1185029Spjd/*- 2185029Spjd * Copyright (c) 2007 Doug Rabson 3185029Spjd * All rights reserved. 4185029Spjd * 5185029Spjd * Redistribution and use in source and binary forms, with or without 6185029Spjd * modification, are permitted provided that the following conditions 7185029Spjd * are met: 8185029Spjd * 1. Redistributions of source code must retain the above copyright 9185029Spjd * notice, this list of conditions and the following disclaimer. 10185029Spjd * 2. Redistributions in binary form must reproduce the above copyright 11185029Spjd * notice, this list of conditions and the following disclaimer in the 12185029Spjd * documentation and/or other materials provided with the distribution. 13185029Spjd * 14185029Spjd * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15185029Spjd * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16185029Spjd * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17185029Spjd * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18185029Spjd * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19185029Spjd * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20185029Spjd * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21185029Spjd * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22185029Spjd * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23185029Spjd * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24185029Spjd * SUCH DAMAGE. 25185029Spjd */ 26185029Spjd 27185029Spjd#include <sys/cdefs.h> 28185029Spjd__FBSDID("$FreeBSD: stable/10/sys/boot/zfs/zfsimpl.c 316323 2017-03-31 04:51:08Z ngie $"); 29185029Spjd 30185029Spjd/* 31185029Spjd * Stand-alone ZFS file reader. 32185029Spjd */ 33185029Spjd 34219089Spjd#include <sys/stat.h> 35235361Savg#include <sys/stdint.h> 36219089Spjd 37185029Spjd#include "zfsimpl.h" 38185029Spjd#include "zfssubr.c" 39185029Spjd 40235329Savg 41235329Savgstruct zfsmount { 42235390Savg const spa_t *spa; 43235329Savg objset_phys_t objset; 44235329Savg uint64_t rootobj; 45235329Savg}; 46235329Savg 47185029Spjd/* 48185029Spjd * List of all vdevs, chained through v_alllink. 49185029Spjd */ 50185029Spjdstatic vdev_list_t zfs_vdevs; 51185029Spjd 52236884Smm /* 53236884Smm * List of ZFS features supported for read 54236884Smm */ 55236884Smmstatic const char *features_for_read[] = { 56263397Sdelphij "org.illumos:lz4_compress", 57263397Sdelphij "com.delphix:hole_birth", 58263397Sdelphij "com.delphix:extensible_dataset", 59268649Sdelphij "com.delphix:embedded_data", 60276081Sdelphij "org.open-zfs:large_blocks", 61236884Smm NULL 62236884Smm}; 63236884Smm 64185029Spjd/* 65185029Spjd * List of all pools, chained through spa_link. 66185029Spjd */ 67185029Spjdstatic spa_list_t zfs_pools; 68185029Spjd 69185029Spjdstatic const dnode_phys_t *dnode_cache_obj = 0; 70185029Spjdstatic uint64_t dnode_cache_bn; 71185029Spjdstatic char *dnode_cache_buf; 72185029Spjdstatic char *zap_scratch; 73192194Sdfrstatic char *zfs_temp_buf, *zfs_temp_end, *zfs_temp_ptr; 74185029Spjd 75200309Sjhb#define TEMP_SIZE (1024 * 1024) 76185029Spjd 77235390Savgstatic int zio_read(const spa_t *spa, const blkptr_t *bp, void *buf); 78241291Savgstatic int zfs_get_root(const spa_t *spa, uint64_t *objid); 79241291Savgstatic int zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result); 80198420Srnoland 81185029Spjdstatic void 82185029Spjdzfs_init(void) 83185029Spjd{ 84185029Spjd STAILQ_INIT(&zfs_vdevs); 85185029Spjd STAILQ_INIT(&zfs_pools); 86185029Spjd 87192194Sdfr zfs_temp_buf = malloc(TEMP_SIZE); 88192194Sdfr zfs_temp_end = zfs_temp_buf + TEMP_SIZE; 89192194Sdfr zfs_temp_ptr = zfs_temp_buf; 90192194Sdfr dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE); 91192194Sdfr zap_scratch = malloc(SPA_MAXBLOCKSIZE); 92185029Spjd 93185029Spjd zfs_init_crc(); 94185029Spjd} 95185029Spjd 96219089Spjdstatic void * 97219089Spjdzfs_alloc(size_t size) 98192194Sdfr{ 99219089Spjd char *ptr; 100192194Sdfr 101219089Spjd if (zfs_temp_ptr + size > zfs_temp_end) { 102192194Sdfr printf("ZFS: out of temporary buffer space\n"); 103192194Sdfr for (;;) ; 104192194Sdfr } 105219089Spjd ptr = zfs_temp_ptr; 106219089Spjd zfs_temp_ptr += size; 107192194Sdfr 108219089Spjd return (ptr); 109192194Sdfr} 110192194Sdfr 111192194Sdfrstatic void 112219089Spjdzfs_free(void *ptr, size_t size) 113192194Sdfr{ 114192194Sdfr 115219089Spjd zfs_temp_ptr -= size; 116219089Spjd if (zfs_temp_ptr != ptr) { 117219089Spjd printf("ZFS: zfs_alloc()/zfs_free() mismatch\n"); 118219089Spjd for (;;) ; 119219089Spjd } 120192194Sdfr} 121192194Sdfr 122185029Spjdstatic int 123185029Spjdxdr_int(const unsigned char **xdr, int *ip) 124185029Spjd{ 125185029Spjd *ip = ((*xdr)[0] << 24) 126185029Spjd | ((*xdr)[1] << 16) 127185029Spjd | ((*xdr)[2] << 8) 128185029Spjd | ((*xdr)[3] << 0); 129185029Spjd (*xdr) += 4; 130185029Spjd return (0); 131185029Spjd} 132185029Spjd 133185029Spjdstatic int 134185029Spjdxdr_u_int(const unsigned char **xdr, u_int *ip) 135185029Spjd{ 136185029Spjd *ip = ((*xdr)[0] << 24) 137185029Spjd | ((*xdr)[1] << 16) 138185029Spjd | ((*xdr)[2] << 8) 139185029Spjd | ((*xdr)[3] << 0); 140185029Spjd (*xdr) += 4; 141185029Spjd return (0); 142185029Spjd} 143185029Spjd 144185029Spjdstatic int 145185029Spjdxdr_uint64_t(const unsigned char **xdr, uint64_t *lp) 146185029Spjd{ 147185029Spjd u_int hi, lo; 148185029Spjd 149185029Spjd xdr_u_int(xdr, &hi); 150185029Spjd xdr_u_int(xdr, &lo); 151185029Spjd *lp = (((uint64_t) hi) << 32) | lo; 152185029Spjd return (0); 153185029Spjd} 154185029Spjd 155185029Spjdstatic int 156185029Spjdnvlist_find(const unsigned char *nvlist, const char *name, int type, 157185029Spjd int* elementsp, void *valuep) 158185029Spjd{ 159185029Spjd const unsigned char *p, *pair; 160185029Spjd int junk; 161185029Spjd int encoded_size, decoded_size; 162185029Spjd 163185029Spjd p = nvlist; 164185029Spjd xdr_int(&p, &junk); 165185029Spjd xdr_int(&p, &junk); 166185029Spjd 167185029Spjd pair = p; 168185029Spjd xdr_int(&p, &encoded_size); 169185029Spjd xdr_int(&p, &decoded_size); 170185029Spjd while (encoded_size && decoded_size) { 171185029Spjd int namelen, pairtype, elements; 172185029Spjd const char *pairname; 173185029Spjd 174185029Spjd xdr_int(&p, &namelen); 175185029Spjd pairname = (const char*) p; 176185029Spjd p += roundup(namelen, 4); 177185029Spjd xdr_int(&p, &pairtype); 178185029Spjd 179185029Spjd if (!memcmp(name, pairname, namelen) && type == pairtype) { 180185029Spjd xdr_int(&p, &elements); 181185029Spjd if (elementsp) 182185029Spjd *elementsp = elements; 183185029Spjd if (type == DATA_TYPE_UINT64) { 184185029Spjd xdr_uint64_t(&p, (uint64_t *) valuep); 185185029Spjd return (0); 186185029Spjd } else if (type == DATA_TYPE_STRING) { 187185029Spjd int len; 188185029Spjd xdr_int(&p, &len); 189185029Spjd (*(const char**) valuep) = (const char*) p; 190185029Spjd return (0); 191185029Spjd } else if (type == DATA_TYPE_NVLIST 192185029Spjd || type == DATA_TYPE_NVLIST_ARRAY) { 193185029Spjd (*(const unsigned char**) valuep) = 194185029Spjd (const unsigned char*) p; 195185029Spjd return (0); 196185029Spjd } else { 197185029Spjd return (EIO); 198185029Spjd } 199185029Spjd } else { 200185029Spjd /* 201185029Spjd * Not the pair we are looking for, skip to the next one. 202185029Spjd */ 203185029Spjd p = pair + encoded_size; 204185029Spjd } 205185029Spjd 206185029Spjd pair = p; 207185029Spjd xdr_int(&p, &encoded_size); 208185029Spjd xdr_int(&p, &decoded_size); 209185029Spjd } 210185029Spjd 211185029Spjd return (EIO); 212185029Spjd} 213185029Spjd 214236884Smmstatic int 215236884Smmnvlist_check_features_for_read(const unsigned char *nvlist) 216236884Smm{ 217236884Smm const unsigned char *p, *pair; 218236884Smm int junk; 219236884Smm int encoded_size, decoded_size; 220236884Smm int rc; 221236884Smm 222236884Smm rc = 0; 223236884Smm 224236884Smm p = nvlist; 225236884Smm xdr_int(&p, &junk); 226236884Smm xdr_int(&p, &junk); 227236884Smm 228236884Smm pair = p; 229236884Smm xdr_int(&p, &encoded_size); 230236884Smm xdr_int(&p, &decoded_size); 231236884Smm while (encoded_size && decoded_size) { 232236884Smm int namelen, pairtype; 233236884Smm const char *pairname; 234236884Smm int i, found; 235236884Smm 236236884Smm found = 0; 237236884Smm 238236884Smm xdr_int(&p, &namelen); 239236884Smm pairname = (const char*) p; 240236884Smm p += roundup(namelen, 4); 241236884Smm xdr_int(&p, &pairtype); 242236884Smm 243236884Smm for (i = 0; features_for_read[i] != NULL; i++) { 244236884Smm if (!memcmp(pairname, features_for_read[i], namelen)) { 245236884Smm found = 1; 246236884Smm break; 247236884Smm } 248236884Smm } 249236884Smm 250236884Smm if (!found) { 251236884Smm printf("ZFS: unsupported feature: %s\n", pairname); 252236884Smm rc = EIO; 253236884Smm } 254236884Smm 255236884Smm p = pair + encoded_size; 256236884Smm 257236884Smm pair = p; 258236884Smm xdr_int(&p, &encoded_size); 259236884Smm xdr_int(&p, &decoded_size); 260236884Smm } 261236884Smm 262236884Smm return (rc); 263236884Smm} 264236884Smm 265185029Spjd/* 266185029Spjd * Return the next nvlist in an nvlist array. 