2 * Copyright (c) 2007 Doug Rabson
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
31 * Stand-alone ZFS file reader.
35 #include <sys/stdint.h>
48 * List of all vdevs, chained through v_alllink.
50 static vdev_list_t zfs_vdevs;
53 * List of ZFS features supported for read
55 static const char *features_for_read[] = {
56 "org.illumos:lz4_compress",
57 "com.delphix:hole_birth",
58 "com.delphix:extensible_dataset",
59 "com.delphix:embedded_data",
60 "org.open-zfs:large_blocks",
67 * List of all pools, chained through spa_link.
69 static spa_list_t zfs_pools;
71 static uint64_t zfs_crc64_table[256];
72 static const dnode_phys_t *dnode_cache_obj = 0;
73 static uint64_t dnode_cache_bn;
74 static char *dnode_cache_buf;
75 static char *zap_scratch;
76 static char *zfs_temp_buf, *zfs_temp_end, *zfs_temp_ptr;
78 #define TEMP_SIZE (1024 * 1024)
80 static int zio_read(const spa_t *spa, const blkptr_t *bp, void *buf);
81 static int zfs_get_root(const spa_t *spa, uint64_t *objid);
82 static int zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result);
83 static int zap_lookup(const spa_t *spa, const dnode_phys_t *dnode,
84 const char *name, uint64_t integer_size, uint64_t num_integers,
90 STAILQ_INIT(&zfs_vdevs);
91 STAILQ_INIT(&zfs_pools);
93 zfs_temp_buf = malloc(TEMP_SIZE);
94 zfs_temp_end = zfs_temp_buf + TEMP_SIZE;
95 zfs_temp_ptr = zfs_temp_buf;
96 dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE);
97 zap_scratch = malloc(SPA_MAXBLOCKSIZE);
103 zfs_alloc(size_t size)
107 if (zfs_temp_ptr + size > zfs_temp_end) {
108 printf("ZFS: out of temporary buffer space\n");
112 zfs_temp_ptr += size;
118 zfs_free(void *ptr, size_t size)
121 zfs_temp_ptr -= size;
122 if (zfs_temp_ptr != ptr) {
123 printf("ZFS: zfs_alloc()/zfs_free() mismatch\n");
129 xdr_int(const unsigned char **xdr, int *ip)
131 *ip = ((*xdr)[0] << 24)
140 xdr_u_int(const unsigned char **xdr, u_int *ip)
142 *ip = ((*xdr)[0] << 24)
151 xdr_uint64_t(const unsigned char **xdr, uint64_t *lp)
157 *lp = (((uint64_t) hi) << 32) | lo;
162 nvlist_find(const unsigned char *nvlist, const char *name, int type,
163 int* elementsp, void *valuep)
165 const unsigned char *p, *pair;
167 int encoded_size, decoded_size;
174 xdr_int(&p, &encoded_size);
175 xdr_int(&p, &decoded_size);
176 while (encoded_size && decoded_size) {
177 int namelen, pairtype, elements;
178 const char *pairname;
180 xdr_int(&p, &namelen);
181 pairname = (const char*) p;
182 p += roundup(namelen, 4);
183 xdr_int(&p, &pairtype);
185 if (!memcmp(name, pairname, namelen) && type == pairtype) {
186 xdr_int(&p, &elements);
188 *elementsp = elements;
189 if (type == DATA_TYPE_UINT64) {
190 xdr_uint64_t(&p, (uint64_t *) valuep);
192 } else if (type == DATA_TYPE_STRING) {
195 (*(const char**) valuep) = (const char*) p;
197 } else if (type == DATA_TYPE_NVLIST
198 || type == DATA_TYPE_NVLIST_ARRAY) {
199 (*(const unsigned char**) valuep) =
200 (const unsigned char*) p;
207 * Not the pair we are looking for, skip to the next one.
