 479b1b8ab34Slling } 480b1b8ab34Slling 481b1b8ab34Slling return (ERR_FSYS_CORRUPT); 482b1b8ab34Slling } 483b1b8ab34Slling 484b1b8ab34Slling static uint64_t 485b1b8ab34Slling zap_hash(uint64_t salt, const char *name) 486b1b8ab34Slling { 487b1b8ab34Slling static uint64_t table[256]; 488b1b8ab34Slling const uint8_t *cp; 489b1b8ab34Slling uint8_t c; 490b1b8ab34Slling uint64_t crc = salt; 491b1b8ab34Slling 492b1b8ab34Slling if (table[128] == 0) { 493b1b8ab34Slling uint64_t *ct; 494b1b8ab34Slling int i, j; 495b1b8ab34Slling for (i = 0; i < 256; i++) { 496b1b8ab34Slling for (ct = table + i, *ct = i, j = 8; j > 0; j--) 497b1b8ab34Slling *ct = (*ct >> 1) ^ (-(*ct & 1) & 498b1b8ab34Slling ZFS_CRC64_POLY); 499b1b8ab34Slling } 500b1b8ab34Slling } 501b1b8ab34Slling 502b1b8ab34Slling if (crc == 0 || table[128] != ZFS_CRC64_POLY) { 503b1b8ab34Slling errnum = ERR_FSYS_CORRUPT; 504b1b8ab34Slling return (0); 505b1b8ab34Slling } 506b1b8ab34Slling 507b1b8ab34Slling for (cp = (const uint8_t *)name; (c = *cp) != '\0'; cp++) 508b1b8ab34Slling crc = (crc >> 8) ^ table[(crc ^ c) & 0xFF]; 509b1b8ab34Slling 510b1b8ab34Slling /* 511b1b8ab34Slling * Only use 28 bits, since we need 4 bits in the cookie for the 512b1b8ab34Slling * collision differentiator. We MUST use the high bits, since 513b1b8ab34Slling * those are the onces that we first pay attention to when 514b1b8ab34Slling * chosing the bucket. 515b1b8ab34Slling */ 516b1b8ab34Slling crc &= ~((1ULL << (64 - ZAP_HASHBITS)) - 1); 517b1b8ab34Slling 518b1b8ab34Slling return (crc); 519b1b8ab34Slling } 520b1b8ab34Slling 521b1b8ab34Slling /* 522b1b8ab34Slling * Only to be used on 8-bit arrays. 523b1b8ab34Slling * array_len is actual len in bytes (not encoded le_value_length). 524b1b8ab34Slling * buf is null-terminated. 525b1b8ab34Slling */ 526b1b8ab34Slling static int 527b1b8ab34Slling zap_leaf_array_equal(zap_leaf_phys_t *l, int blksft, int chunk, 528b1b8ab34Slling int array_len, const char *buf) 529b1b8ab34Slling { 530b1b8ab34Slling int bseen = 0; 531b1b8ab34Slling 532b1b8ab34Slling while (bseen < array_len) { 533b1b8ab34Slling struct zap_leaf_array *la = 534b1b8ab34Slling &ZAP_LEAF_CHUNK(l, blksft, chunk).l_array; 535b1b8ab34Slling int toread = MIN(array_len - bseen, ZAP_LEAF_ARRAY_BYTES); 536b1b8ab34Slling 537b1b8ab34Slling if (chunk >= ZAP_LEAF_NUMCHUNKS(blksft)) 538b1b8ab34Slling return (0); 539b1b8ab34Slling 540b1b8ab34Slling if (zfs_bcmp(la->la_array, buf + bseen, toread) != 0) 541b1b8ab34Slling break; 542b1b8ab34Slling chunk = la->la_next; 543b1b8ab34Slling bseen += toread; 544b1b8ab34Slling } 545b1b8ab34Slling return (bseen == array_len); 546b1b8ab34Slling } 547b1b8ab34Slling 548b1b8ab34Slling /* 549b1b8ab34Slling * Given a zap_leaf_phys_t, walk thru the zap leaf chunks to get the 550b1b8ab34Slling * value for the property "name". 551b1b8ab34Slling * 552b1b8ab34Slling * Return: 553b1b8ab34Slling * 0 - success 554b1b8ab34Slling * errnum - failure 555b1b8ab34Slling */ 556e7cbe64fSgw static int 557b1b8ab34Slling zap_leaf_lookup(zap_leaf_phys_t *l, int blksft, uint64_t h, 558b1b8ab34Slling const char *name, uint64_t *value) 559b1b8ab34Slling { 560b1b8ab34Slling uint16_t chunk; 561b1b8ab34Slling struct zap_leaf_entry *le; 562b1b8ab34Slling 563b1b8ab34Slling /* Verify if this is a valid leaf block */ 564b1b8ab34Slling if (l->l_hdr.lh_block_type != ZBT_LEAF) 565b1b8ab34Slling return (ERR_FSYS_CORRUPT); 566b1b8ab34Slling if (l->l_hdr.lh_magic != ZAP_LEAF_MAGIC) 567b1b8ab34Slling return (ERR_FSYS_CORRUPT); 568b1b8ab34Slling 569b1b8ab34Slling for (chunk = l->l_hash[LEAF_HASH(blksft, h)]; 570b1b8ab34Slling chunk != CHAIN_END; chunk = le->le_next) { 571b1b8ab34Slling 572b1b8ab34Slling if (chunk >= ZAP_LEAF_NUMCHUNKS(blksft)) 573b1b8ab34Slling return (ERR_FSYS_CORRUPT); 574b1b8ab34Slling 575b1b8ab34Slling le = ZAP_LEAF_ENTRY(l, blksft, chunk); 576b1b8ab34Slling 577b1b8ab34Slling /* Verify the chunk entry */ 578b1b8ab34Slling if (le->le_type != ZAP_CHUNK_ENTRY) 579b1b8ab34Slling return (ERR_FSYS_CORRUPT); 580b1b8ab34Slling 581b1b8ab34Slling if (le->le_hash != h) 582b1b8ab34Slling continue; 583b1b8ab34Slling 584b1b8ab34Slling if (zap_leaf_array_equal(l, blksft, le->le_name_chunk, 585b1b8ab34Slling le->le_name_length, name)) { 586b1b8ab34Slling 587b1b8ab34Slling struct zap_leaf_array *la; 588b1b8ab34Slling uint8_t *ip; 589b1b8ab34Slling 590b1b8ab34Slling if (le->le_int_size != 8 || le->le_value_length != 1) 591e37b211cStaylor return (ERR_FSYS_CORRUPT); 592b1b8ab34Slling 593b1b8ab34Slling /* get the uint64_t property value */ 594b1b8ab34Slling la = &ZAP_LEAF_CHUNK(l, blksft, 595b1b8ab34Slling le->le_value_chunk).l_array; 596b1b8ab34Slling ip = la->la_array; 597b1b8ab34Slling 598b1b8ab34Slling *value = (uint64_t)ip[0] << 56 | (uint64_t)ip[1] << 48 | 599b1b8ab34Slling (uint64_t)ip[2] << 40 | (uint64_t)ip[3] << 32 | 600b1b8ab34Slling (uint64_t)ip[4] << 24 | (uint64_t)ip[5] << 16 | 601b1b8ab34Slling (uint64_t)ip[6] << 8 | (uint64_t)ip[7]; 602b1b8ab34Slling 603b1b8ab34Slling return (0); 604b1b8ab34Slling } 605b1b8ab34Slling } 606b1b8ab34Slling 607b1b8ab34Slling return (ERR_FSYS_CORRUPT); 608b1b8ab34Slling } 609b1b8ab34Slling 610b1b8ab34Slling /* 611b1b8ab34Slling * Fat ZAP lookup 612b1b8ab34Slling * 613b1b8ab34Slling * Return: 614b1b8ab34Slling * 0 - success 615b1b8ab34Slling * errnum - failure 616b1b8ab34Slling */ 617e7cbe64fSgw static int 618b1b8ab34Slling fzap_lookup(dnode_phys_t *zap_dnode, zap_phys_t *zap, 619b1b8ab34Slling char *name, uint64_t *value, char *stack) 620b1b8ab34Slling { 621b1b8ab34Slling zap_leaf_phys_t *l; 622b1b8ab34Slling uint64_t hash, idx, blkid; 623b1b8ab34Slling int blksft = zfs_log2(zap_dnode->dn_datablkszsec << DNODE_SHIFT); 624b1b8ab34Slling 625b1b8ab34Slling /* Verify if this is a fat zap header block */ 626b1b8ab34Slling if (zap->zap_magic != (uint64_t)ZAP_MAGIC) 627b1b8ab34Slling return (ERR_FSYS_CORRUPT); 628b1b8ab34Slling 629b1b8ab34Slling hash = zap_hash(zap->zap_salt, name); 630b1b8ab34Slling if (errnum) 631b1b8ab34Slling return (errnum); 632b1b8ab34Slling 633b1b8ab34Slling /* get block id from index */ 634b1b8ab34Slling if (zap->zap_ptrtbl.zt_numblks != 0) { 635b1b8ab34Slling /* external pointer tables not supported */ 636b1b8ab34Slling return (ERR_FSYS_CORRUPT); 637b1b8ab34Slling } 638b1b8ab34Slling idx = ZAP_HASH_IDX(hash, zap->zap_ptrtbl.zt_shift); 639b1b8ab34Slling blkid = ((uint64_t *)zap)[idx + (1<<(blksft-3-1))]; 640b1b8ab34Slling 641b1b8ab34Slling /* Get the leaf block */ 642b1b8ab34Slling l = (zap_leaf_phys_t *)stack; 643b1b8ab34Slling stack += 1<<blksft; 644051aabe6Staylor if ((1<<blksft) < sizeof (zap_leaf_phys_t)) 645e37b211cStaylor return (ERR_FSYS_CORRUPT); 646b1b8ab34Slling if (errnum = dmu_read(zap_dnode, blkid, l, stack)) 647b1b8ab34Slling return (errnum); 648b1b8ab34Slling 649b1b8ab34Slling return (zap_leaf_lookup(l, blksft, hash, name, value)); 650b1b8ab34Slling } 651b1b8ab34Slling 652b1b8ab34Slling /* 653b1b8ab34Slling * Read in the data of a zap object and find the value for a matching 654b1b8ab34Slling * property name. 655b1b8ab34Slling * 656b1b8ab34Slling * Return: 657b1b8ab34Slling * 0 - success 658b1b8ab34Slling * errnum - failure 659b1b8ab34Slling */ 660b1b8ab34Slling static int 661b1b8ab34Slling zap_lookup(dnode_phys_t *zap_dnode, char *name, uint64_t *val, char *stack) 662b1b8ab34Slling { 663b1b8ab34Slling uint64_t block_type; 664b1b8ab34Slling int size; 665b1b8ab34Slling void *zapbuf; 666b1b8ab34Slling 667b1b8ab34Slling /* Read in the first block of the zap object data. */ 668b1b8ab34Slling zapbuf = stack; 669b1b8ab34Slling size = zap_dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 670b1b8ab34Slling stack += size; 671b1b8ab34Slling if (errnum = dmu_read(zap_dnode, 0, zapbuf, stack)) 672b1b8ab34Slling return (errnum); 673b1b8ab34Slling 674b1b8ab34Slling block_type = *((uint64_t *)zapbuf); 675b1b8ab34Slling 676b1b8ab34Slling if (block_type == ZBT_MICRO) { 677b1b8ab34Slling return (mzap_lookup(zapbuf, size, name, val)); 678b1b8ab34Slling } else if (block_type == ZBT_HEADER) { 679b1b8ab34Slling /* this is a fat zap */ 680b1b8ab34Slling return (fzap_lookup(zap_dnode, zapbuf, name, 681b1b8ab34Slling val, stack)); 682b1b8ab34Slling } 683b1b8ab34Slling 684b1b8ab34Slling return (ERR_FSYS_CORRUPT); 685b1b8ab34Slling } 686b1b8ab34Slling 687b1b8ab34Slling /* 688b1b8ab34Slling * Get the dnode of an object number from the metadnode of an object set. 689b1b8ab34Slling * 690b1b8ab34Slling * Input 691b1b8ab34Slling * mdn - metadnode to get the object dnode 692b1b8ab34Slling * objnum - object number for the object dnode 693b1b8ab34Slling * buf - data buffer that holds the returning dnode 694b1b8ab34Slling * stack - scratch area 695b1b8ab34Slling * 696b1b8ab34Slling * Return: 697b1b8ab34Slling * 0 - success 698b1b8ab34Slling * errnum - failure 699b1b8ab34Slling */ 700b1b8ab34Slling static int 701b1b8ab34Slling dnode_get(dnode_phys_t *mdn, uint64_t objnum, uint8_t type, dnode_phys_t *buf, 702b1b8ab34Slling char *stack) 703b1b8ab34Slling { 704b1b8ab34Slling uint64_t blkid, blksz; /* the block id this object dnode is in */ 705b1b8ab34Slling int epbs; /* shift of number of dnodes in a block */ 706b1b8ab34Slling int idx; /* index within a block */ 707b1b8ab34Slling dnode_phys_t *dnbuf; 708b1b8ab34Slling 709b1b8ab34Slling blksz = mdn->dn_datablkszsec << SPA_MINBLOCKSHIFT; 710b1b8ab34Slling epbs = zfs_log2(blksz) - DNODE_SHIFT; 711b1b8ab34Slling blkid = objnum >> epbs; 712b1b8ab34Slling idx = objnum & ((1<<epbs)-1); 713b1b8ab34Slling 714b1b8ab34Slling if (dnode_buf != NULL && dnode_mdn == mdn && 715b1b8ab34Slling objnum >= dnode_start && objnum < dnode_end) { 716b1b8ab34Slling grub_memmove(buf, &dnode_buf[idx], DNODE_SIZE); 717b1b8ab34Slling VERIFY_DN_TYPE(buf, type); 718b1b8ab34Slling return (0); 719b1b8ab34Slling } 720b1b8ab34Slling 721b1b8ab34Slling if (dnode_buf && blksz == 1<<DNODE_BLOCK_SHIFT) { 722b1b8ab34Slling dnbuf = dnode_buf; 723b1b8ab34Slling dnode_mdn = mdn; 724b1b8ab34Slling dnode_start = blkid << epbs; 725b1b8ab34Slling dnode_end = (blkid + 1) << epbs; 726b1b8ab34Slling } else { 727b1b8ab34Slling dnbuf = (dnode_phys_t *)stack; 728b1b8ab34Slling stack += blksz; 729b1b8ab34Slling } 730b1b8ab34Slling 731b1b8ab34Slling if (errnum = dmu_read(mdn, blkid, (char *)dnbuf, stack)) 732b1b8ab34Slling return (errnum); 733b1b8ab34Slling 734b1b8ab34Slling grub_memmove(buf, &dnbuf[idx], DNODE_SIZE); 735b1b8ab34Slling VERIFY_DN_TYPE(buf, type); 736b1b8ab34Slling 737b1b8ab34Slling return (0); 738b1b8ab34Slling } 739b1b8ab34Slling 740b1b8ab34Slling /* 741eb2bd662Svikram * Check if this is a special file that resides at the top 742eb2bd662Svikram * dataset of the pool. Currently this is the GRUB menu, 743eb2bd662Svikram * boot signature and boot signature backup. 