267185029Spjd */ 268185029Spjdstatic const unsigned char * 269185029Spjdnvlist_next(const unsigned char *nvlist) 270185029Spjd{ 271185029Spjd const unsigned char *p, *pair; 272185029Spjd int junk; 273185029Spjd int encoded_size, decoded_size; 274185029Spjd 275185029Spjd p = nvlist; 276185029Spjd xdr_int(&p, &junk); 277185029Spjd xdr_int(&p, &junk); 278185029Spjd 279185029Spjd pair = p; 280185029Spjd xdr_int(&p, &encoded_size); 281185029Spjd xdr_int(&p, &decoded_size); 282185029Spjd while (encoded_size && decoded_size) { 283185029Spjd p = pair + encoded_size; 284185029Spjd 285185029Spjd pair = p; 286185029Spjd xdr_int(&p, &encoded_size); 287185029Spjd xdr_int(&p, &decoded_size); 288185029Spjd } 289185029Spjd 290185029Spjd return p; 291185029Spjd} 292185029Spjd 293185029Spjd#ifdef TEST 294185029Spjd 295185029Spjdstatic const unsigned char * 296185029Spjdnvlist_print(const unsigned char *nvlist, unsigned int indent) 297185029Spjd{ 298185029Spjd static const char* typenames[] = { 299185029Spjd "DATA_TYPE_UNKNOWN", 300185029Spjd "DATA_TYPE_BOOLEAN", 301185029Spjd "DATA_TYPE_BYTE", 302185029Spjd "DATA_TYPE_INT16", 303185029Spjd "DATA_TYPE_UINT16", 304185029Spjd "DATA_TYPE_INT32", 305185029Spjd "DATA_TYPE_UINT32", 306185029Spjd "DATA_TYPE_INT64", 307185029Spjd "DATA_TYPE_UINT64", 308185029Spjd "DATA_TYPE_STRING", 309185029Spjd "DATA_TYPE_BYTE_ARRAY", 310185029Spjd "DATA_TYPE_INT16_ARRAY", 311185029Spjd "DATA_TYPE_UINT16_ARRAY", 312185029Spjd "DATA_TYPE_INT32_ARRAY", 313185029Spjd "DATA_TYPE_UINT32_ARRAY", 314185029Spjd "DATA_TYPE_INT64_ARRAY", 315185029Spjd "DATA_TYPE_UINT64_ARRAY", 316185029Spjd "DATA_TYPE_STRING_ARRAY", 317185029Spjd "DATA_TYPE_HRTIME", 318185029Spjd "DATA_TYPE_NVLIST", 319185029Spjd "DATA_TYPE_NVLIST_ARRAY", 320185029Spjd "DATA_TYPE_BOOLEAN_VALUE", 321185029Spjd "DATA_TYPE_INT8", 322185029Spjd "DATA_TYPE_UINT8", 323185029Spjd "DATA_TYPE_BOOLEAN_ARRAY", 324185029Spjd "DATA_TYPE_INT8_ARRAY", 325185029Spjd "DATA_TYPE_UINT8_ARRAY" 326185029Spjd }; 327185029Spjd 328185029Spjd unsigned int i, j; 329185029Spjd const unsigned char *p, *pair; 330185029Spjd int junk; 331185029Spjd int encoded_size, decoded_size; 332185029Spjd 333185029Spjd p = nvlist; 334185029Spjd xdr_int(&p, &junk); 335185029Spjd xdr_int(&p, &junk); 336185029Spjd 337185029Spjd pair = p; 338185029Spjd xdr_int(&p, &encoded_size); 339185029Spjd xdr_int(&p, &decoded_size); 340185029Spjd while (encoded_size && decoded_size) { 341185029Spjd int namelen, pairtype, elements; 342185029Spjd const char *pairname; 343185029Spjd 344185029Spjd xdr_int(&p, &namelen); 345185029Spjd pairname = (const char*) p; 346185029Spjd p += roundup(namelen, 4); 347185029Spjd xdr_int(&p, &pairtype); 348185029Spjd 349185029Spjd for (i = 0; i < indent; i++) 350185029Spjd printf(" "); 351185029Spjd printf("%s %s", typenames[pairtype], pairname); 352185029Spjd 353185029Spjd xdr_int(&p, &elements); 354185029Spjd switch (pairtype) { 355185029Spjd case DATA_TYPE_UINT64: { 356185029Spjd uint64_t val; 357185029Spjd xdr_uint64_t(&p, &val); 358235361Savg printf(" = 0x%jx\n", (uintmax_t)val); 359185029Spjd break; 360185029Spjd } 361185029Spjd 362185029Spjd case DATA_TYPE_STRING: { 363185029Spjd int len; 364185029Spjd xdr_int(&p, &len); 365185029Spjd printf(" = \"%s\"\n", p); 366185029Spjd break; 367185029Spjd } 368185029Spjd 369185029Spjd case DATA_TYPE_NVLIST: 370185029Spjd printf("\n"); 371185029Spjd nvlist_print(p, indent + 1); 372185029Spjd break; 373185029Spjd 374185029Spjd case DATA_TYPE_NVLIST_ARRAY: 375185029Spjd for (j = 0; j < elements; j++) { 376185029Spjd printf("[%d]\n", j); 377185029Spjd p = nvlist_print(p, indent + 1); 378185029Spjd if (j != elements - 1) { 379185029Spjd for (i = 0; i < indent; i++) 380185029Spjd printf(" "); 381185029Spjd printf("%s %s", typenames[pairtype], pairname); 382185029Spjd } 383185029Spjd } 384185029Spjd break; 385185029Spjd 386185029Spjd default: 387185029Spjd printf("\n"); 388185029Spjd } 389185029Spjd 390185029Spjd p = pair + encoded_size; 391185029Spjd 392185029Spjd pair = p; 393185029Spjd xdr_int(&p, &encoded_size); 394185029Spjd xdr_int(&p, &decoded_size); 395185029Spjd } 396185029Spjd 397185029Spjd return p; 398185029Spjd} 399185029Spjd 400185029Spjd#endif 401185029Spjd 402185029Spjdstatic int 403192194Sdfrvdev_read_phys(vdev_t *vdev, const blkptr_t *bp, void *buf, 404192194Sdfr off_t offset, size_t size) 405185029Spjd{ 406192194Sdfr size_t psize; 407192194Sdfr int rc; 408192194Sdfr 409211091Smm if (!vdev->v_phys_read) 410211091Smm return (EIO); 411211091Smm 412192194Sdfr if (bp) { 413192194Sdfr psize = BP_GET_PSIZE(bp); 414192194Sdfr } else { 415192194Sdfr psize = size; 416192194Sdfr } 417192194Sdfr 418235361Savg /*printf("ZFS: reading %d bytes at 0x%jx to %p\n", psize, (uintmax_t)offset, buf);*/ 419192194Sdfr rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize); 420192194Sdfr if (rc) 421192194Sdfr return (rc); 422226568Spjd if (bp && zio_checksum_verify(bp, buf)) 423192194Sdfr return (EIO); 424192194Sdfr 425192194Sdfr return (0); 426192194Sdfr} 427192194Sdfr 428192194Sdfrstatic int 429192194Sdfrvdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf, 430192194Sdfr off_t offset, size_t bytes) 431192194Sdfr{ 432192194Sdfr 433192194Sdfr return (vdev_read_phys(vdev, bp, buf, 434192194Sdfr offset + VDEV_LABEL_START_SIZE, bytes)); 435192194Sdfr} 436192194Sdfr 437192194Sdfr 438192194Sdfrstatic int 439192194Sdfrvdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf, 440192194Sdfr off_t offset, size_t bytes) 441192194Sdfr{ 442185029Spjd vdev_t *kid; 443185029Spjd int rc; 444185029Spjd 445185029Spjd rc = EIO; 446185029Spjd STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { 447185029Spjd if (kid->v_state != VDEV_STATE_HEALTHY) 448185029Spjd continue; 449192194Sdfr rc = kid->v_read(kid, bp, buf, offset, bytes); 450185029Spjd if (!rc) 451185029Spjd return (0); 452185029Spjd } 453185029Spjd 454185029Spjd return (rc); 455185029Spjd} 456185029Spjd 457212383Spjdstatic int 458212383Spjdvdev_replacing_read(vdev_t *vdev, const blkptr_t *bp, void *buf, 459212383Spjd off_t offset, size_t bytes) 460212383Spjd{ 461212383Spjd vdev_t *kid; 462212383Spjd 463212383Spjd /* 464212383Spjd * Here we should have two kids: 465212383Spjd * First one which is the one we are replacing and we can trust 466212383Spjd * only this one to have valid data, but it might not be present. 467212383Spjd * Second one is that one we are replacing with. It is most likely 468212383Spjd * healthy, but we can't trust it has needed data, so we won't use it. 469212383Spjd */ 470212383Spjd kid = STAILQ_FIRST(&vdev->v_children); 471212383Spjd if (kid == NULL) 472212383Spjd return (EIO); 473212383Spjd if (kid->v_state != VDEV_STATE_HEALTHY) 474212383Spjd return (EIO); 475212383Spjd return (kid->v_read(kid, bp, buf, offset, bytes)); 476212383Spjd} 477212383Spjd 478185029Spjdstatic vdev_t * 479185029Spjdvdev_find(uint64_t guid) 480185029Spjd{ 481185029Spjd vdev_t *vdev; 482185029Spjd 483185029Spjd STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink) 484185029Spjd if (vdev->v_guid == guid) 485185029Spjd return (vdev); 486185029Spjd 487185029Spjd return (0); 488185029Spjd} 489185029Spjd 490185029Spjdstatic vdev_t * 491316323Sngievdev_create(uint64_t guid, vdev_read_t *_read) 492185029Spjd{ 493185029Spjd vdev_t *vdev; 494185029Spjd 495185029Spjd vdev = malloc(sizeof(vdev_t)); 496185029Spjd memset(vdev, 0, sizeof(vdev_t)); 497185029Spjd STAILQ_INIT(&vdev->v_children); 498185029Spjd vdev->v_guid = guid; 499185029Spjd vdev->v_state = VDEV_STATE_OFFLINE; 500316323Sngie vdev->v_read = _read; 501192194Sdfr vdev->v_phys_read = 0; 502192194Sdfr vdev->v_read_priv = 0; 503185029Spjd STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink); 504185029Spjd 505185029Spjd return (vdev); 506185029Spjd} 507185029Spjd 508185029Spjdstatic int 509219089Spjdvdev_init_from_nvlist(const unsigned char *nvlist, vdev_t *pvdev, 510219089Spjd vdev_t **vdevp, int is_newer) 511185029Spjd{ 512185029Spjd int rc; 513192194Sdfr uint64_t guid, id, ashift, nparity; 514185029Spjd const char *type; 515185029Spjd const char *path; 516185029Spjd vdev_t *vdev, *kid; 517185029Spjd const unsigned char *kids; 518201689Sdelphij int nkids, i, is_new; 519212383Spjd uint64_t is_offline, is_faulted, is_degraded, is_removed, isnt_present; 520185029Spjd 521185029Spjd if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, 522185029Spjd DATA_TYPE_UINT64, 0, &guid) 523185029Spjd || nvlist_find(nvlist, ZPOOL_CONFIG_ID, 524185029Spjd DATA_TYPE_UINT64, 0, &id) 525185029Spjd || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE, 526185029Spjd DATA_TYPE_STRING, 0, &type)) { 527185029Spjd printf("ZFS: can't find vdev details\n"); 528185029Spjd return (ENOENT); 529185029Spjd } 530185029Spjd 531185029Spjd if (strcmp(type, VDEV_TYPE_MIRROR) 532192194Sdfr && strcmp(type, VDEV_TYPE_DISK) 533228266Savg#ifdef ZFS_TEST 534228266Savg && strcmp(type, VDEV_TYPE_FILE) 535228266Savg#endif 536212383Spjd && strcmp(type, VDEV_TYPE_RAIDZ) 537212383Spjd && strcmp(type, VDEV_TYPE_REPLACING)) { 538219089Spjd printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n"); 539185029Spjd return (EIO); 540185029Spjd } 541185029Spjd 542212383Spjd is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0; 543201689Sdelphij 544201689Sdelphij nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, 0, 545201689Sdelphij &is_offline); 546201689Sdelphij nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, 0, 547201689Sdelphij &is_removed); 548201689Sdelphij nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, 0, 549201689Sdelphij &is_faulted); 550201689Sdelphij nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, 0, 551201689Sdelphij &is_degraded); 552212383Spjd nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, 0, 553212383Spjd &isnt_present); 554201689Sdelphij 555201689Sdelphij vdev = vdev_find(guid); 556201689Sdelphij if (!vdev) { 557201689Sdelphij is_new = 1; 558201689Sdelphij 559201690Sdelphij if (!strcmp(type, VDEV_TYPE_MIRROR)) 560201690Sdelphij vdev = vdev_create(guid, vdev_mirror_read); 561201690Sdelphij else if (!strcmp(type, VDEV_TYPE_RAIDZ)) 562201690Sdelphij vdev = vdev_create(guid, vdev_raidz_read); 563212383Spjd else if (!strcmp(type, VDEV_TYPE_REPLACING)) 564212383Spjd vdev = vdev_create(guid, vdev_replacing_read); 565201690Sdelphij else 566201690Sdelphij vdev = vdev_create(guid, vdev_disk_read); 567185029Spjd 568201690Sdelphij vdev->v_id = id; 569219089Spjd vdev->v_top = pvdev != NULL ? pvdev : vdev; 570201690Sdelphij