209 p = pair + encoded_size;
213 xdr_int(&p, &encoded_size);
214 xdr_int(&p, &decoded_size);
221 nvlist_check_features_for_read(const unsigned char *nvlist)
223 const unsigned char *p, *pair;
225 int encoded_size, decoded_size;
235 xdr_int(&p, &encoded_size);
236 xdr_int(&p, &decoded_size);
237 while (encoded_size && decoded_size) {
238 int namelen, pairtype;
239 const char *pairname;
244 xdr_int(&p, &namelen);
245 pairname = (const char*) p;
246 p += roundup(namelen, 4);
247 xdr_int(&p, &pairtype);
249 for (i = 0; features_for_read[i] != NULL; i++) {
250 if (!memcmp(pairname, features_for_read[i], namelen)) {
257 printf("ZFS: unsupported feature: %s\n", pairname);
261 p = pair + encoded_size;
264 xdr_int(&p, &encoded_size);
265 xdr_int(&p, &decoded_size);
272 * Return the next nvlist in an nvlist array.
274 static const unsigned char *
275 nvlist_next(const unsigned char *nvlist)
277 const unsigned char *p, *pair;
279 int encoded_size, decoded_size;
286 xdr_int(&p, &encoded_size);
287 xdr_int(&p, &decoded_size);
288 while (encoded_size && decoded_size) {
289 p = pair + encoded_size;
292 xdr_int(&p, &encoded_size);
293 xdr_int(&p, &decoded_size);
301 static const unsigned char *
302 nvlist_print(const unsigned char *nvlist, unsigned int indent)
304 static const char* typenames[] = {
315 "DATA_TYPE_BYTE_ARRAY",
316 "DATA_TYPE_INT16_ARRAY",
317 "DATA_TYPE_UINT16_ARRAY",
318 "DATA_TYPE_INT32_ARRAY",
319 "DATA_TYPE_UINT32_ARRAY",
320 "DATA_TYPE_INT64_ARRAY",
321 "DATA_TYPE_UINT64_ARRAY",
322 "DATA_TYPE_STRING_ARRAY",
325 "DATA_TYPE_NVLIST_ARRAY",
326 "DATA_TYPE_BOOLEAN_VALUE",
329 "DATA_TYPE_BOOLEAN_ARRAY",
330 "DATA_TYPE_INT8_ARRAY",
331 "DATA_TYPE_UINT8_ARRAY"
335 const unsigned char *p, *pair;
337 int encoded_size, decoded_size;
344 xdr_int(&p, &encoded_size);
345 xdr_int(&p, &decoded_size);
346 while (encoded_size && decoded_size) {
347 int namelen, pairtype, elements;
348 const char *pairname;
350 xdr_int(&p, &namelen);
351 pairname = (const char*) p;
352 p += roundup(namelen, 4);
353 xdr_int(&p, &pairtype);
355 for (i = 0; i < indent; i++)
357 printf("%s %s", typenames[pairtype], pairname);
359 xdr_int(&p, &elements);
361 case DATA_TYPE_UINT64: {
363 xdr_uint64_t(&p, &val);
364 printf(" = 0x%jx\n", (uintmax_t)val);
368 case DATA_TYPE_STRING: {
371 printf(" = \"%s\"\n", p);
375 case DATA_TYPE_NVLIST:
377 nvlist_print(p, indent + 1);
380 case DATA_TYPE_NVLIST_ARRAY:
381 for (j = 0; j < elements; j++) {
383 p = nvlist_print(p, indent + 1);
384 if (j != elements - 1) {
385 for (i = 0; i < indent; i++)
387 printf("%s %s", typenames[pairtype], pairname);
396 p = pair + encoded_size;
399 xdr_int(&p, &encoded_size);
400 xdr_int(&p, &decoded_size);
409 vdev_read_phys(vdev_t *vdev, const blkptr_t *bp, void *buf,
410 off_t offset, size_t size)
415 if (!vdev->v_phys_read)
419 psize = BP_GET_PSIZE(bp);
424 /*printf("ZFS: reading %d bytes at 0x%jx to %p\n", psize, (uintmax_t)offset, buf);*/
425 rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize);
428 if (bp && zio_checksum_verify(vdev->spa, bp, buf))
435 vdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
436 off_t offset, size_t bytes)
439 return (vdev_read_phys(vdev, bp, buf,
440 offset + VDEV_LABEL_START_SIZE, bytes));
445 vdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
446 off_t offset, size_t bytes)
452 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
453 if (kid->v_state != VDEV_STATE_HEALTHY)
455 rc = kid->v_read(kid, bp, buf, offset, bytes);
464 vdev_replacing_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
465 off_t offset, size_t bytes)
470 * Here we should have two kids:
471 * First one which is the one we are replacing and we can trust
472 * only this one to have valid data, but it might not be present.