744b1b8ab34Slling * str starts with '/'. 745b1b8ab34Slling */ 746b1b8ab34Slling static int 747eb2bd662Svikram is_top_dataset_file(char *str) 748b1b8ab34Slling { 749b1b8ab34Slling char *tptr; 750b1b8ab34Slling 751b1b8ab34Slling if ((tptr = grub_strstr(str, "menu.lst")) && 752b1b8ab34Slling (tptr[8] == '\0' || tptr[8] == ' ') && 753b1b8ab34Slling *(tptr-1) == '/') 754b1b8ab34Slling return (1); 755b1b8ab34Slling 756eb2bd662Svikram if (grub_strncmp(str, BOOTSIGN_DIR"/", 7571183b401Svikram grub_strlen(BOOTSIGN_DIR) + 1) == 0) 758eb2bd662Svikram return (1); 759eb2bd662Svikram 760eb2bd662Svikram if (grub_strcmp(str, BOOTSIGN_BACKUP) == 0) 761eb2bd662Svikram return (1); 762eb2bd662Svikram 763b1b8ab34Slling return (0); 764b1b8ab34Slling } 765b1b8ab34Slling 766b1b8ab34Slling /* 767b1b8ab34Slling * Get the file dnode for a given file name where mdn is the meta dnode 768b1b8ab34Slling * for this ZFS object set. When found, place the file dnode in dn. 769b1b8ab34Slling * The 'path' argument will be mangled. 770b1b8ab34Slling * 771b1b8ab34Slling * Return: 772b1b8ab34Slling * 0 - success 773b1b8ab34Slling * errnum - failure 774b1b8ab34Slling */ 775b1b8ab34Slling static int 776b1b8ab34Slling dnode_get_path(dnode_phys_t *mdn, char *path, dnode_phys_t *dn, 777b1b8ab34Slling char *stack) 778b1b8ab34Slling { 779e7437265Sahrens uint64_t objnum, version; 780b1b8ab34Slling char *cname, ch; 781b1b8ab34Slling 782b1b8ab34Slling if (errnum = dnode_get(mdn, MASTER_NODE_OBJ, DMU_OT_MASTER_NODE, 783b1b8ab34Slling dn, stack)) 784b1b8ab34Slling return (errnum); 785b1b8ab34Slling 786e7437265Sahrens if (errnum = zap_lookup(dn, ZPL_VERSION_STR, &version, stack)) 787e7437265Sahrens return (errnum); 788e7437265Sahrens if (version > ZPL_VERSION) 789e7437265Sahrens return (-1); 790e7437265Sahrens 791b1b8ab34Slling if (errnum = zap_lookup(dn, ZFS_ROOT_OBJ, &objnum, stack)) 792b1b8ab34Slling return (errnum); 793b1b8ab34Slling 794b1b8ab34Slling if (errnum = dnode_get(mdn, objnum, DMU_OT_DIRECTORY_CONTENTS, 795b1b8ab34Slling dn, stack)) 796b1b8ab34Slling return (errnum); 797b1b8ab34Slling 798b1b8ab34Slling /* skip leading slashes */ 799b1b8ab34Slling while (*path == '/') 800b1b8ab34Slling path++; 801b1b8ab34Slling 802b1b8ab34Slling while (*path && !isspace(*path)) { 803b1b8ab34Slling 804b1b8ab34Slling /* get the next component name */ 805b1b8ab34Slling cname = path; 806b1b8ab34Slling while (*path && !isspace(*path) && *path != '/') 807b1b8ab34Slling path++; 808b1b8ab34Slling ch = *path; 809b1b8ab34Slling *path = 0; /* ensure null termination */ 810b1b8ab34Slling 811b1b8ab34Slling if (errnum = zap_lookup(dn, cname, &objnum, stack)) 812b1b8ab34Slling return (errnum); 813b1b8ab34Slling 814e7437265Sahrens objnum = ZFS_DIRENT_OBJ(objnum); 815b1b8ab34Slling if (errnum = dnode_get(mdn, objnum, 0, dn, stack)) 816b1b8ab34Slling return (errnum); 817b1b8ab34Slling 818b1b8ab34Slling *path = ch; 819b1b8ab34Slling while (*path == '/') 820b1b8ab34Slling path++; 821b1b8ab34Slling } 822b1b8ab34Slling 823b1b8ab34Slling /* We found the dnode for this file. Verify if it is a plain file. */ 824b1b8ab34Slling VERIFY_DN_TYPE(dn, DMU_OT_PLAIN_FILE_CONTENTS); 825b1b8ab34Slling 826b1b8ab34Slling return (0); 827b1b8ab34Slling } 828b1b8ab34Slling 829b1b8ab34Slling /* 830b1b8ab34Slling * Get the default 'bootfs' property value from the rootpool. 831b1b8ab34Slling * 832b1b8ab34Slling * Return: 833b1b8ab34Slling * 0 - success 834b1b8ab34Slling * errnum -failure 835b1b8ab34Slling */ 836b1b8ab34Slling static int 837b1b8ab34Slling get_default_bootfsobj(dnode_phys_t *mosmdn, uint64_t *obj, char *stack) 838b1b8ab34Slling { 839b1b8ab34Slling uint64_t objnum = 0; 840b1b8ab34Slling dnode_phys_t *dn = (dnode_phys_t *)stack; 841b1b8ab34Slling stack += DNODE_SIZE; 842b1b8ab34Slling 843ae8180dbSlling if (errnum = dnode_get(mosmdn, DMU_POOL_DIRECTORY_OBJECT, 844b1b8ab34Slling DMU_OT_OBJECT_DIRECTORY, dn, stack)) 845ae8180dbSlling return (errnum); 846b1b8ab34Slling 847b1b8ab34Slling /* 848b1b8ab34Slling * find the object number for 'pool_props', and get the dnode 849b1b8ab34Slling * of the 'pool_props'. 850b1b8ab34Slling */ 851b1b8ab34Slling if (zap_lookup(dn, DMU_POOL_PROPS, &objnum, stack)) 852b1b8ab34Slling return (ERR_FILESYSTEM_NOT_FOUND); 853b1b8ab34Slling 854ae8180dbSlling if (errnum = dnode_get(mosmdn, objnum, DMU_OT_POOL_PROPS, dn, stack)) 855ae8180dbSlling return (errnum); 856b1b8ab34Slling 857b1b8ab34Slling if (zap_lookup(dn, ZPOOL_PROP_BOOTFS, &objnum, stack)) 858b1b8ab34Slling return (ERR_FILESYSTEM_NOT_FOUND); 859b1b8ab34Slling 860b1b8ab34Slling if (!objnum) 861b1b8ab34Slling return (ERR_FILESYSTEM_NOT_FOUND); 862b1b8ab34Slling 863b1b8ab34Slling *obj = objnum; 864b1b8ab34Slling return (0); 865b1b8ab34Slling } 866b1b8ab34Slling 867b1b8ab34Slling /* 868b1b8ab34Slling * Given a MOS metadnode, get the metadnode of a given filesystem name (fsname), 869b1b8ab34Slling * e.g. pool/rootfs, or a given object number (obj), e.g. the object number 870b1b8ab34Slling * of pool/rootfs. 