if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT, 571201690Sdelphij DATA_TYPE_UINT64, 0, &ashift) == 0) 572201690Sdelphij vdev->v_ashift = ashift; 573201690Sdelphij else 574201690Sdelphij vdev->v_ashift = 0; 575201690Sdelphij if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY, 576201690Sdelphij DATA_TYPE_UINT64, 0, &nparity) == 0) 577201690Sdelphij vdev->v_nparity = nparity; 578201690Sdelphij else 579201690Sdelphij vdev->v_nparity = 0; 580201690Sdelphij if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH, 581201690Sdelphij DATA_TYPE_STRING, 0, &path) == 0) { 582213136Spjd if (strncmp(path, "/dev/", 5) == 0) 583201690Sdelphij path += 5; 584201690Sdelphij vdev->v_name = strdup(path); 585192194Sdfr } else { 586201690Sdelphij if (!strcmp(type, "raidz")) { 587201690Sdelphij if (vdev->v_nparity == 1) 588201690Sdelphij vdev->v_name = "raidz1"; 589219089Spjd else if (vdev->v_nparity == 2) 590201690Sdelphij vdev->v_name = "raidz2"; 591219089Spjd else if (vdev->v_nparity == 3) 592219089Spjd vdev->v_name = "raidz3"; 593219089Spjd else { 594219089Spjd printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n"); 595219089Spjd return (EIO); 596219089Spjd } 597201690Sdelphij } else { 598201690Sdelphij vdev->v_name = strdup(type); 599201690Sdelphij } 600192194Sdfr } 601212382Spjd } else { 602212382Spjd is_new = 0; 603212382Spjd } 604201689Sdelphij 605212382Spjd if (is_new || is_newer) { 606212382Spjd /* 607212382Spjd * This is either new vdev or we've already seen this vdev, 608212382Spjd * but from an older vdev label, so let's refresh its state 609212382Spjd * from the newer label. 610212382Spjd */ 611201689Sdelphij if (is_offline) 612201689Sdelphij vdev->v_state = VDEV_STATE_OFFLINE; 613201689Sdelphij else if (is_removed) 614201689Sdelphij vdev->v_state = VDEV_STATE_REMOVED; 615201689Sdelphij else if (is_faulted) 616201689Sdelphij vdev->v_state = VDEV_STATE_FAULTED; 617201689Sdelphij else if (is_degraded) 618201689Sdelphij vdev->v_state = VDEV_STATE_DEGRADED; 619212383Spjd else if (isnt_present) 620212383Spjd vdev->v_state = VDEV_STATE_CANT_OPEN; 621201689Sdelphij } 622201689Sdelphij 623185029Spjd rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, 624185029Spjd DATA_TYPE_NVLIST_ARRAY, &nkids, &kids); 625185029Spjd /* 626185029Spjd * Its ok if we don't have any kids. 627185029Spjd */ 628185029Spjd if (rc == 0) { 629192194Sdfr vdev->v_nchildren = nkids; 630185029Spjd for (i = 0; i < nkids; i++) { 631219089Spjd rc = vdev_init_from_nvlist(kids, vdev, &kid, is_newer); 632185029Spjd if (rc) 633185029Spjd return (rc); 634201689Sdelphij if (is_new) 635201689Sdelphij STAILQ_INSERT_TAIL(&vdev->v_children, kid, 636201689Sdelphij v_childlink); 637185029Spjd kids = nvlist_next(kids); 638185029Spjd } 639192194Sdfr } else { 640192194Sdfr vdev->v_nchildren = 0; 641185029Spjd } 642185029Spjd 643185029Spjd if (vdevp) 644185029Spjd *vdevp = vdev; 645185029Spjd return (0); 646185029Spjd} 647185029Spjd 648185029Spjdstatic void 649185029Spjdvdev_set_state(vdev_t *vdev) 650185029Spjd{ 651185029Spjd vdev_t *kid; 652185029Spjd int good_kids; 653185029Spjd int bad_kids; 654185029Spjd 655185029Spjd /* 656192194Sdfr * A mirror or raidz is healthy if all its kids are healthy. A 657192194Sdfr * mirror is degraded if any of its kids is healthy; a raidz 658192194Sdfr * is degraded if at most nparity kids are offline. 659185029Spjd */ 660185029Spjd if (STAILQ_FIRST(&vdev->v_children)) { 661185029Spjd good_kids = 0; 662185029Spjd bad_kids = 0; 663185029Spjd STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { 664185029Spjd if (kid->v_state == VDEV_STATE_HEALTHY) 665185029Spjd good_kids++; 666185029Spjd else 667185029Spjd bad_kids++; 668185029Spjd } 669192194Sdfr if (bad_kids == 0) { 670192194Sdfr vdev->v_state = VDEV_STATE_HEALTHY; 671185029Spjd } else { 672192194Sdfr if (vdev->v_read == vdev_mirror_read) { 673192194Sdfr if (good_kids) { 674192194Sdfr vdev->v_state = VDEV_STATE_DEGRADED; 675192194Sdfr } else { 676192194Sdfr vdev->v_state = VDEV_STATE_OFFLINE; 677192194Sdfr } 678192194Sdfr } else if (vdev->v_read == vdev_raidz_read) { 679192194Sdfr if (bad_kids > vdev->v_nparity) { 680192194Sdfr vdev->v_state = VDEV_STATE_OFFLINE; 681192194Sdfr } else { 682192194Sdfr vdev->v_state = VDEV_STATE_DEGRADED; 683192194Sdfr } 684192194Sdfr } 685185029Spjd } 686185029Spjd } 687185029Spjd} 688185029Spjd 689185029Spjdstatic spa_t * 690185029Spjdspa_find_by_guid(uint64_t guid) 691185029Spjd{ 692185029Spjd spa_t *spa; 693185029Spjd 694185029Spjd STAILQ_FOREACH(spa, &zfs_pools, spa_link) 695185029Spjd if (spa->spa_guid == guid) 696185029Spjd return (spa); 697185029Spjd 698185029Spjd return (0); 699185029Spjd} 700185029Spjd 701185029Spjdstatic spa_t * 702185029Spjdspa_find_by_name(const char *name) 703185029Spjd{ 704185029Spjd spa_t *spa; 705185029Spjd 706185029Spjd STAILQ_FOREACH(spa, &zfs_pools, spa_link) 707185029Spjd if (!strcmp(spa->spa_name, name)) 708185029Spjd return (spa); 709185029Spjd 710185029Spjd return (0); 711185029Spjd} 712185029Spjd 713241293Savg#ifdef BOOT2 714235329Savgstatic spa_t * 715241293Savgspa_get_primary(void) 716241293Savg{ 717241293Savg 718241293Savg return (STAILQ_FIRST(&zfs_pools)); 719241293Savg} 720241293Savg 721241293Savgstatic vdev_t * 722241293Savgspa_get_primary_vdev(const spa_t *spa) 723241293Savg{ 724241293Savg vdev_t *vdev; 725241293Savg vdev_t *kid; 726241293Savg 727241293Savg if (spa == NULL) 728241293Savg spa = spa_get_primary(); 729241293Savg if (spa == NULL) 730241293Savg return (NULL); 731241293Savg vdev = STAILQ_FIRST(&spa->spa_vdevs); 732241293Savg if (vdev == NULL) 733241293Savg return (NULL); 734241293Savg for (kid = STAILQ_FIRST(&vdev->v_children); kid != NULL; 735241293Savg kid = STAILQ_FIRST(&vdev->v_children)) 736241293Savg vdev = kid; 737241293Savg return (vdev); 738241293Savg} 739241293Savg#endif 740241293Savg 741241293Savgstatic spa_t * 742185029Spjdspa_create(uint64_t guid) 743185029Spjd{ 744185029Spjd spa_t *spa; 745185029Spjd 746185029Spjd spa = malloc(sizeof(spa_t)); 747185029Spjd memset(spa, 0, sizeof(spa_t)); 748185029Spjd STAILQ_INIT(&spa->spa_vdevs); 749185029Spjd spa->spa_guid = guid; 750185029Spjd STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link); 751185029Spjd 752185029Spjd return (spa); 753185029Spjd} 754185029Spjd 755185029Spjdstatic const char * 756185029Spjdstate_name(vdev_state_t state) 757185029Spjd{ 758185029Spjd static const char* names[] = { 759185029Spjd "UNKNOWN", 760185029Spjd "CLOSED", 761185029Spjd "OFFLINE", 762201689Sdelphij "REMOVED", 763185029Spjd "CANT_OPEN", 764201689Sdelphij "FAULTED", 765185029Spjd "DEGRADED", 766185029Spjd "ONLINE" 767185029Spjd }; 768185029Spjd return names[state]; 769185029Spjd} 770185029Spjd 771185029Spjd#ifdef BOOT2 772185029Spjd 773185029Spjd#define pager_printf printf 774185029Spjd 775185029Spjd#else 776185029Spjd 777185029Spjdstatic void 778185029Spjdpager_printf(const char *fmt, ...) 779185029Spjd{ 780185029Spjd char line[80]; 781185029Spjd va_list args; 782185029Spjd 783185029Spjd va_start(args, fmt); 784185029Spjd vsprintf(line, fmt, args); 785185029Spjd va_end(args); 786185029Spjd pager_output(line); 787185029Spjd} 788185029Spjd 789185029Spjd#endif 790185029Spjd 791213136Spjd#define STATUS_FORMAT " %s %s\n" 792185029Spjd 793185029Spjdstatic void 794185029Spjdprint_state(int indent, const char *name, vdev_state_t state) 795185029Spjd{ 796185029Spjd int i; 797185029Spjd char buf[512]; 798185029Spjd 799185029Spjd buf[0] = 0; 800185029Spjd for (i = 0; i < indent; i++) 801185029Spjd strcat(buf, " "); 802185029Spjd strcat(buf, name); 803185029Spjd pager_printf(STATUS_FORMAT, buf, state_name(state)); 804185029Spjd 805185029Spjd} 806185029Spjd 807185029Spjdstatic void 808185029Spjdvdev_status(vdev_t *vdev, int indent) 809185029Spjd{ 810185029Spjd vdev_t *kid; 811185029Spjd print_state(indent, vdev->v_name, vdev->v_state); 812185029Spjd 813185029Spjd STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) { 814185029Spjd vdev_status(kid, indent + 1); 815185029Spjd } 816185029Spjd} 817185029Spjd 818185029Spjdstatic void 819185029Spjdspa_status(spa_t *spa) 820185029Spjd{ 821241291Savg static char bootfs[ZFS_MAXNAMELEN]; 822241291Savg uint64_t rootid; 823185029Spjd vdev_t *vdev; 824185029Spjd int good_kids, bad_kids, degraded_kids; 825185029Spjd vdev_state_t state; 826185029Spjd 827185029Spjd pager_printf(" pool: %s\n", spa->spa_name); 828241291Savg if (zfs_get_root(spa, &rootid) == 0 && 829241291Savg zfs_rlookup(spa, rootid, bootfs) == 0) { 830241291Savg if (bootfs[0] == '\0') 831241291Savg pager_printf("bootfs: %s\n", spa->spa_name); 832241291Savg else 833241291Savg pager_printf("bootfs: %s/%s\n", spa->spa_name, bootfs); 834241291Savg } 835185029Spjd pager_printf("config:\n\n"); 836185029Spjd pager_printf(STATUS_FORMAT, "NAME", "STATE"); 837185029Spjd 838185029Spjd good_kids = 0; 839185029Spjd degraded_kids = 0; 840185029Spjd bad_kids = 0; 841185029Spjd STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) { 842185029Spjd if (vdev->v_state == VDEV_STATE_HEALTHY) 843185029Spjd good_kids++; 844185029Spjd else if (vdev->v_state == VDEV_STATE_DEGRADED) 845185029Spjd degraded_kids++; 846185029Spjd else 847185029Spjd bad_kids++; 848185029Spjd } 849185029Spjd 850185029Spjd state = VDEV_STATE_CLOSED; 851185029Spjd if (good_kids > 0 && (degraded_kids + bad_kids) == 0) 852185029Spjd state = VDEV_STATE_HEALTHY; 853185029Spjd else if ((good_kids + degraded_kids) > 0) 854185029Spjd state = VDEV_STATE_DEGRADED; 855185029Spjd 856185029Spjd print_state(0, spa->spa_name, state); 857185029Spjd STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) { 858185029Spjd vdev_status(vdev, 1); 859185029Spjd } 860185029Spjd} 861185029Spjd 862185029Spjdstatic void 863185029Spjdspa_all_status(void) 864185029Spjd{ 865185029Spjd spa_t *spa; 866185029Spjd int first = 1; 867185029Spjd 868185029Spjd STAILQ_FOREACH(spa, &zfs_pools, spa_link) { 869185029Spjd if (!first) 870185029Spjd pager_printf("\n"); 871185029Spjd first = 0; 872185029Spjd spa_status(spa); 873185029Spjd } 874185029Spjd} 875185029Spjd 876185029Spjdstatic int 877316323Sngievdev_probe(vdev_phys_read_t *_read, void *read_priv, spa_t **spap) 878185029Spjd{ 879185029Spjd vdev_t vtmp; 880185029Spjd vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch; 881185029Spjd spa_t *spa; 882185029Spjd vdev_t *vdev, *top_vdev, *pool_vdev; 883185029Spjd off_t off; 884185029Spjd blkptr_t bp; 885185029Spjd const unsigned char *nvlist; 886185029Spjd uint64_t val; 887185029Spjd uint64_t guid; 888185029Spjd uint64_t pool_txg, pool_guid; 889212384Spjd uint64_t is_log; 890185029Spjd const char *pool_name; 891185029Spjd const unsigned char *vdevs; 892236884Smm const unsigned char *features; 893201689Sdelphij int i, rc, is_newer; 894219089Spjd char *upbuf; 895185029Spjd const struct uberblock *up; 896185029Spjd 897185029Spjd /* 898185029Spjd * Load the vdev label and figure out which 899185029Spjd * uberblock is most current. 