473 * Second one is that one we are replacing with. It is most likely
474 * healthy, but we can't trust it has needed data, so we won't use it.
476 kid = STAILQ_FIRST(&vdev->v_children);
479 if (kid->v_state != VDEV_STATE_HEALTHY)
481 return (kid->v_read(kid, bp, buf, offset, bytes));
485 vdev_find(uint64_t guid)
489 STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink)
490 if (vdev->v_guid == guid)
497 vdev_create(uint64_t guid, vdev_read_t *read)
501 vdev = malloc(sizeof(vdev_t));
502 memset(vdev, 0, sizeof(vdev_t));
503 STAILQ_INIT(&vdev->v_children);
505 vdev->v_state = VDEV_STATE_OFFLINE;
507 vdev->v_phys_read = 0;
508 vdev->v_read_priv = 0;
509 STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink);
515 vdev_init_from_nvlist(const unsigned char *nvlist, vdev_t *pvdev,
516 vdev_t **vdevp, int is_newer)
519 uint64_t guid, id, ashift, nparity;
523 const unsigned char *kids;
524 int nkids, i, is_new;
525 uint64_t is_offline, is_faulted, is_degraded, is_removed, isnt_present;
527 if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID,
528 DATA_TYPE_UINT64, 0, &guid)
529 || nvlist_find(nvlist, ZPOOL_CONFIG_ID,
530 DATA_TYPE_UINT64, 0, &id)
531 || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE,
532 DATA_TYPE_STRING, 0, &type)) {
533 printf("ZFS: can't find vdev details\n");
537 if (strcmp(type, VDEV_TYPE_MIRROR)
538 && strcmp(type, VDEV_TYPE_DISK)
540 && strcmp(type, VDEV_TYPE_FILE)
542 && strcmp(type, VDEV_TYPE_RAIDZ)
543 && strcmp(type, VDEV_TYPE_REPLACING)) {
544 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
548 is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0;
550 nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, 0,
552 nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, 0,
554 nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, 0,
556 nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, 0,
558 nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, 0,
561 vdev = vdev_find(guid);
565 if (!strcmp(type, VDEV_TYPE_MIRROR))
566 vdev = vdev_create(guid, vdev_mirror_read);
567 else if (!strcmp(type, VDEV_TYPE_RAIDZ))
568 vdev = vdev_create(guid, vdev_raidz_read);
569 else if (!strcmp(type, VDEV_TYPE_REPLACING))
570 vdev = vdev_create(guid, vdev_replacing_read);
572 vdev = vdev_create(guid, vdev_disk_read);
575 vdev->v_top = pvdev != NULL ? pvdev : vdev;
576 if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT,
577 DATA_TYPE_UINT64, 0, &ashift) == 0)
578 vdev->v_ashift = ashift;
581 if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY,
582 DATA_TYPE_UINT64, 0, &nparity) == 0)
583 vdev->v_nparity = nparity;
586 if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH,
587 DATA_TYPE_STRING, 0, &path) == 0) {
588 if (strncmp(path, "/dev/", 5) == 0)
590 vdev->v_name = strdup(path);
592 if (!strcmp(type, "raidz")) {
593 if (vdev->v_nparity == 1)
594 vdev->v_name = "raidz1";
595 else if (vdev->v_nparity == 2)
596 vdev->v_name = "raidz2";
597 else if (vdev->v_nparity == 3)
598 vdev->v_name = "raidz3";
600 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
604 vdev->v_name = strdup(type);
611 if (is_new || is_newer) {
613 * This is either new vdev or we've already seen this vdev,
614 * but from an older vdev label, so let's refresh its state
615 * from the newer label.