871b1b8ab34Slling * 872b1b8ab34Slling * If no fsname and no obj are given, return the DSL_DIR metadnode. 873b1b8ab34Slling * If fsname is given, return its metadnode and its matching object number. 874b1b8ab34Slling * If only obj is given, return the metadnode for this object number. 875b1b8ab34Slling * 876b1b8ab34Slling * Return: 877b1b8ab34Slling * 0 - success 878b1b8ab34Slling * errnum - failure 879b1b8ab34Slling */ 880b1b8ab34Slling static int 881b1b8ab34Slling get_objset_mdn(dnode_phys_t *mosmdn, char *fsname, uint64_t *obj, 882b1b8ab34Slling dnode_phys_t *mdn, char *stack) 883b1b8ab34Slling { 884b1b8ab34Slling uint64_t objnum, headobj; 885b1b8ab34Slling char *cname, ch; 886b1b8ab34Slling blkptr_t *bp; 887b1b8ab34Slling objset_phys_t *osp; 888fe3e2633SEric Taylor int issnapshot = 0; 889fe3e2633SEric Taylor char *snapname; 890b1b8ab34Slling 891b1b8ab34Slling if (fsname == NULL && obj) { 892b1b8ab34Slling headobj = *obj; 893b1b8ab34Slling goto skip; 894b1b8ab34Slling } 895b1b8ab34Slling 896b1b8ab34Slling if (errnum = dnode_get(mosmdn, DMU_POOL_DIRECTORY_OBJECT, 897b1b8ab34Slling DMU_OT_OBJECT_DIRECTORY, mdn, stack)) 898b1b8ab34Slling return (errnum); 899b1b8ab34Slling 900b1b8ab34Slling if (errnum = zap_lookup(mdn, DMU_POOL_ROOT_DATASET, &objnum, 901b1b8ab34Slling stack)) 902b1b8ab34Slling return (errnum); 903b1b8ab34Slling 904b1b8ab34Slling if (errnum = dnode_get(mosmdn, objnum, DMU_OT_DSL_DIR, mdn, stack)) 905b1b8ab34Slling return (errnum); 906b1b8ab34Slling 907b1b8ab34Slling if (fsname == NULL) { 908b1b8ab34Slling headobj = 909b1b8ab34Slling ((dsl_dir_phys_t *)DN_BONUS(mdn))->dd_head_dataset_obj; 910b1b8ab34Slling goto skip; 911b1b8ab34Slling } 912b1b8ab34Slling 913b1b8ab34Slling /* take out the pool name */ 914b1b8ab34Slling while (*fsname && !isspace(*fsname) && *fsname != '/') 915b1b8ab34Slling fsname++; 916b1b8ab34Slling 917b1b8ab34Slling while (*fsname && !isspace(*fsname)) { 918b1b8ab34Slling uint64_t childobj; 919b1b8ab34Slling 920b1b8ab34Slling while (*fsname == '/') 921b1b8ab34Slling fsname++; 922b1b8ab34Slling 923b1b8ab34Slling cname = fsname; 924b1b8ab34Slling while (*fsname && !isspace(*fsname) && *fsname != '/') 925b1b8ab34Slling fsname++; 926b1b8ab34Slling ch = *fsname; 927b1b8ab34Slling *fsname = 0; 928b1b8ab34Slling 929fe3e2633SEric Taylor snapname = cname; 930fe3e2633SEric Taylor while (*snapname && !isspace(*snapname) && *snapname != '@') 931fe3e2633SEric Taylor snapname++; 932fe3e2633SEric Taylor if (*snapname == '@') { 933fe3e2633SEric Taylor issnapshot = 1; 934fe3e2633SEric Taylor *snapname = 0; 935fe3e2633SEric Taylor } 936b1b8ab34Slling childobj = 937b1b8ab34Slling ((dsl_dir_phys_t *)DN_BONUS(mdn))->dd_child_dir_zapobj; 938b1b8ab34Slling if (errnum = dnode_get(mosmdn, childobj, 939b1b8ab34Slling DMU_OT_DSL_DIR_CHILD_MAP, mdn, stack)) 940b1b8ab34Slling return (errnum); 941b1b8ab34Slling 942ae8180dbSlling if (zap_lookup(mdn, cname, &objnum, stack)) 943ae8180dbSlling return (ERR_FILESYSTEM_NOT_FOUND); 944b1b8ab34Slling 945b1b8ab34Slling if (errnum = dnode_get(mosmdn, objnum, DMU_OT_DSL_DIR, 946b1b8ab34Slling mdn, stack)) 947b1b8ab34Slling return (errnum); 948b1b8ab34Slling 949b1b8ab34Slling *fsname = ch; 950fe3e2633SEric Taylor if (issnapshot) 951fe3e2633SEric Taylor *snapname = '@'; 952b1b8ab34Slling } 953b1b8ab34Slling headobj = ((dsl_dir_phys_t *)DN_BONUS(mdn))->dd_head_dataset_obj; 954b1b8ab34Slling if (obj) 955b1b8ab34Slling *obj = headobj; 956b1b8ab34Slling 957b1b8ab34Slling skip: 958b1b8ab34Slling if (errnum = dnode_get(mosmdn, headobj, DMU_OT_DSL_DATASET, mdn, stack)) 959b1b8ab34Slling return (errnum); 960fe3e2633SEric Taylor if (issnapshot) { 961fe3e2633SEric Taylor uint64_t snapobj; 962fe3e2633SEric Taylor 963fe3e2633SEric Taylor snapobj = ((dsl_dataset_phys_t *)DN_BONUS(mdn))-> 964fe3e2633SEric Taylor ds_snapnames_zapobj; 965b1b8ab34Slling 966fe3e2633SEric Taylor if (errnum = dnode_get(mosmdn, snapobj, 967fe3e2633SEric Taylor DMU_OT_DSL_DS_SNAP_MAP, mdn, stack)) 968fe3e2633SEric Taylor return (errnum); 969fe3e2633SEric Taylor if (zap_lookup(mdn, snapname + 1, &headobj, stack)) 970fe3e2633SEric Taylor return (ERR_FILESYSTEM_NOT_FOUND); 971fe3e2633SEric Taylor if (errnum = dnode_get(mosmdn, headobj, 972fe3e2633SEric Taylor DMU_OT_DSL_DATASET, mdn, stack)) 973fe3e2633SEric Taylor return (errnum); 974fe3e2633SEric Taylor if (obj) 975fe3e2633SEric Taylor *obj = headobj; 976fe3e2633SEric Taylor } 977b1b8ab34Slling 978b1b8ab34Slling bp = &((dsl_dataset_phys_t *)DN_BONUS(mdn))->ds_bp; 979b1b8ab34Slling osp = (objset_phys_t *)stack; 980b1b8ab34Slling stack += sizeof (objset_phys_t); 981b1b8ab34Slling if (errnum = zio_read(bp, osp, stack)) 982b1b8ab34Slling return (errnum); 983b1b8ab34Slling 984b1b8ab34Slling grub_memmove((char *)mdn, (char *)&osp->os_meta_dnode, DNODE_SIZE); 985b1b8ab34Slling 986b1b8ab34Slling return (0); 