900185029Spjd */ 901185029Spjd memset(&vtmp, 0, sizeof(vtmp)); 902316323Sngie vtmp.v_phys_read = _read; 903185029Spjd vtmp.v_read_priv = read_priv; 904185029Spjd off = offsetof(vdev_label_t, vl_vdev_phys); 905185029Spjd BP_ZERO(&bp); 906185029Spjd BP_SET_LSIZE(&bp, sizeof(vdev_phys_t)); 907185029Spjd BP_SET_PSIZE(&bp, sizeof(vdev_phys_t)); 908185029Spjd BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); 909185029Spjd BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); 910226568Spjd DVA_SET_OFFSET(BP_IDENTITY(&bp), off); 911185029Spjd ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0); 912192194Sdfr if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0)) 913185029Spjd return (EIO); 914185029Spjd 915185029Spjd if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) { 916185029Spjd return (EIO); 917185029Spjd } 918185029Spjd 919185029Spjd nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4; 920185029Spjd 921185029Spjd if (nvlist_find(nvlist, 922185029Spjd ZPOOL_CONFIG_VERSION, 923185029Spjd DATA_TYPE_UINT64, 0, &val)) { 924185029Spjd return (EIO); 925185029Spjd } 926185029Spjd 927236884Smm if (!SPA_VERSION_IS_SUPPORTED(val)) { 928185097Sdfr printf("ZFS: unsupported ZFS version %u (should be %u)\n", 929185097Sdfr (unsigned) val, (unsigned) SPA_VERSION); 930185029Spjd return (EIO); 931185029Spjd } 932185029Spjd 933236884Smm /* Check ZFS features for read */ 934237001Smm if (nvlist_find(nvlist, 935237001Smm ZPOOL_CONFIG_FEATURES_FOR_READ, 936237001Smm DATA_TYPE_NVLIST, 0, &features) == 0 937237001Smm && nvlist_check_features_for_read(features) != 0) 938236884Smm return (EIO); 939236884Smm 940185029Spjd if (nvlist_find(nvlist, 941185029Spjd ZPOOL_CONFIG_POOL_STATE, 942185029Spjd DATA_TYPE_UINT64, 0, &val)) { 943185029Spjd return (EIO); 944185029Spjd } 945185029Spjd 946219089Spjd if (val == POOL_STATE_DESTROYED) { 947219089Spjd /* We don't boot only from destroyed pools. */ 948185029Spjd return (EIO); 949185029Spjd } 950185029Spjd 951185029Spjd if (nvlist_find(nvlist, 952185029Spjd ZPOOL_CONFIG_POOL_TXG, 953185029Spjd DATA_TYPE_UINT64, 0, &pool_txg) 954185029Spjd || nvlist_find(nvlist, 955185029Spjd ZPOOL_CONFIG_POOL_GUID, 956185029Spjd DATA_TYPE_UINT64, 0, &pool_guid) 957185029Spjd || nvlist_find(nvlist, 958185029Spjd ZPOOL_CONFIG_POOL_NAME, 959185029Spjd DATA_TYPE_STRING, 0, &pool_name)) { 960192194Sdfr /* 961192194Sdfr * Cache and spare devices end up here - just ignore 962192194Sdfr * them. 963192194Sdfr */ 964192194Sdfr /*printf("ZFS: can't find pool details\n");*/ 965185029Spjd return (EIO); 966185029Spjd } 967185029Spjd 968212384Spjd is_log = 0; 969212384Spjd (void) nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, 0, 970212384Spjd &is_log); 971212384Spjd if (is_log) 972212384Spjd return (EIO); 973212384Spjd 974185029Spjd /* 975185029Spjd * Create the pool if this is the first time we've seen it. 976185029Spjd */ 977185029Spjd spa = spa_find_by_guid(pool_guid); 978185029Spjd if (!spa) { 979185029Spjd spa = spa_create(pool_guid); 980185029Spjd spa->spa_name = strdup(pool_name); 981185029Spjd } 982201689Sdelphij if (pool_txg > spa->spa_txg) { 983185029Spjd spa->spa_txg = pool_txg; 984201689Sdelphij is_newer = 1; 985201689Sdelphij } else 986201689Sdelphij is_newer = 0; 987185029Spjd 988185029Spjd /* 989185029Spjd * Get the vdev tree and create our in-core copy of it. 990201689Sdelphij * If we already have a vdev with this guid, this must 991185029Spjd * be some kind of alias (overlapping slices, dangerously dedicated 992185029Spjd * disks etc). 993185029Spjd */ 994185029Spjd if (nvlist_find(nvlist, 995185029Spjd ZPOOL_CONFIG_GUID, 996185029Spjd DATA_TYPE_UINT64, 0, &guid)) { 997185029Spjd return (EIO); 998185029Spjd } 999185029Spjd vdev = vdev_find(guid); 1000201689Sdelphij if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */ 1001185029Spjd return (EIO); 1002185029Spjd 1003185029Spjd if (nvlist_find(nvlist, 1004185029Spjd ZPOOL_CONFIG_VDEV_TREE, 1005185029Spjd DATA_TYPE_NVLIST, 0, &vdevs)) { 1006185029Spjd return (EIO); 1007185029Spjd } 1008201689Sdelphij 1009219089Spjd rc = vdev_init_from_nvlist(vdevs, NULL, &top_vdev, is_newer); 1010185852Sdfr if (rc) 1011185852Sdfr return (rc); 1012185029Spjd 1013185029Spjd /* 1014185029Spjd * Add the toplevel vdev to the pool if its not already there. 1015185029Spjd */ 1016185029Spjd STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink) 1017185029Spjd if (top_vdev == pool_vdev) 1018185029Spjd break; 1019185029Spjd if (!pool_vdev && top_vdev) 1020185029Spjd STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink); 1021185029Spjd 1022185029Spjd /* 1023185029Spjd * We should already have created an incomplete vdev for this 1024185029Spjd * vdev. Find it and initialise it with our read proc. 1025185029Spjd */ 1026185029Spjd vdev = vdev_find(guid); 1027185029Spjd if (vdev) { 1028316323Sngie vdev->v_phys_read = _read; 1029185029Spjd vdev->v_read_priv = read_priv; 1030226551Spjd vdev->v_state = VDEV_STATE_HEALTHY; 1031185029Spjd } else { 1032185029Spjd printf("ZFS: inconsistent nvlist contents\n"); 1033185029Spjd return (EIO); 1034185029Spjd } 1035185029Spjd 1036185029Spjd /* 1037185029Spjd * Re-evaluate top-level vdev state. 1038185029Spjd */ 1039185029Spjd vdev_set_state(top_vdev); 1040185029Spjd 1041185029Spjd /* 1042185029Spjd * Ok, we are happy with the pool so far. Lets find 1043185029Spjd * the best uberblock and then we can actually access 1044185029Spjd * the contents of the pool. 1045185029Spjd */ 1046219089Spjd upbuf = zfs_alloc(VDEV_UBERBLOCK_SIZE(vdev)); 1047219089Spjd up = (const struct uberblock *)upbuf; 1048185029Spjd for (i = 0; 1049219089Spjd i < VDEV_UBERBLOCK_COUNT(vdev); 1050185029Spjd i++) { 1051219089Spjd off = VDEV_UBERBLOCK_OFFSET(vdev, i); 1052185029Spjd BP_ZERO(&bp); 1053185029Spjd DVA_SET_OFFSET(&bp.blk_dva[0], off); 1054219089Spjd BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); 1055219089Spjd BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev)); 1056185029Spjd BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL); 1057185029Spjd BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF); 1058185029Spjd ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0); 1059219089Spjd 1060226568Spjd if (vdev_read_phys(vdev, &bp, upbuf, off, 0)) 1061185029Spjd continue; 1062185029Spjd 1063185029Spjd if (up->ub_magic != UBERBLOCK_MAGIC) 1064185029Spjd continue; 1065185029Spjd if (up->ub_txg < spa->spa_txg) 1066185029Spjd continue; 1067185029Spjd if (up->ub_txg > spa->spa_uberblock.ub_txg) { 1068185029Spjd spa->spa_uberblock = *up; 1069185029Spjd } else if (up->ub_txg == spa->spa_uberblock.ub_txg) { 1070185029Spjd if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp) 1071185029Spjd spa->spa_uberblock = *up; 1072185029Spjd } 1073185029Spjd } 1074219089Spjd zfs_free(upbuf, VDEV_UBERBLOCK_SIZE(vdev)); 1075185029Spjd 1076185029Spjd if (spap) 1077185029Spjd *spap = spa; 1078185029Spjd return (0); 1079185029Spjd} 1080185029Spjd 1081185029Spjdstatic int 1082185029Spjdilog2(int n) 1083185029Spjd{ 1084185029Spjd int v; 1085185029Spjd 1086185029Spjd for (v = 0; v < 32; v++) 1087185029Spjd if (n == (1 << v)) 1088185029Spjd return v; 1089185029Spjd return -1; 1090185029Spjd} 1091185029Spjd 1092185029Spjdstatic int 1093235390Savgzio_read_gang(const spa_t *spa, const blkptr_t *bp, void *buf) 1094198420Srnoland{ 1095226568Spjd blkptr_t gbh_bp; 1096198420Srnoland zio_gbh_phys_t zio_gb; 1097226568Spjd char *pbuf; 1098198420Srnoland int i; 1099198420Srnoland 1100226568Spjd /* Artificial BP for gang block header. */ 1101226568Spjd gbh_bp = *bp; 1102226568Spjd BP_SET_PSIZE(&gbh_bp, SPA_GANGBLOCKSIZE); 1103226568Spjd BP_SET_LSIZE(&gbh_bp, SPA_GANGBLOCKSIZE); 1104226568Spjd BP_SET_CHECKSUM(&gbh_bp, ZIO_CHECKSUM_GANG_HEADER); 1105226568Spjd BP_SET_COMPRESS(&gbh_bp, ZIO_COMPRESS_OFF); 1106226568Spjd for (i = 0; i < SPA_DVAS_PER_BP; i++) 1107226568Spjd DVA_SET_GANG(&gbh_bp.blk_dva[i], 0); 1108226552Spjd 1109226568Spjd /* Read gang header block using the artificial BP. */ 1110226568Spjd if (zio_read(spa, &gbh_bp, &zio_gb)) 1111198420Srnoland return (EIO); 1112198420Srnoland 1113226568Spjd pbuf = buf; 1114198420Srnoland for (i = 0; i < SPA_GBH_NBLKPTRS; i++) { 1115208610Savg blkptr_t *gbp = &zio_gb.zg_blkptr[i]; 1116208610Savg 1117208610Savg if (BP_IS_HOLE(gbp)) 1118208610Savg continue; 1119226568Spjd if (zio_read(spa, gbp, pbuf)) 1120198420Srnoland return (EIO); 1121226568Spjd pbuf += BP_GET_PSIZE(gbp); 1122198420Srnoland } 1123226568Spjd 1124226568Spjd if (zio_checksum_verify(bp, buf)) 1125226568Spjd return (EIO); 1126198420Srnoland return (0); 1127198420Srnoland} 1128198420Srnoland 