987b1b8ab34Slling } 988b1b8ab34Slling 989b1b8ab34Slling /* 990e7cbe64fSgw * For a given XDR packed nvlist, verify the first 4 bytes and move on. 991b1b8ab34Slling * 992e7cbe64fSgw * An XDR packed nvlist is encoded as (comments from nvs_xdr_create) : 993b1b8ab34Slling * 994b1b8ab34Slling * encoding method/host endian (4 bytes) 995b1b8ab34Slling * nvl_version (4 bytes) 996b1b8ab34Slling * nvl_nvflag (4 bytes) 997b1b8ab34Slling * encoded nvpairs: 998b1b8ab34Slling * encoded size of the nvpair (4 bytes) 999b1b8ab34Slling * decoded size of the nvpair (4 bytes) 1000b1b8ab34Slling * name string size (4 bytes) 1001b1b8ab34Slling * name string data (sizeof(NV_ALIGN4(string)) 1002b1b8ab34Slling * data type (4 bytes) 1003b1b8ab34Slling * # of elements in the nvpair (4 bytes) 1004b1b8ab34Slling * data 1005b1b8ab34Slling * 2 zero's for the last nvpair 1006b1b8ab34Slling * (end of the entire list) (8 bytes) 1007b1b8ab34Slling * 1008b1b8ab34Slling * Return: 1009b1b8ab34Slling * 0 - success 1010b1b8ab34Slling * 1 - failure 1011b1b8ab34Slling */ 1012e7cbe64fSgw static int 1013e7cbe64fSgw nvlist_unpack(char *nvlist, char **out) 1014b1b8ab34Slling { 1015b1b8ab34Slling /* Verify if the 1st and 2nd byte in the nvlist are valid. */ 1016b1b8ab34Slling if (nvlist[0] != NV_ENCODE_XDR || nvlist[1] != HOST_ENDIAN) 1017b1b8ab34Slling return (1); 1018b1b8ab34Slling 1019e7cbe64fSgw nvlist += 4; 1020e7cbe64fSgw *out = nvlist; 1021e7cbe64fSgw return (0); 1022e7cbe64fSgw } 1023e7cbe64fSgw 1024e7cbe64fSgw static char * 1025e7cbe64fSgw nvlist_array(char *nvlist, int index) 1026e7cbe64fSgw { 1027e7cbe64fSgw int i, encode_size; 1028e7cbe64fSgw 1029e7cbe64fSgw for (i = 0; i < index; i++) { 1030e7cbe64fSgw /* skip the header, nvl_version, and nvl_nvflag */ 1031e7cbe64fSgw nvlist = nvlist + 4 * 2; 1032e7cbe64fSgw 1033e7cbe64fSgw while (encode_size = BSWAP_32(*(uint32_t *)nvlist)) 1034e7cbe64fSgw nvlist += encode_size; /* goto the next nvpair */ 1035e7cbe64fSgw 1036e7cbe64fSgw nvlist = nvlist + 4 * 2; /* skip the ending 2 zeros - 8 bytes */ 1037e7cbe64fSgw } 1038e7cbe64fSgw 1039e7cbe64fSgw return (nvlist); 1040e7cbe64fSgw } 1041e7cbe64fSgw 1042e7cbe64fSgw static int 1043e7cbe64fSgw nvlist_lookup_value(char *nvlist, char *name, void *val, int valtype, 1044e7cbe64fSgw int *nelmp) 1045e7cbe64fSgw { 1046e7cbe64fSgw int name_len, type, slen, encode_size; 1047e7cbe64fSgw char *nvpair, *nvp_name, *strval = val; 1048e7cbe64fSgw uint64_t *intval = val; 1049e7cbe64fSgw 1050b1b8ab34Slling /* skip the header, nvl_version, and nvl_nvflag */ 1051e7cbe64fSgw nvlist = nvlist + 4 * 2; 1052b1b8ab34Slling 1053b1b8ab34Slling /* 1054b1b8ab34Slling * Loop thru the nvpair list 1055b1b8ab34Slling * The XDR representation of an integer is in big-endian byte order. 1056b1b8ab34Slling */ 1057b1b8ab34Slling while (encode_size = BSWAP_32(*(uint32_t *)nvlist)) { 1058b1b8ab34Slling 1059b1b8ab34Slling nvpair = nvlist + 4 * 2; /* skip the encode/decode size */ 1060b1b8ab34Slling 1061b1b8ab34Slling name_len = BSWAP_32(*(uint32_t *)nvpair); 1062b1b8ab34Slling nvpair += 4; 1063b1b8ab34Slling 1064b1b8ab34Slling nvp_name = nvpair; 1065b1b8ab34Slling nvpair = nvpair + ((name_len + 3) & ~3); /* align */ 1066b1b8ab34Slling 1067b1b8ab34Slling type = BSWAP_32(*(uint32_t *)nvpair); 1068b1b8ab34Slling nvpair += 4; 1069b1b8ab34Slling 1070b1b8ab34Slling if ((grub_strncmp(nvp_name, name, name_len) == 0) && 1071b1b8ab34Slling type == valtype) { 1072e7cbe64fSgw int nelm; 1073b1b8ab34Slling 1074e7cbe64fSgw if ((nelm = BSWAP_32(*(uint32_t *)nvpair)) < 1) 1075b1b8ab34Slling return (1); 1076b1b8ab34Slling nvpair += 4; 1077b1b8ab34Slling 1078b1b8ab34Slling switch (valtype) { 1079b1b8ab34Slling case DATA_TYPE_STRING: 1080b1b8ab34Slling slen = BSWAP_32(*(uint32_t *)nvpair); 1081b1b8ab34Slling nvpair += 4; 1082b1b8ab34Slling grub_memmove(strval, nvpair, slen); 1083b1b8ab34Slling strval[slen] = '\0'; 1084b1b8ab34Slling return (0); 1085b1b8ab34Slling 1086b1b8ab34Slling case DATA_TYPE_UINT64: 1087b1b8ab34Slling *intval = BSWAP_64(*(uint64_t *)nvpair); 1088b1b8ab34Slling return (0); 1089e7cbe64fSgw 1090e7cbe64fSgw case DATA_TYPE_NVLIST: 1091e7cbe64fSgw *(void **)val = (void *)nvpair; 1092e7cbe64fSgw return (0); 1093e7cbe64fSgw 1094e7cbe64fSgw case DATA_TYPE_NVLIST_ARRAY: 1095e7cbe64fSgw *(void **)val = (void *)nvpair; 1096e7cbe64fSgw if (nelmp) 1097e7cbe64fSgw *nelmp = nelm; 1098e7cbe64fSgw return (0); 1099b1b8ab34Slling } 1100b1b8ab34Slling } 1101b1b8ab34Slling 1102b1b8ab34Slling nvlist += encode_size; /* goto the next nvpair */ 1103b1b8ab34Slling } 1104b1b8ab34Slling 1105b1b8ab34Slling return (1); 1106b1b8ab34Slling } 1107b1b8ab34Slling 1108b1b8ab34Slling /* 1109e7cbe64fSgw * Check if this vdev is online and is in a good state. 1110e7cbe64fSgw */ 1111e7cbe64fSgw static int 1112e7cbe64fSgw vdev_validate(char *nv) 1113e7cbe64fSgw { 1114e7cbe64fSgw uint64_t ival; 1115e7cbe64fSgw 1116e7cbe64fSgw if (nvlist_lookup_value(nv, ZPOOL_CONFIG_OFFLINE, &ival, 1117e7cbe64fSgw DATA_TYPE_UINT64, NULL) == 0 || 1118e7cbe64fSgw nvlist_lookup_value(nv, ZPOOL_CONFIG_FAULTED, &ival, 1119e7cbe64fSgw DATA_TYPE_UINT64, NULL) == 0 || 1120e7cbe64fSgw nvlist_lookup_value(nv, ZPOOL_CONFIG_REMOVED, &ival, 1121e7cbe64fSgw DATA_TYPE_UINT64, NULL) == 0) 1122e7cbe64fSgw return (ERR_DEV_VALUES); 1123e7cbe64fSgw 1124e7cbe64fSgw return (0); 1125e7cbe64fSgw } 1126e7cbe64fSgw 1127e7cbe64fSgw /* 112821ecdf64SLin Ling * Get a valid vdev pathname/devid from the boot device. 1129ffb5616eSLin Ling * The caller should already allocate MAXPATHLEN memory for bootpath and devid. 1130e7cbe64fSgw */ 113121ecdf64SLin Ling static int 113221ecdf64SLin Ling vdev_get_bootpath(char *nv, uint64_t inguid, char *devid, char *bootpath, 113321ecdf64SLin Ling int is_spare) 1134e7cbe64fSgw { 1135e7cbe64fSgw char type[16]; 1136e7cbe64fSgw 1137e7cbe64fSgw if (nvlist_lookup_value(nv, ZPOOL_CONFIG_TYPE, &type, DATA_TYPE_STRING, 1138e7cbe64fSgw NULL)) 1139e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1140e7cbe64fSgw 1141e7cbe64fSgw if (strcmp(type, VDEV_TYPE_DISK) == 0) { 1142ffb5616eSLin Ling uint64_t guid; 1143ffb5616eSLin Ling 1144ffb5616eSLin Ling if (vdev_validate(nv) != 0) 1145ffb5616eSLin Ling return (ERR_NO_BOOTPATH); 1146ffb5616eSLin Ling 1147ffb5616eSLin Ling if (nvlist_lookup_value(nv, ZPOOL_CONFIG_GUID, 1148ffb5616eSLin Ling &guid, DATA_TYPE_UINT64, NULL) != 0) 1149ffb5616eSLin Ling return (ERR_NO_BOOTPATH); 1150ffb5616eSLin Ling 1151ffb5616eSLin Ling if (guid != inguid) 1152e7cbe64fSgw return (ERR_NO_BOOTPATH); 1153e7cbe64fSgw 115421ecdf64SLin Ling /* for a spare vdev, pick the disk labeled with "is_spare" */ 115521ecdf64SLin Ling if (is_spare) { 115621ecdf64SLin Ling uint64_t spare = 0; 115721ecdf64SLin Ling (void) nvlist_lookup_value(nv, ZPOOL_CONFIG_IS_SPARE, 115821ecdf64SLin Ling &spare, DATA_TYPE_UINT64, NULL); 115921ecdf64SLin Ling if (!spare) 116021ecdf64SLin Ling return (ERR_NO_BOOTPATH); 116121ecdf64SLin Ling } 116221ecdf64SLin Ling 1163ffb5616eSLin Ling if (nvlist_lookup_value(nv, ZPOOL_CONFIG_PHYS_PATH, 1164ffb5616eSLin Ling bootpath, DATA_TYPE_STRING, NULL) != 0) 1165ffb5616eSLin Ling bootpath[0] = '\0'; 1166ffb5616eSLin Ling 1167ffb5616eSLin Ling if (nvlist_lookup_value(nv, ZPOOL_CONFIG_DEVID, 1168ffb5616eSLin Ling devid, DATA_TYPE_STRING, NULL) != 0) 1169ffb5616eSLin Ling devid[0] = '\0'; 1170ffb5616eSLin Ling 1171ffb5616eSLin Ling if (strlen(bootpath) >= MAXPATHLEN || 1172ffb5616eSLin Ling strlen(devid) >= MAXPATHLEN) 1173ffb5616eSLin Ling return (ERR_WONT_FIT); 1174ffb5616eSLin Ling 1175ffb5616eSLin Ling return (0); 1176ffb5616eSLin Ling 117721ecdf64SLin Ling } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || 117821ecdf64SLin Ling strcmp(type, VDEV_TYPE_REPLACING) == 0 || 117921ecdf64SLin Ling (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { 1180e7cbe64fSgw int nelm, i; 1181e7cbe64fSgw char *child; 1182e7cbe64fSgw 1183e7cbe64fSgw if (nvlist_lookup_value(nv, ZPOOL_CONFIG_CHILDREN, &child, 1184e7cbe64fSgw DATA_TYPE_NVLIST_ARRAY, &nelm)) 1185e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1186e7cbe64fSgw 1187e7cbe64fSgw for (i = 0; i < nelm; i++) { 1188e7cbe64fSgw char *child_i; 1189e7cbe64fSgw 1190e7cbe64fSgw child_i = nvlist_array(child, i); 1191ffb5616eSLin Ling if (vdev_get_bootpath(child_i, inguid, devid, 119221ecdf64SLin Ling bootpath, is_spare) == 0) 1193ffb5616eSLin Ling return (0); 1194e7cbe64fSgw } 1195e7cbe64fSgw } 1196e7cbe64fSgw 1197ffb5616eSLin Ling return (ERR_NO_BOOTPATH); 1198e7cbe64fSgw } 1199e7cbe64fSgw 1200e7cbe64fSgw /* 1201e7cbe64fSgw * Check the disk label information and retrieve needed vdev name-value pairs. 1202b1b8ab34Slling * 1203b1b8ab34Slling * Return: 1204b1b8ab34Slling * 0 - success 1205e7cbe64fSgw * ERR_* - failure 1206b1b8ab34Slling */ 1207051aabe6Staylor int 1208*e23347b1SEric Taylor check_pool_label(uint64_t sector, char *stack, char *outdevid, 1209*e23347b1SEric Taylor char *outpath, uint64_t *outguid) 1210b1b8ab34Slling { 1211b1b8ab34Slling vdev_phys_t *vdev; 1212*e23347b1SEric Taylor uint64_t pool_state, txg = 0; 1213e7cbe64fSgw char *nvlist, *nv; 1214051aabe6Staylor uint64_t diskguid; 1215fe3e2633SEric Taylor uint64_t version; 1216b1b8ab34Slling 1217*e23347b1SEric Taylor sector += (VDEV_SKIP_SIZE >> SPA_MINBLOCKSHIFT); 1218b1b8ab34Slling 1219b1b8ab34Slling /* Read in the vdev name-value pair list (112K). */ 1220b1b8ab34Slling if (devread(sector, 0, VDEV_PHYS_SIZE, stack) == 0) 1221b1b8ab34Slling return (ERR_READ); 1222b1b8ab34Slling 1223b1b8ab34Slling vdev = (vdev_phys_t *)stack; 1224e4c3b53dStaylor stack += sizeof (vdev_phys_t); 1225b1b8ab34Slling 1226e7cbe64fSgw if (nvlist_unpack(vdev->vp_nvlist, &nvlist)) 1227b1b8ab34Slling return (ERR_FSYS_CORRUPT); 1228e7cbe64fSgw 1229e7cbe64fSgw if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_POOL_STATE, &pool_state, 1230e7cbe64fSgw DATA_TYPE_UINT64, NULL)) 1231e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1232e7cbe64fSgw 1233e7cbe64fSgw if (pool_state == POOL_STATE_DESTROYED) 1234e7cbe64fSgw return (ERR_FILESYSTEM_NOT_FOUND); 1235e7cbe64fSgw 1236e7cbe64fSgw if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_POOL_NAME, 1237e7cbe64fSgw current_rootpool, DATA_TYPE_STRING, NULL)) 1238e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1239e7cbe64fSgw 1240e7cbe64fSgw if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_POOL_TXG, &txg, 