1129198420Srnolandstatic int 1130235390Savgzio_read(const spa_t *spa, const blkptr_t *bp, void *buf) 1131185029Spjd{ 1132185029Spjd int cpfunc = BP_GET_COMPRESS(bp); 1133219089Spjd uint64_t align, size; 1134192194Sdfr void *pbuf; 1135219089Spjd int i, error; 1136185029Spjd 1137268649Sdelphij /* 1138268649Sdelphij * Process data embedded in block pointer 1139268649Sdelphij */ 1140268649Sdelphij if (BP_IS_EMBEDDED(bp)) { 1141268649Sdelphij ASSERT(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA); 1142268649Sdelphij 1143268649Sdelphij size = BPE_GET_PSIZE(bp); 1144268649Sdelphij ASSERT(size <= BPE_PAYLOAD_SIZE); 1145268649Sdelphij 1146268649Sdelphij if (cpfunc != ZIO_COMPRESS_OFF) 1147268649Sdelphij pbuf = zfs_alloc(size); 1148268649Sdelphij else 1149268649Sdelphij pbuf = buf; 1150268649Sdelphij 1151268649Sdelphij decode_embedded_bp_compressed(bp, pbuf); 1152268649Sdelphij error = 0; 1153268649Sdelphij 1154268649Sdelphij if (cpfunc != ZIO_COMPRESS_OFF) { 1155268649Sdelphij error = zio_decompress_data(cpfunc, pbuf, 1156268649Sdelphij size, buf, BP_GET_LSIZE(bp)); 1157268649Sdelphij zfs_free(pbuf, size); 1158268649Sdelphij } 1159268649Sdelphij if (error != 0) 1160268649Sdelphij printf("ZFS: i/o error - unable to decompress block pointer data, error %d\n", 1161268649Sdelphij error); 1162268649Sdelphij return (error); 1163268649Sdelphij } 1164268649Sdelphij 1165219089Spjd error = EIO; 1166185029Spjd 1167185029Spjd for (i = 0; i < SPA_DVAS_PER_BP; i++) { 1168185029Spjd const dva_t *dva = &bp->blk_dva[i]; 1169185029Spjd vdev_t *vdev; 1170185029Spjd int vdevid; 1171185029Spjd off_t offset; 1172185029Spjd 1173185029Spjd if (!dva->dva_word[0] && !dva->dva_word[1]) 1174185029Spjd continue; 1175185029Spjd 1176226568Spjd vdevid = DVA_GET_VDEV(dva); 1177226568Spjd offset = DVA_GET_OFFSET(dva); 1178226568Spjd STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) { 1179226568Spjd if (vdev->v_id == vdevid) 1180226568Spjd break; 1181226568Spjd } 1182226568Spjd if (!vdev || !vdev->v_read) 1183226568Spjd continue; 1184185029Spjd 1185226568Spjd size = BP_GET_PSIZE(bp); 1186226568Spjd if (vdev->v_read == vdev_raidz_read) { 1187219089Spjd align = 1ULL << vdev->v_top->v_ashift; 1188219089Spjd if (P2PHASE(size, align) != 0) 1189219089Spjd size = P2ROUNDUP(size, align); 1190226568Spjd } 1191226568Spjd if (size != BP_GET_PSIZE(bp) || cpfunc != ZIO_COMPRESS_OFF) 1192226568Spjd pbuf = zfs_alloc(size); 1193226568Spjd else 1194226568Spjd pbuf = buf; 1195219089Spjd 1196226568Spjd if (DVA_GET_GANG(dva)) 1197226568Spjd error = zio_read_gang(spa, bp, pbuf); 1198226568Spjd else 1199219089Spjd error = vdev->v_read(vdev, bp, pbuf, offset, size); 1200226568Spjd if (error == 0) { 1201226568Spjd if (cpfunc != ZIO_COMPRESS_OFF) 1202226568Spjd error = zio_decompress_data(cpfunc, pbuf, 1203226568Spjd BP_GET_PSIZE(bp), buf, BP_GET_LSIZE(bp)); 1204226568Spjd else if (size != BP_GET_PSIZE(bp)) 1205226568Spjd bcopy(pbuf, buf, BP_GET_PSIZE(bp)); 1206192194Sdfr } 1207226568Spjd if (buf != pbuf) 1208226568Spjd zfs_free(pbuf, size); 1209226568Spjd if (error == 0) 1210226568Spjd break; 1211185029Spjd } 1212219089Spjd if (error != 0) 1213219089Spjd printf("ZFS: i/o error - all block copies unavailable\n"); 1214219089Spjd return (error); 1215185029Spjd} 1216185029Spjd 1217185029Spjdstatic int 1218235390Savgdnode_read(const spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen) 1219185029Spjd{ 1220185029Spjd int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT; 1221185029Spjd int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1222185029Spjd int nlevels = dnode->dn_nlevels; 1223185029Spjd int i, rc; 1224185029Spjd 1225276081Sdelphij if (bsize > SPA_MAXBLOCKSIZE) { 1226276081Sdelphij printf("ZFS: I/O error - blocks larger than 128K are not supported\n"); 1227276081Sdelphij return (EIO); 1228276081Sdelphij } 1229276081Sdelphij 1230185029Spjd /* 1231186243Sdfr * Note: bsize may not be a power of two here so we need to do an 1232186243Sdfr * actual divide rather than a bitshift. 1233185029Spjd */ 1234185029Spjd while (buflen > 0) { 1235186243Sdfr uint64_t bn = offset / bsize; 1236186243Sdfr int boff = offset % bsize; 1237185029Spjd int ibn; 1238185029Spjd const blkptr_t *indbp; 1239185029Spjd blkptr_t bp; 1240185029Spjd 1241185029Spjd if (bn > dnode->dn_maxblkid) 1242185029Spjd return (EIO); 1243185029Spjd 1244185029Spjd if (dnode == dnode_cache_obj && bn == dnode_cache_bn) 1245185029Spjd goto cached; 1246185029Spjd 1247185029Spjd indbp = dnode->dn_blkptr; 1248185029Spjd for (i = 0; i < nlevels; i++) { 1249185029Spjd /* 1250185029Spjd * Copy the bp from the indirect array so that 1251185029Spjd * we can re-use the scratch buffer for multi-level 1252185029Spjd * objects. 1253185029Spjd */ 1254185029Spjd ibn = bn >> ((nlevels - i - 1) * ibshift); 1255185029Spjd ibn &= ((1 << ibshift) - 1); 1256185029Spjd bp = indbp[ibn]; 1257284509Savg if (BP_IS_HOLE(&bp)) { 1258284509Savg memset(dnode_cache_buf, 0, bsize); 1259284509Savg break; 1260284509Savg } 1261185029Spjd rc = zio_read(spa, &bp, dnode_cache_buf); 1262185029Spjd if (rc) 1263185029Spjd return (rc); 1264185029Spjd indbp = (const blkptr_t *) dnode_cache_buf; 1265185029Spjd } 1266185029Spjd dnode_cache_obj = dnode; 1267185029Spjd dnode_cache_bn = bn; 1268185029Spjd cached: 1269185029Spjd 1270185029Spjd /* 1271185029Spjd * The buffer contains our data block. Copy what we 1272185029Spjd * need from it and loop. 1273185029Spjd */ 1274185029Spjd i = bsize - boff; 1275185029Spjd if (i > buflen) i = buflen; 1276185029Spjd memcpy(buf, &dnode_cache_buf[boff], i); 1277185029Spjd buf = ((char*) buf) + i; 1278185029Spjd offset += i; 1279185029Spjd buflen -= i; 1280185029Spjd } 1281185029Spjd 1282185029Spjd return (0); 1283185029Spjd} 1284185029Spjd 1285185029Spjd/* 1286185029Spjd * Lookup a value in a microzap directory. Assumes that the zap 1287185029Spjd * scratch buffer contains the directory contents. 1288185029Spjd */ 1289185029Spjdstatic int 1290235390Savgmzap_lookup(const dnode_phys_t *dnode, const char *name, uint64_t *value) 1291185029Spjd{ 1292185029Spjd const mzap_phys_t *mz; 1293185029Spjd const mzap_ent_phys_t *mze; 1294185029Spjd size_t size; 1295185029Spjd int chunks, i; 1296185029Spjd 1297185029Spjd /* 1298185029Spjd * Microzap objects use exactly one block. Read the whole 1299185029Spjd * thing. 1300185029Spjd */ 1301185029Spjd size = dnode->dn_datablkszsec * 512; 1302185029Spjd 1303185029Spjd mz = (const mzap_phys_t *) zap_scratch; 1304185029Spjd chunks = size / MZAP_ENT_LEN - 1; 1305185029Spjd 1306185029Spjd for (i = 0; i < chunks; i++) { 1307185029Spjd mze = &mz->mz_chunk[i]; 1308185029Spjd if (!strcmp(mze->mze_name, name)) { 1309185029Spjd *value = mze->mze_value; 1310185029Spjd return (0); 1311185029Spjd } 1312185029Spjd } 1313185029Spjd 1314185029Spjd return (ENOENT); 1315185029Spjd} 1316185029Spjd 1317185029Spjd/* 1318185029Spjd * Compare a name with a zap leaf entry. Return non-zero if the name 1319185029Spjd * matches. 1320185029Spjd */ 1321185029Spjdstatic int 1322185029Spjdfzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name) 1323185029Spjd{ 1324185029Spjd size_t namelen; 1325185029Spjd const zap_leaf_chunk_t *nc; 1326185029Spjd const char *p; 1327185029Spjd 1328240346Savg namelen = zc->l_entry.le_name_numints; 1329185029Spjd 1330185029Spjd nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk); 1331185029Spjd p = name; 1332185029Spjd while (namelen > 0) { 1333185029Spjd size_t len; 1334185029Spjd len = namelen; 1335185029Spjd if (len > ZAP_LEAF_ARRAY_BYTES) 1336185029Spjd len = ZAP_LEAF_ARRAY_BYTES; 1337185029Spjd if (memcmp(p, nc->l_array.la_array, len)) 1338185029Spjd return (0); 1339185029Spjd p += len; 1340185029Spjd namelen -= len; 1341185029Spjd nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next); 1342185029Spjd } 1343185029Spjd 1344185029Spjd return 1; 1345185029Spjd} 1346185029Spjd 1347185029Spjd/* 1348185029Spjd * Extract a uint64_t value from a zap leaf entry. 1349185029Spjd */ 1350185029Spjdstatic uint64_t 1351185029Spjdfzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc) 1352185029Spjd{ 1353185029Spjd const zap_leaf_chunk_t *vc; 1354185029Spjd int i; 1355185029Spjd uint64_t value; 1356185029Spjd const uint8_t *p; 1357185029Spjd 1358185029Spjd vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk); 1359185029Spjd for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) { 1360185029Spjd value = (value << 8) | p[i]; 1361185029Spjd } 1362185029Spjd 1363185029Spjd return value; 1364185029Spjd} 1365185029Spjd 1366185029Spjd/* 1367185029Spjd * Lookup a value in a fatzap directory. Assumes that the zap scratch 1368185029Spjd * buffer contains the directory header. 1369185029Spjd */ 1370185029Spjdstatic int 1371235390Savgfzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value) 1372185029Spjd{ 1373185029Spjd int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1374185029Spjd zap_phys_t zh = *(zap_phys_t *) zap_scratch; 1375185029Spjd fat_zap_t z; 1376185029Spjd uint64_t *ptrtbl; 1377185029Spjd uint64_t hash; 1378185029Spjd int rc; 1379185029Spjd 1380185029Spjd if (zh.zap_magic != ZAP_MAGIC) 1381185029Spjd return (EIO); 1382185029Spjd 1383185029Spjd z.zap_block_shift = ilog2(bsize); 1384185029Spjd z.zap_phys = (zap_phys_t *) zap_scratch; 1385185029Spjd 1386185029Spjd /* 1387185029Spjd * Figure out where the pointer table is and read it in if necessary. 