1241e7cbe64fSgw DATA_TYPE_UINT64, NULL)) 1242e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1243e7cbe64fSgw 1244e7cbe64fSgw /* not an active device */ 1245e7cbe64fSgw if (txg == 0) 1246e7cbe64fSgw return (ERR_NO_BOOTPATH); 1247e7cbe64fSgw 1248fe3e2633SEric Taylor if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_VERSION, &version, 1249fe3e2633SEric Taylor DATA_TYPE_UINT64, NULL)) 1250fe3e2633SEric Taylor return (ERR_FSYS_CORRUPT); 1251fe3e2633SEric Taylor if (version > SPA_VERSION) 1252fe3e2633SEric Taylor return (ERR_NEWER_VERSION); 1253e7cbe64fSgw if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_VDEV_TREE, &nv, 1254e7cbe64fSgw DATA_TYPE_NVLIST, NULL)) 1255e7cbe64fSgw return (ERR_FSYS_CORRUPT); 1256051aabe6Staylor if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_GUID, &diskguid, 1257051aabe6Staylor DATA_TYPE_UINT64, NULL)) 1258051aabe6Staylor return (ERR_FSYS_CORRUPT); 125921ecdf64SLin Ling if (vdev_get_bootpath(nv, diskguid, outdevid, outpath, 0)) 1260e7cbe64fSgw return (ERR_NO_BOOTPATH); 1261*e23347b1SEric Taylor if (nvlist_lookup_value(nvlist, ZPOOL_CONFIG_POOL_GUID, outguid, 1262*e23347b1SEric Taylor DATA_TYPE_UINT64, NULL)) 1263*e23347b1SEric Taylor return (ERR_FSYS_CORRUPT); 1264e7cbe64fSgw return (0); 1265b1b8ab34Slling } 1266b1b8ab34Slling 1267b1b8ab34Slling /* 1268b1b8ab34Slling * zfs_mount() locates a valid uberblock of the root pool and read in its MOS 1269b1b8ab34Slling * to the memory address MOS. 1270b1b8ab34Slling * 1271b1b8ab34Slling * Return: 1272b1b8ab34Slling * 1 - success 1273b1b8ab34Slling * 0 - failure 1274b1b8ab34Slling */ 1275b1b8ab34Slling int 1276b1b8ab34Slling zfs_mount(void) 1277b1b8ab34Slling { 1278b1b8ab34Slling char *stack; 1279b1b8ab34Slling int label = 0; 1280*e23347b1SEric Taylor uberblock_phys_t *ub_array, *ubbest; 1281b1b8ab34Slling objset_phys_t *osp; 1282051aabe6Staylor char tmp_bootpath[MAXNAMELEN]; 1283051aabe6Staylor char tmp_devid[MAXNAMELEN]; 1284*e23347b1SEric Taylor uint64_t tmp_guid; 1285*e23347b1SEric Taylor uint64_t adjpl = (uint64_t)part_length << SPA_MINBLOCKSHIFT; 1286051aabe6Staylor 1287051aabe6Staylor /* if it's our first time here, zero the best uberblock out */ 1288*e23347b1SEric Taylor if (best_drive == 0 && best_part == 0 && find_best_root) { 1289e37b211cStaylor grub_memset(¤t_uberblock, 0, sizeof (uberblock_t)); 1290*e23347b1SEric Taylor pool_guid = 0; 1291*e23347b1SEric Taylor } 1292b1b8ab34Slling 1293b1b8ab34Slling stackbase = ZFS_SCRATCH; 1294b1b8ab34Slling stack = stackbase; 1295b1b8ab34Slling ub_array = (uberblock_phys_t *)stack; 1296b1b8ab34Slling stack += VDEV_UBERBLOCK_RING; 1297b1b8ab34Slling 1298b1b8ab34Slling osp = (objset_phys_t *)stack; 1299b1b8ab34Slling stack += sizeof (objset_phys_t); 1300*e23347b1SEric Taylor adjpl = P2ALIGN(adjpl, (uint64_t)sizeof (vdev_label_t)); 1301b1b8ab34Slling 1302*e23347b1SEric Taylor for (label = 0; label < VDEV_LABELS; label++) { 1303*e23347b1SEric Taylor uint64_t sector = vdev_label_start(adjpl, 1304*e23347b1SEric Taylor label) >> SPA_MINBLOCKSHIFT; 1305b1b8ab34Slling 1306b1b8ab34Slling /* Read in the uberblock ring (128K). */ 1307*e23347b1SEric Taylor if (devread(sector + 1308*e23347b1SEric Taylor ((VDEV_SKIP_SIZE + VDEV_PHYS_SIZE) >> 1309*e23347b1SEric Taylor SPA_MINBLOCKSHIFT), 0, VDEV_UBERBLOCK_RING, 1310b1b8ab34Slling (char *)ub_array) == 0) 1311b1b8ab34Slling continue; 1312b1b8ab34Slling 1313*e23347b1SEric Taylor if ((ubbest = find_bestub(ub_array, sector)) != NULL && 1314b1b8ab34Slling zio_read(&ubbest->ubp_uberblock.ub_rootbp, osp, stack) 1315b1b8ab34Slling == 0) { 1316b1b8ab34Slling 1317b1b8ab34Slling VERIFY_OS_TYPE(osp, DMU_OST_META); 1318b1b8ab34Slling 1319*e23347b1SEric Taylor if (check_pool_label(sector, stack, tmp_devid, 1320*e23347b1SEric Taylor tmp_bootpath, &tmp_guid)) 1321*e23347b1SEric Taylor continue; 1322*e23347b1SEric Taylor if (pool_guid == 0) 1323*e23347b1SEric Taylor pool_guid = tmp_guid; 1324b1b8ab34Slling 1325*e23347b1SEric Taylor if (find_best_root && ((pool_guid != tmp_guid) || 1326051aabe6Staylor vdev_uberblock_compare(&ubbest->ubp_uberblock, 1327*e23347b1SEric Taylor &(current_uberblock)) <= 0)) 1328051aabe6Staylor continue; 1329ffb5616eSLin Ling 1330051aabe6Staylor /* Got the MOS. Save it at the memory addr MOS. */ 1331051aabe6Staylor grub_memmove(MOS, &osp->os_meta_dnode, DNODE_SIZE); 1332051aabe6Staylor grub_memmove(¤t_uberblock, 1333ffb5616eSLin Ling &ubbest->ubp_uberblock, sizeof (uberblock_t)); 1334051aabe6Staylor grub_memmove(current_bootpath, tmp_bootpath, 1335051aabe6Staylor MAXNAMELEN); 1336051aabe6Staylor grub_memmove(current_devid, tmp_devid, 1337051aabe6Staylor grub_strlen(tmp_devid)); 1338b1b8ab34Slling is_zfs_mount = 1; 1339b1b8ab34Slling return (1); 1340b1b8ab34Slling } 1341b1b8ab34Slling } 1342b1b8ab34Slling 1343b1b8ab34Slling return (0); 1344b1b8ab34Slling } 1345b1b8ab34Slling 1346b1b8ab34Slling /* 1347b1b8ab34Slling * zfs_open() locates a file in the rootpool by following the 1348b1b8ab34Slling * MOS and places the dnode of the file in the memory address DNODE. 1349b1b8ab34Slling * 1350b1b8ab34Slling * Return: 1351b1b8ab34Slling * 1 - success 1352b1b8ab34Slling * 0 - failure 1353b1b8ab34Slling */ 1354b1b8ab34Slling int 1355b1b8ab34Slling zfs_open(char *filename) 1356b1b8ab34Slling { 1357b1b8ab34Slling char *stack; 1358b1b8ab34Slling dnode_phys_t *mdn; 1359b1b8ab34Slling 1360b1b8ab34Slling file_buf = NULL; 1361b1b8ab34Slling stackbase = ZFS_SCRATCH; 1362b1b8ab34Slling stack = stackbase; 1363b1b8ab34Slling 1364b1b8ab34Slling mdn = (dnode_phys_t *)stack; 1365b1b8ab34Slling stack += sizeof (dnode_phys_t); 1366b1b8ab34Slling 1367b1b8ab34Slling dnode_mdn = NULL; 1368b1b8ab34Slling dnode_buf = (dnode_phys_t *)stack; 1369b1b8ab34Slling stack += 1<<DNODE_BLOCK_SHIFT; 1370b1b8ab34Slling 1371b1b8ab34Slling /* 1372b1b8ab34Slling * menu.lst is placed at the root pool filesystem level, 1373b1b8ab34Slling * do not goto 'current_bootfs'. 1374b1b8ab34Slling */ 1375eb2bd662Svikram if (is_top_dataset_file(filename)) { 1376b1b8ab34Slling if (errnum = get_objset_mdn(MOS, NULL, NULL, mdn, stack)) 1377b1b8ab34Slling return (0); 1378b1b8ab34Slling 1379b1b8ab34Slling current_bootfs_obj = 0; 1380b1b8ab34Slling } else { 1381b1b8ab34Slling if (current_bootfs[0] == '\0') { 1382b1b8ab34Slling /* Get the default root filesystem object number */ 1383ae8180dbSlling if (errnum = get_default_bootfsobj(MOS, 1384ae8180dbSlling ¤t_bootfs_obj, stack)) 1385b1b8ab34Slling return (0); 1386b1b8ab34Slling 1387b1b8ab34Slling if (errnum = get_objset_mdn(MOS, NULL, 1388b1b8ab34Slling ¤t_bootfs_obj, mdn, stack)) 1389b1b8ab34Slling return (0); 1390b1b8ab34Slling } else { 1391b35c6776Staylor if (errnum = get_objset_mdn(MOS, current_bootfs, 1392b35c6776Staylor ¤t_bootfs_obj, mdn, stack)) { 1393051aabe6Staylor grub_memset(current_bootfs, 0, MAXNAMELEN); 1394b1b8ab34Slling return (0); 1395b35c6776Staylor } 1396b1b8ab34Slling } 1397b1b8ab34Slling } 1398b1b8ab34Slling 1399b1b8ab34Slling if (dnode_get_path(mdn, filename, DNODE, stack)) { 1400b1b8ab34Slling errnum = ERR_FILE_NOT_FOUND; 1401b1b8ab34Slling return (0); 1402b1b8ab34Slling } 1403b1b8ab34Slling 1404b1b8ab34Slling /* get the file size and set the file position to 0 */ 1405b1b8ab34Slling filemax = ((znode_phys_t *)DN_BONUS(DNODE))->zp_size; 1406b1b8ab34Slling filepos = 0; 1407b1b8ab34Slling 1408b1b8ab34Slling dnode_buf = NULL; 1409b1b8ab34Slling return (1); 1410b1b8ab34Slling } 1411b1b8ab34Slling 1412b1b8ab34Slling /* 1413b1b8ab34Slling * zfs_read reads in the data blocks pointed by the DNODE. 1414b1b8ab34Slling * 1415b1b8ab34Slling * Return: 1416b1b8ab34Slling * len - the length successfully read in to the buffer 1417b1b8ab34Slling * 0 - failure 1418b1b8ab34Slling */ 1419b1b8ab34Slling int 1420b1b8ab34Slling zfs_read(char *buf, int len) 1421b1b8ab34Slling { 1422b1b8ab34Slling char *stack; 1423b1b8ab34Slling char *tmpbuf; 1424b1b8ab34Slling int blksz, length, movesize; 1425b1b8ab34Slling 1426b1b8ab34Slling if (file_buf == NULL) { 1427b1b8ab34Slling file_buf = stackbase; 1428b1b8ab34Slling stackbase += SPA_MAXBLOCKSIZE; 1429b1b8ab34Slling file_start = file_end = 0; 1430b1b8ab34Slling } 1431b1b8ab34Slling stack = stackbase; 1432b1b8ab34Slling 1433b1b8ab34Slling /* 1434b1b8ab34Slling * If offset is in memory, move it into the buffer provided and return. 1435b1b8ab34Slling */ 1436b1b8ab34Slling if (filepos >= file_start && filepos+len <= file_end) { 1437b1b8ab34Slling grub_memmove(buf, file_buf + filepos - file_start, len); 1438b1b8ab34Slling filepos += len; 1439b1b8ab34Slling return (len); 1440b1b8ab34Slling } 1441b1b8ab34Slling 1442b1b8ab34Slling blksz = DNODE->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1443b1b8ab34Slling 1444b1b8ab34Slling /* 1445b1b8ab34Slling * Entire Dnode is too big to fit into the space available. We 1446b1b8ab34Slling * will need to read it in chunks. This could be optimized to 1447b1b8ab34Slling * read in as large a chunk as there is space available, but for 1448b1b8ab34Slling * now, this only reads in one data block at a time. 1449b1b8ab34Slling */ 1450b1b8ab34Slling length = len; 1451b1b8ab34Slling while (length) { 1452b1b8ab34Slling /* 1453b1b8ab34Slling * Find requested blkid and the offset within that block. 1454b1b8ab34Slling */ 1455b1b8ab34Slling uint64_t blkid = filepos / blksz; 1456b1b8ab34Slling 1457b1b8ab34Slling if (errnum = dmu_read(DNODE, blkid, file_buf, stack)) 1458b1b8ab34Slling return (0); 1459b1b8ab34Slling 1460b1b8ab34Slling file_start = blkid * blksz; 1461b1b8ab34Slling file_end = file_start + blksz; 1462b1b8ab34Slling 1463b1b8ab34Slling movesize = MIN(length, file_end - filepos); 1464b1b8ab34Slling 1465b1b8ab34Slling grub_memmove(buf, file_buf + filepos - file_start, 1466b1b8ab34Slling movesize); 1467b1b8ab34Slling buf += movesize; 1468b1b8ab34Slling length -= movesize; 1469b1b8ab34Slling filepos += movesize; 1470b1b8ab34Slling } 1471b1b8ab34Slling 1472b1b8ab34Slling return (len); 1473b1b8ab34Slling } 1474b1b8ab34Slling 1475b1b8ab34Slling /* 1476b1b8ab34Slling * No-Op 1477b1b8ab34Slling */ 1478b1b8ab34Slling int 1479b1b8ab34Slling zfs_embed(int *start_sector, int needed_sectors) 1480b1b8ab34Slling { 1481b1b8ab34Slling return (1); 1482b1b8ab34Slling } 1483b1b8ab34Slling 1484b1b8ab34Slling #endif /* FSYS_ZFS */ 1485