1388185029Spjd */ 1389185029Spjd if (zh.zap_ptrtbl.zt_blk) { 1390185029Spjd rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize, 1391185029Spjd zap_scratch, bsize); 1392185029Spjd if (rc) 1393185029Spjd return (rc); 1394185029Spjd ptrtbl = (uint64_t *) zap_scratch; 1395185029Spjd } else { 1396185029Spjd ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0); 1397185029Spjd } 1398185029Spjd 1399185029Spjd hash = zap_hash(zh.zap_salt, name); 1400185029Spjd 1401185029Spjd zap_leaf_t zl; 1402185029Spjd zl.l_bs = z.zap_block_shift; 1403185029Spjd 1404185029Spjd off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs; 1405185029Spjd zap_leaf_chunk_t *zc; 1406185029Spjd 1407185029Spjd rc = dnode_read(spa, dnode, off, zap_scratch, bsize); 1408185029Spjd if (rc) 1409185029Spjd return (rc); 1410185029Spjd 1411185029Spjd zl.l_phys = (zap_leaf_phys_t *) zap_scratch; 1412185029Spjd 1413185029Spjd /* 1414185029Spjd * Make sure this chunk matches our hash. 1415185029Spjd */ 1416185029Spjd if (zl.l_phys->l_hdr.lh_prefix_len > 0 1417185029Spjd && zl.l_phys->l_hdr.lh_prefix 1418185029Spjd != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len)) 1419185029Spjd return (ENOENT); 1420185029Spjd 1421185029Spjd /* 1422185029Spjd * Hash within the chunk to find our entry. 1423185029Spjd */ 1424185029Spjd int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len); 1425185029Spjd int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1); 1426185029Spjd h = zl.l_phys->l_hash[h]; 1427185029Spjd if (h == 0xffff) 1428185029Spjd return (ENOENT); 1429185029Spjd zc = &ZAP_LEAF_CHUNK(&zl, h); 1430185029Spjd while (zc->l_entry.le_hash != hash) { 1431185029Spjd if (zc->l_entry.le_next == 0xffff) { 1432185029Spjd zc = 0; 1433185029Spjd break; 1434185029Spjd } 1435185029Spjd zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next); 1436185029Spjd } 1437185029Spjd if (fzap_name_equal(&zl, zc, name)) { 1438240349Savg if (zc->l_entry.le_value_intlen * zc->l_entry.le_value_numints > 8) 1439240349Savg return (E2BIG); 1440185029Spjd *value = fzap_leaf_value(&zl, zc); 1441185029Spjd return (0); 1442185029Spjd } 1443185029Spjd 1444185029Spjd return (ENOENT); 1445185029Spjd} 1446185029Spjd 1447185029Spjd/* 1448185029Spjd * Lookup a name in a zap object and return its value as a uint64_t. 1449185029Spjd */ 1450185029Spjdstatic int 1451235390Savgzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value) 1452185029Spjd{ 1453185029Spjd int rc; 1454185029Spjd uint64_t zap_type; 1455219089Spjd size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1456185029Spjd 1457185029Spjd rc = dnode_read(spa, dnode, 0, zap_scratch, size); 1458185029Spjd if (rc) 1459185029Spjd return (rc); 1460185029Spjd 1461185029Spjd zap_type = *(uint64_t *) zap_scratch; 1462185029Spjd if (zap_type == ZBT_MICRO) 1463235390Savg return mzap_lookup(dnode, name, value); 1464219089Spjd else if (zap_type == ZBT_HEADER) 1465185029Spjd return fzap_lookup(spa, dnode, name, value); 1466219089Spjd printf("ZFS: invalid zap_type=%d\n", (int)zap_type); 1467219089Spjd return (EIO); 1468185029Spjd} 1469185029Spjd 1470185029Spjd/* 1471185029Spjd * List a microzap directory. Assumes that the zap scratch buffer contains 1472185029Spjd * the directory contents. 1473185029Spjd */ 1474185029Spjdstatic int 1475293802Sallanjudemzap_list(const dnode_phys_t *dnode, int (*callback)(const char *)) 1476185029Spjd{ 1477185029Spjd const mzap_phys_t *mz; 1478185029Spjd const mzap_ent_phys_t *mze; 1479185029Spjd size_t size; 1480185029Spjd int chunks, i; 1481185029Spjd 1482185029Spjd /* 1483185029Spjd * Microzap objects use exactly one block. Read the whole 1484185029Spjd * thing. 1485185029Spjd */ 1486185029Spjd size = dnode->dn_datablkszsec * 512; 1487185029Spjd mz = (const mzap_phys_t *) zap_scratch; 1488185029Spjd chunks = size / MZAP_ENT_LEN - 1; 1489185029Spjd 1490185029Spjd for (i = 0; i < chunks; i++) { 1491185029Spjd mze = &mz->mz_chunk[i]; 1492185029Spjd if (mze->mze_name[0]) 1493235361Savg //printf("%-32s 0x%jx\n", mze->mze_name, (uintmax_t)mze->mze_value); 1494293802Sallanjude callback(mze->mze_name); 1495185029Spjd } 1496185029Spjd 1497185029Spjd return (0); 1498185029Spjd} 1499185029Spjd 1500185029Spjd/* 1501185029Spjd * List a fatzap directory. Assumes that the zap scratch buffer contains 1502185029Spjd * the directory header. 1503185029Spjd */ 1504185029Spjdstatic int 1505293802Sallanjudefzap_list(const spa_t *spa, const dnode_phys_t *dnode, int (*callback)(const char *)) 1506185029Spjd{ 1507185029Spjd int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1508185029Spjd zap_phys_t zh = *(zap_phys_t *) zap_scratch; 1509185029Spjd fat_zap_t z; 1510185029Spjd int i, j; 1511185029Spjd 1512185029Spjd if (zh.zap_magic != ZAP_MAGIC) 1513185029Spjd return (EIO); 1514185029Spjd 1515185029Spjd z.zap_block_shift = ilog2(bsize); 1516185029Spjd z.zap_phys = (zap_phys_t *) zap_scratch; 1517185029Spjd 1518185029Spjd /* 1519185029Spjd * This assumes that the leaf blocks start at block 1. The 1520185029Spjd * documentation isn't exactly clear on this. 1521185029Spjd */ 1522185029Spjd zap_leaf_t zl; 1523185029Spjd zl.l_bs = z.zap_block_shift; 1524185029Spjd for (i = 0; i < zh.zap_num_leafs; i++) { 1525185029Spjd off_t off = (i + 1) << zl.l_bs; 1526185029Spjd char name[256], *p; 1527185029Spjd uint64_t value; 1528185029Spjd 1529185029Spjd if (dnode_read(spa, dnode, off, zap_scratch, bsize)) 1530185029Spjd return (EIO); 1531185029Spjd 1532185029Spjd zl.l_phys = (zap_leaf_phys_t *) zap_scratch; 1533185029Spjd 1534185029Spjd for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) { 1535185029Spjd zap_leaf_chunk_t *zc, *nc; 1536185029Spjd int namelen; 1537185029Spjd 1538185029Spjd zc = &ZAP_LEAF_CHUNK(&zl, j); 1539185029Spjd if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY) 1540185029Spjd continue; 1541240346Savg namelen = zc->l_entry.le_name_numints; 1542185029Spjd if (namelen > sizeof(name)) 1543185029Spjd namelen = sizeof(name); 1544235390Savg 1545185029Spjd /* 1546185029Spjd * Paste the name back together. 1547185029Spjd */ 1548185029Spjd nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk); 1549185029Spjd p = name; 1550185029Spjd while (namelen > 0) { 1551185029Spjd int len; 1552185029Spjd len = namelen; 1553185029Spjd if (len > ZAP_LEAF_ARRAY_BYTES) 1554185029Spjd len = ZAP_LEAF_ARRAY_BYTES; 1555185029Spjd memcpy(p, nc->l_array.la_array, len); 1556185029Spjd p += len; 1557185029Spjd namelen -= len; 1558185029Spjd nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next); 1559185029Spjd } 1560185029Spjd 1561185029Spjd /* 1562185029Spjd * Assume the first eight bytes of the value are 1563185029Spjd * a uint64_t. 1564185029Spjd */ 1565185029Spjd value = fzap_leaf_value(&zl, zc); 1566185029Spjd 1567240348Savg //printf("%s 0x%jx\n", name, (uintmax_t)value); 1568293802Sallanjude callback((const char *)name); 1569185029Spjd } 1570185029Spjd } 1571185029Spjd 1572185029Spjd return (0); 1573185029Spjd} 1574185029Spjd 1575293802Sallanjudestatic int zfs_printf(const char *name) 1576293802Sallanjude{ 1577293802Sallanjude 1578293802Sallanjude printf("%s\n", name); 1579293802Sallanjude 1580293802Sallanjude return (0); 1581293802Sallanjude} 1582293802Sallanjude 1583185029Spjd/* 1584185029Spjd * List a zap directory. 1585185029Spjd */ 1586185029Spjdstatic int 1587235390Savgzap_list(const spa_t *spa, const dnode_phys_t *dnode) 1588185029Spjd{ 1589185029Spjd uint64_t zap_type; 1590185029Spjd size_t size = dnode->dn_datablkszsec * 512; 1591185029Spjd 1592185029Spjd if (dnode_read(spa, dnode, 0, zap_scratch, size)) 1593185029Spjd return (EIO); 1594185029Spjd 1595185029Spjd zap_type = *(uint64_t *) zap_scratch; 1596185029Spjd if (zap_type == ZBT_MICRO) 1597293802Sallanjude return mzap_list(dnode, zfs_printf); 1598185029Spjd else 1599293802Sallanjude return fzap_list(spa, dnode, zfs_printf); 1600185029Spjd} 1601185029Spjd 1602185029Spjdstatic int 1603235390Savgobjset_get_dnode(const spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode) 1604185029Spjd{ 1605185029Spjd off_t offset; 1606185029Spjd 1607185029Spjd offset = objnum * sizeof(dnode_phys_t); 1608185029Spjd return dnode_read(spa, &os->os_meta_dnode, offset, 1609185029Spjd dnode, sizeof(dnode_phys_t)); 1610185029Spjd} 1611185029Spjd 1612235329Savgstatic int 1613235390Savgmzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value) 1614235329Savg{ 1615235329Savg const mzap_phys_t *mz; 1616235329Savg const mzap_ent_phys_t *mze; 1617235329Savg size_t size; 1618235329Savg int chunks, i; 1619235329Savg 1620235329Savg /* 1621235329Savg * Microzap objects use exactly one block. Read the whole 1622235329Savg * thing. 1623235329Savg */ 1624235329Savg size = dnode->dn_datablkszsec * 512; 1625235329Savg 1626235329Savg mz = (const mzap_phys_t *) zap_scratch; 1627235329Savg chunks = size / MZAP_ENT_LEN - 1; 1628235329Savg 1629235329Savg for (i = 0; i < chunks; i++) { 1630235329Savg mze = &mz->mz_chunk[i]; 1631235329Savg if (value == mze->mze_value) { 1632235329Savg strcpy(name, mze->mze_name); 1633235329Savg return (0); 1634235329Savg } 1635235329Savg } 1636235329Savg 1637235329Savg return (ENOENT); 1638235329Savg} 1639235329Savg 1640235329Savgstatic void 1641235329Savgfzap_name_copy(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, char *name) 1642235329Savg{ 1643235329Savg size_t namelen; 1644235329Savg const zap_leaf_chunk_t *nc; 1645235329Savg char *p; 1646235329Savg 1647240346Savg namelen = zc->l_entry.le_name_numints; 1648235329Savg 1649235329Savg nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk); 1650235329Savg p = name; 1651235329Savg while (namelen > 0) { 1652235329Savg size_t len; 1653235329Savg len = namelen; 1654235329Savg if (len > ZAP_LEAF_ARRAY_BYTES) 1655235329Savg len = ZAP_LEAF_ARRAY_BYTES; 1656235329Savg memcpy(p, nc->l_array.la_array, len); 1657235329Savg p += len; 1658235329Savg namelen -= len; 1659235329Savg nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next); 1660235329Savg } 1661235329Savg 1662235329Savg *p = '\0'; 1663235329Savg} 1664235329Savg 1665235329Savgstatic int 1666235390Savgfzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value) 1667235329Savg{ 1668235329Savg int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT; 1669235329Savg zap_phys_t zh = *(zap_phys_t *) zap_scratch; 1670235329Savg fat_zap_t z; 1671240347Savg int i, j; 1672235329Savg 1673235329Savg if (zh.zap_magic != ZAP_MAGIC) 1674235329Savg return (EIO); 1675235329Savg 1676235329Savg z.zap_block_shift = ilog2(bsize); 1677235329Savg z.zap_phys = (zap_phys_t *) zap_scratch; 1678235329Savg 1679235329Savg /* 1680240347Savg * This assumes that the leaf blocks start at block 1. The 1681240347Savg * documentation isn't exactly clear on this. 1682235329Savg */ 1683235329Savg zap_leaf_t zl; 1684235329Savg zl.l_bs = z.zap_block_shift; 1685240347Savg for (i = 0; i < zh.zap_num_leafs; i++) { 1686240347Savg off_t off = (i + 1) << zl.l_bs; 1687235329Savg 1688240347Savg if (dnode_read(spa, dnode, off, zap_scratch, bsize)) 1689240347Savg return (EIO); 1690235329Savg 1691240347Savg zl.l_phys = (zap_leaf_phys_t *) zap_scratch; 1692235329Savg 1693240347Savg for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) { 1694240347Savg zap_leaf_chunk_t *zc; 1695235329Savg 1696240347Savg zc = &ZAP_LEAF_CHUNK(&zl, j); 1697240347Savg if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY) 1698240347Savg continue; 1699240347Savg if (zc->l_entry.le_value_intlen != 8 || 1700240347Savg zc->l_entry.le_value_numints != 1) 1701240347Savg continue; 1702235329Savg 1703240347Savg if (fzap_leaf_value(&zl, zc) == value) { 1704240347Savg fzap_name_copy(&zl, zc, name); 1705240347Savg return (0); 1706240347Savg } 1707235329Savg } 1708235329Savg } 1709235329Savg 1710235329Savg return (ENOENT); 1711235329Savg} 1712235329Savg 1713235329Savgstatic int 1714235390Savgzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value) 1715235329Savg{ 1716235329Savg int rc; 1717235329Savg uint64_t zap_type; 1718235329Savg size_t size = dnode->dn_datablkszsec * 512; 1719235329Savg 1720235329Savg rc = dnode_read(spa, dnode, 0, zap_scratch, size); 1721235329Savg if (rc) 1722235329Savg return (rc); 1723235329Savg 1724235329Savg zap_type = *(uint64_t *) zap_scratch; 1725235329Savg if (zap_type == ZBT_MICRO) 1726235329Savg return mzap_rlookup(spa, dnode, name, value); 1727235329Savg else 1728235329Savg return fzap_rlookup(spa, dnode, name, value); 1729235329Savg} 1730235329Savg 1731235329Savgstatic int 1732235390Savgzfs_rlookup(const spa_t *spa, uint64_t objnum, char *result) 1733235329Savg{ 1734235329Savg char name[256]; 1735235329Savg char component[256]; 1736235329Savg uint64_t dir_obj, parent_obj, child_dir_zapobj; 1737235329Savg dnode_phys_t child_dir_zap, dataset, dir, parent; 1738235329Savg dsl_dir_phys_t *dd; 1739235329Savg dsl_dataset_phys_t *ds; 1740235329Savg char *p; 1741235329Savg int len; 1742235329Savg 1743235329Savg p = &name[sizeof(name) - 1]; 1744235329Savg *p = '\0'; 1745235329Savg 1746235329Savg if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { 1747235361Savg printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); 1748235329Savg return (EIO); 1749235329Savg } 1750235329Savg ds = (dsl_dataset_phys_t *)&dataset.dn_bonus; 1751235329Savg dir_obj = ds->ds_dir_obj; 1752235329Savg 1753235329Savg for (;;) { 1754235329Savg if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir) != 0) 1755235329Savg return (EIO); 1756235329Savg dd = (dsl_dir_phys_t *)&dir.dn_bonus; 1757235329Savg 1758235329Savg /* Actual loop condition. */ 1759235329Savg parent_obj = dd->dd_parent_obj; 1760235329Savg if (parent_obj == 0) 1761235329Savg break; 1762235329Savg 1763235329Savg if (objset_get_dnode(spa, &spa->spa_mos, parent_obj, &parent) != 0) 1764235329Savg return (EIO); 1765235329Savg dd = (dsl_dir_phys_t *)&parent.dn_bonus; 1766235329Savg child_dir_zapobj = dd->dd_child_dir_zapobj; 1767235329Savg if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) 1768235329Savg return (EIO); 1769235329Savg if (zap_rlookup(spa, &child_dir_zap, component, dir_obj) != 0) 1770235329Savg return (EIO); 1771235329Savg 1772235329Savg len = strlen(component); 1773235329Savg p -= len; 1774235329Savg memcpy(p, component, len); 1775235329Savg --p; 1776235329Savg *p = '/'; 1777235329Savg 1778235329Savg /* Actual loop iteration. */ 1779235329Savg dir_obj = parent_obj; 1780235329Savg } 1781235329Savg 1782235329Savg if (*p != '\0') 1783235329Savg ++p; 1784235329Savg strcpy(result, p); 1785235329Savg 1786235329Savg return (0); 1787235329Savg} 1788235329Savg 1789235329Savgstatic int 1790235390Savgzfs_lookup_dataset(const spa_t *spa, const char *name, uint64_t *objnum) 1791235329Savg{ 1792235329Savg char element[256]; 1793235329Savg uint64_t dir_obj, child_dir_zapobj; 1794235329Savg dnode_phys_t child_dir_zap, dir; 1795235329Savg dsl_dir_phys_t *dd; 1796235329Savg const char *p, *q; 1797235329Savg 1798235329Savg if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) 1799235329Savg return (EIO); 1800235329Savg if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &dir_obj)) 1801235329Savg return (EIO); 1802235329Savg 1803235329Savg p = name; 1804235329Savg for (;;) { 1805235329Savg if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) 1806235329Savg return (EIO); 1807235329Savg dd = (dsl_dir_phys_t *)&dir.dn_bonus; 1808235329Savg 1809235329Savg while (*p == '/') 1810235329Savg p++; 1811235329Savg /* Actual loop condition #1. */ 1812235329Savg if (*p == '\0') 1813235329Savg break; 1814235329Savg 1815235329Savg q = strchr(p, '/'); 1816235329Savg if (q) { 1817235329Savg memcpy(element, p, q - p); 1818235329Savg element[q - p] = '\0'; 1819235329Savg p = q + 1; 1820235329Savg } else { 1821235329Savg strcpy(element, p); 1822235329Savg p += strlen(p); 1823235329Savg } 1824235329Savg 1825235329Savg child_dir_zapobj = dd->dd_child_dir_zapobj; 1826235329Savg if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) 1827235329Savg return (EIO); 1828235329Savg 1829235329Savg /* Actual loop condition #2. */ 1830235329Savg if (zap_lookup(spa, &child_dir_zap, element, &dir_obj) != 0) 1831235329Savg return (ENOENT); 1832235329Savg } 1833235329Savg 1834235329Savg *objnum = dd->dd_head_dataset_obj; 1835235329Savg return (0); 1836235329Savg} 1837235329Savg 1838241283Savg#ifndef BOOT2 1839241283Savgstatic int 1840241283Savgzfs_list_dataset(const spa_t *spa, uint64_t objnum/*, int pos, char *entry*/) 1841241283Savg{ 1842241283Savg uint64_t dir_obj, child_dir_zapobj; 1843241283Savg dnode_phys_t child_dir_zap, dir, dataset; 1844241283Savg dsl_dataset_phys_t *ds; 1845241283Savg dsl_dir_phys_t *dd; 1846241283Savg 1847241283Savg if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { 1848241283Savg printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); 1849241283Savg return (EIO); 1850241283Savg } 1851241283Savg ds = (dsl_dataset_phys_t *) &dataset.dn_bonus; 1852241283Savg dir_obj = ds->ds_dir_obj; 1853241283Savg 1854241283Savg if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) { 1855241283Savg printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj); 1856241283Savg return (EIO); 1857241283Savg } 1858241283Savg dd = (dsl_dir_phys_t *)&dir.dn_bonus; 1859241283Savg 1860241283Savg child_dir_zapobj = dd->dd_child_dir_zapobj; 1861241283Savg if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) { 1862241283Savg printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj); 1863241283Savg return (EIO); 1864241283Savg } 1865241283Savg 1866241283Savg return (zap_list(spa, &child_dir_zap) != 0); 1867241283Savg} 1868293802Sallanjude 1869293802Sallanjudeint 1870293802Sallanjudezfs_callback_dataset(const spa_t *spa, uint64_t objnum, int (*callback)(const char *name)) 1871293802Sallanjude{ 1872293802Sallanjude uint64_t dir_obj, child_dir_zapobj, zap_type; 1873293802Sallanjude dnode_phys_t child_dir_zap, dir, dataset; 1874293802Sallanjude dsl_dataset_phys_t *ds; 1875293802Sallanjude dsl_dir_phys_t *dd; 1876293802Sallanjude int err; 1877293802Sallanjude 1878293802Sallanjude err = objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset); 1879293802Sallanjude if (err != 0) { 1880293802Sallanjude printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); 1881293802Sallanjude return (err); 1882293802Sallanjude } 1883293802Sallanjude ds = (dsl_dataset_phys_t *) &dataset.dn_bonus; 1884293802Sallanjude dir_obj = ds->ds_dir_obj; 1885293802Sallanjude 1886293802Sallanjude err = objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir); 1887293802Sallanjude if (err != 0) { 1888293802Sallanjude printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj); 1889293802Sallanjude return (err); 1890293802Sallanjude } 1891293802Sallanjude dd = (dsl_dir_phys_t *)&dir.dn_bonus; 1892293802Sallanjude 1893293802Sallanjude child_dir_zapobj = dd->dd_child_dir_zapobj; 1894293802Sallanjude err = objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap); 1895293802Sallanjude if (err != 0) { 1896293802Sallanjude printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj); 1897293802Sallanjude return (err); 1898293802Sallanjude } 1899293802Sallanjude 1900293802Sallanjude err = dnode_read(spa, &child_dir_zap, 0, zap_scratch, child_dir_zap.dn_datablkszsec * 512); 1901293802Sallanjude if (err != 0) 1902293802Sallanjude return (err); 1903293802Sallanjude 1904293802Sallanjude zap_type = *(uint64_t *) zap_scratch; 1905293802Sallanjude if (zap_type == ZBT_MICRO) 1906293802Sallanjude return mzap_list(&child_dir_zap, callback); 1907293802Sallanjude else 1908293802Sallanjude return fzap_list(spa, &child_dir_zap, callback); 1909293802Sallanjude} 1910241283Savg#endif 1911241283Savg 1912185029Spjd/* 1913185029Spjd * Find the object set given the object number of its dataset object 1914185029Spjd * and return its details in *objset 1915185029Spjd */ 1916185029Spjdstatic int 1917235390Savgzfs_mount_dataset(const spa_t *spa, uint64_t objnum, objset_phys_t *objset) 1918185029Spjd{ 1919185029Spjd dnode_phys_t dataset; 1920185029Spjd dsl_dataset_phys_t *ds; 1921185029Spjd 1922185029Spjd if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) { 1923235361Savg printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum); 1924185029Spjd return (EIO); 1925185029Spjd } 1926185029Spjd 1927185029Spjd ds = (dsl_dataset_phys_t *) &dataset.dn_bonus; 1928185029Spjd if (zio_read(spa, &ds->ds_bp, objset)) { 1929235361Savg printf("ZFS: can't read object set for dataset %ju\n", 1930235361Savg (uintmax_t)objnum); 1931185029Spjd return (EIO); 1932185029Spjd } 1933185029Spjd 1934185029Spjd return (0); 1935185029Spjd} 1936185029Spjd 1937185029Spjd/* 1938185029Spjd * Find the object set pointed to by the BOOTFS property or the root 1939185029Spjd * dataset if there is none and return its details in *objset 1940185029Spjd */ 1941185029Spjdstatic int 1942235390Savgzfs_get_root(const spa_t *spa, uint64_t *objid) 1943185029Spjd{ 1944185029Spjd dnode_phys_t dir, propdir; 1945185029Spjd uint64_t props, bootfs, root; 1946185029Spjd 1947235329Savg *objid = 0; 1948235329Savg 1949185029Spjd /* 1950185029Spjd * Start with the MOS directory object. 1951185029Spjd */ 1952185029Spjd if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) { 1953185029Spjd printf("ZFS: can't read MOS object directory\n"); 1954185029Spjd return (EIO); 1955185029Spjd } 1956185029Spjd 1957185029Spjd /* 1958185029Spjd * Lookup the pool_props and see if we can find a bootfs. 1959185029Spjd */ 1960185029Spjd if (zap_lookup(spa, &dir, DMU_POOL_PROPS, &props) == 0 1961185029Spjd && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0 1962198420Srnoland && zap_lookup(spa, &propdir, "bootfs", &bootfs) == 0 1963198420Srnoland && bootfs != 0) 1964235329Savg { 1965235329Savg *objid = bootfs; 1966235329Savg return (0); 1967235329Savg } 1968185029Spjd /* 1969185029Spjd * Lookup the root dataset directory 1970185029Spjd */ 1971185029Spjd if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &root) 1972185029Spjd || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) { 1973185029Spjd printf("ZFS: can't find root dsl_dir\n"); 1974185029Spjd return (EIO); 1975185029Spjd } 1976185029Spjd 1977185029Spjd /* 1978185029Spjd * Use the information from the dataset directory's bonus buffer 1979185029Spjd * to find the dataset object and from that the object set itself. 1980185029Spjd */ 1981185029Spjd dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus; 1982235329Savg *objid = dd->dd_head_dataset_obj; 1983235329Savg return (0); 1984185029Spjd} 1985185029Spjd 1986185029Spjdstatic int 1987235390Savgzfs_mount(const spa_t *spa, uint64_t rootobj, struct zfsmount *mount) 1988185029Spjd{ 1989219089Spjd 1990235329Savg mount->spa = spa; 1991235329Savg 1992185029Spjd /* 1993235329Savg * Find the root object set if not explicitly provided 1994185029Spjd */ 1995235329Savg if (rootobj == 0 && zfs_get_root(spa, &rootobj)) { 1996235329Savg printf("ZFS: can't find root filesystem\n"); 1997185029Spjd return (EIO); 1998185029Spjd } 1999185029Spjd 2000235329Savg if (zfs_mount_dataset(spa, rootobj, &mount->objset)) { 2001235329Savg printf("ZFS: can't open root filesystem\n"); 2002185029Spjd return (EIO); 2003185029Spjd } 2004185029Spjd 2005235329Savg mount->rootobj = rootobj; 2006235329Savg 2007185029Spjd return (0); 2008185029Spjd} 2009185029Spjd 2010219089Spjdstatic int 2011235329Savgzfs_spa_init(spa_t *spa) 2012235329Savg{ 2013235329Savg 2014235329Savg if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) { 2015235329Savg printf("ZFS: can't read MOS of pool %s\n", spa->spa_name); 2016235329Savg return (EIO); 2017235329Savg } 2018241289Savg if (spa->spa_mos.os_type != DMU_OST_META) { 2019241289Savg printf("ZFS: corrupted MOS of pool %s\n", spa->spa_name); 2020241289Savg return (EIO); 2021241289Savg } 2022235329Savg return (0); 2023235329Savg} 2024235329Savg 2025235329Savgstatic int 2026235390Savgzfs_dnode_stat(const spa_t *spa, dnode_phys_t *dn, struct stat *sb) 2027219089Spjd{ 2028219089Spjd 2029219089Spjd if (dn->dn_bonustype != DMU_OT_SA) { 2030219089Spjd znode_phys_t *zp = (znode_phys_t *)dn->dn_bonus; 2031219089Spjd 2032219089Spjd sb->st_mode = zp->zp_mode; 2033219089Spjd sb->st_uid = zp->zp_uid; 2034219089Spjd sb->st_gid = zp->zp_gid; 2035219089Spjd sb->st_size = zp->zp_size; 2036219089Spjd } else { 2037219089Spjd sa_hdr_phys_t *sahdrp; 2038219089Spjd int hdrsize; 2039219089Spjd size_t size = 0; 2040219089Spjd void *buf = NULL; 2041219089Spjd 2042219089Spjd if (dn->dn_bonuslen != 0) 2043219089Spjd sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn); 2044219089Spjd else { 2045219089Spjd if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) != 0) { 2046219089Spjd blkptr_t *bp = &dn->dn_spill; 2047219089Spjd int error; 2048219089Spjd 2049219089Spjd size = BP_GET_LSIZE(bp); 2050219089Spjd buf = zfs_alloc(size); 2051219089Spjd error = zio_read(spa, bp, buf); 2052219089Spjd if (error != 0) { 2053219089Spjd zfs_free(buf, size); 2054219089Spjd return (error); 2055219089Spjd } 2056219089Spjd sahdrp = buf; 2057219089Spjd } else { 2058219089Spjd return (EIO); 2059219089Spjd } 2060219089Spjd } 2061219089Spjd hdrsize = SA_HDR_SIZE(sahdrp); 2062219089Spjd sb->st_mode = *(uint64_t *)((char *)sahdrp + hdrsize + 2063219089Spjd SA_MODE_OFFSET); 2064219089Spjd sb->st_uid = *(uint64_t *)((char *)sahdrp + hdrsize + 2065219089Spjd SA_UID_OFFSET); 2066219089Spjd sb->st_gid = *(uint64_t *)((char *)sahdrp + hdrsize + 2067219089Spjd SA_GID_OFFSET); 2068219089Spjd sb->st_size = *(uint64_t *)((char *)sahdrp + hdrsize + 2069219089Spjd SA_SIZE_OFFSET); 2070219089Spjd if (buf != NULL) 2071219089Spjd zfs_free(buf, size); 2072219089Spjd } 2073219089Spjd 2074219089Spjd return (0); 2075219089Spjd} 2076219089Spjd 2077185029Spjd/* 2078185029Spjd * Lookup a file and return its dnode. 2079185029Spjd */ 2080185029Spjdstatic int 2081235329Savgzfs_lookup(const struct zfsmount *mount, const char *upath, dnode_phys_t *dnode) 2082185029Spjd{ 2083185029Spjd int rc; 2084185029Spjd uint64_t objnum, rootnum, parentnum; 2085235390Savg const spa_t *spa; 2086185029Spjd dnode_phys_t dn; 2087185029Spjd const char *p, *q; 2088185029Spjd char element[256]; 2089185029Spjd char path[1024]; 2090185029Spjd int symlinks_followed = 0; 2091219089Spjd struct stat sb; 2092185029Spjd 2093235329Savg spa = mount->spa; 2094235329Savg if (mount->objset.os_type != DMU_OST_ZFS) { 2095235361Savg printf("ZFS: unexpected object set type %ju\n", 2096235361Savg (uintmax_t)mount->objset.os_type); 2097185029Spjd return (EIO); 2098185029Spjd } 2099185029Spjd 2100185029Spjd /* 2101185029Spjd * Get the root directory dnode. 2102185029Spjd */ 2103235329Savg rc = objset_get_dnode(spa, &mount->objset, MASTER_NODE_OBJ, &dn); 2104185029Spjd if (rc) 2105185029Spjd return (rc); 2106185029Spjd 2107185029Spjd rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, &rootnum); 2108185029Spjd if (rc) 2109185029Spjd return (rc); 2110185029Spjd 2111235329Savg rc = objset_get_dnode(spa, &mount->objset, rootnum, &dn); 2112185029Spjd if (rc) 2113185029Spjd return (rc); 2114185029Spjd 2115185029Spjd objnum = rootnum; 2116185029Spjd p = upath; 2117185029Spjd while (p && *p) { 2118185029Spjd while (*p == '/') 2119185029Spjd p++; 2120185029Spjd if (!*p) 2121185029Spjd break; 2122185029Spjd q = strchr(p, '/'); 2123185029Spjd if (q) { 2124185029Spjd memcpy(element, p, q - p); 2125185029Spjd element[q - p] = 0; 2126185029Spjd p = q; 2127185029Spjd } else { 2128185029Spjd strcpy(element, p); 2129185029Spjd p = 0; 2130185029Spjd } 2131185029Spjd 2132219089Spjd rc = zfs_dnode_stat(spa, &dn, &sb); 2133219089Spjd if (rc) 2134219089Spjd return (rc); 2135219089Spjd if (!S_ISDIR(sb.st_mode)) 2136185029Spjd return (ENOTDIR); 2137185029Spjd 2138185029Spjd parentnum = objnum; 2139185029Spjd rc = zap_lookup(spa, &dn, element, &objnum); 2140185029Spjd if (rc) 2141185029Spjd return (rc); 2142185029Spjd objnum = ZFS_DIRENT_OBJ(objnum); 2143185029Spjd 2144235329Savg rc = objset_get_dnode(spa, &mount->objset, objnum, &dn); 2145185029Spjd if (rc) 2146185029Spjd return (rc); 2147185029Spjd 2148185029Spjd /* 2149185029Spjd * Check for symlink. 2150185029Spjd */ 2151219089Spjd rc = zfs_dnode_stat(spa, &dn, &sb); 2152219089Spjd if (rc) 2153219089Spjd return (rc); 2154219089Spjd if (S_ISLNK(sb.st_mode)) { 2155185029Spjd if (symlinks_followed > 10) 2156185029Spjd return (EMLINK); 2157185029Spjd symlinks_followed++; 2158185029Spjd 2159185029Spjd /* 2160185029Spjd * Read the link value and copy the tail of our 2161185029Spjd * current path onto the end. 2162185029Spjd */ 2163185029Spjd if (p) 2164219089Spjd strcpy(&path[sb.st_size], p); 2165185029Spjd else 2166219089Spjd path[sb.st_size] = 0; 2167294975Ssmh /* 2168294975Ssmh * Second test is purely to silence bogus compiler 2169294975Ssmh * warning about accessing past the end of dn_bonus. 2170294975Ssmh */ 2171294975Ssmh if (sb.st_size + sizeof(znode_phys_t) <= 2172294975Ssmh dn.dn_bonuslen && sizeof(znode_phys_t) <= 2173294975Ssmh sizeof(dn.dn_bonus)) { 2174185029Spjd memcpy(path, &dn.dn_bonus[sizeof(znode_phys_t)], 2175219089Spjd sb.st_size); 2176185029Spjd } else { 2177219089Spjd rc = dnode_read(spa, &dn, 0, path, sb.st_size); 2178185029Spjd if (rc) 2179185029Spjd return (rc); 2180185029Spjd } 2181185029Spjd 2182185029Spjd /* 2183185029Spjd * Restart with the new path, starting either at 2184185029Spjd * the root or at the parent depending whether or 2185185029Spjd * not the link is relative. 2186185029Spjd */ 2187185029Spjd p = path; 2188185029Spjd if (*p == '/') 2189185029Spjd objnum = rootnum; 2190185029Spjd else 2191185029Spjd objnum = parentnum; 2192235329Savg objset_get_dnode(spa, &mount->objset, objnum, &dn); 2193185029Spjd } 2194185029Spjd } 2195185029Spjd 2196185029Spjd *dnode = dn; 2197185029Spjd return (0); 2198185029Spjd} 2199