618 vdev->v_state = VDEV_STATE_OFFLINE;
620 vdev->v_state = VDEV_STATE_REMOVED;
622 vdev->v_state = VDEV_STATE_FAULTED;
623 else if (is_degraded)
624 vdev->v_state = VDEV_STATE_DEGRADED;
625 else if (isnt_present)
626 vdev->v_state = VDEV_STATE_CANT_OPEN;
629 rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN,
630 DATA_TYPE_NVLIST_ARRAY, &nkids, &kids);
632 * Its ok if we don't have any kids.
635 vdev->v_nchildren = nkids;
636 for (i = 0; i < nkids; i++) {
637 rc = vdev_init_from_nvlist(kids, vdev, &kid, is_newer);
641 STAILQ_INSERT_TAIL(&vdev->v_children, kid,
643 kids = nvlist_next(kids);
646 vdev->v_nchildren = 0;
655 vdev_set_state(vdev_t *vdev)
662 * A mirror or raidz is healthy if all its kids are healthy. A
663 * mirror is degraded if any of its kids is healthy; a raidz
664 * is degraded if at most nparity kids are offline.
666 if (STAILQ_FIRST(&vdev->v_children)) {
669 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
670 if (kid->v_state == VDEV_STATE_HEALTHY)
676 vdev->v_state = VDEV_STATE_HEALTHY;
678 if (vdev->v_read == vdev_mirror_read) {
680 vdev->v_state = VDEV_STATE_DEGRADED;
682 vdev->v_state = VDEV_STATE_OFFLINE;
684 } else if (vdev->v_read == vdev_raidz_read) {
685 if (bad_kids > vdev->v_nparity) {
686 vdev->v_state = VDEV_STATE_OFFLINE;
688 vdev->v_state = VDEV_STATE_DEGRADED;
696 spa_find_by_guid(uint64_t guid)
700 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
701 if (spa->spa_guid == guid)
708 spa_find_by_name(const char *name)
712 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
713 if (!strcmp(spa->spa_name, name))
721 spa_get_primary(void)
724 return (STAILQ_FIRST(&zfs_pools));
728 spa_get_primary_vdev(const spa_t *spa)
734 spa = spa_get_primary();
737 vdev = STAILQ_FIRST(&spa->spa_vdevs);
740 for (kid = STAILQ_FIRST(&vdev->v_children); kid != NULL;
741 kid = STAILQ_FIRST(&vdev->v_children))
748 spa_create(uint64_t guid)
752 spa = malloc(sizeof(spa_t));
753 memset(spa, 0, sizeof(spa_t));
754 STAILQ_INIT(&spa->spa_vdevs);
755 spa->spa_guid = guid;
756 STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link);
762 state_name(vdev_state_t state)
764 static const char* names[] = {
779 #define pager_printf printf
784 pager_printf(const char *fmt, ...)
790 vsprintf(line, fmt, args);
792 return (pager_output(line));
797 #define STATUS_FORMAT " %s %s\n"
800 print_state(int indent, const char *name, vdev_state_t state)
806 for (i = 0; i < indent; i++)
809 return (pager_printf(STATUS_FORMAT, buf, state_name(state)));
814 vdev_status(vdev_t *vdev, int indent)
818 ret = print_state(indent, vdev->v_name, vdev->v_state);
822 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
823 ret = vdev_status(kid, indent + 1);
831 spa_status(spa_t *spa)
833 static char bootfs[ZFS_MAXNAMELEN];
836 int good_kids, bad_kids, degraded_kids, ret;
839 ret = pager_printf(" pool: %s\n", spa->spa_name);
843 if (zfs_get_root(spa, &rootid) == 0 &&
844 zfs_rlookup(spa, rootid, bootfs) == 0) {
845 if (bootfs[0] == '\0')
846 ret = pager_printf("bootfs: %s\n", spa->spa_name);
848 ret = pager_printf("bootfs: %s/%s\n", spa->spa_name,
853 ret = pager_printf("config:\n\n");
856 ret = pager_printf(STATUS_FORMAT, "NAME", "STATE");
863 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
864 if (vdev->v_state == VDEV_STATE_HEALTHY)
866 else if (vdev->v_state == VDEV_STATE_DEGRADED)
872 state = VDEV_STATE_CLOSED;
873 if (good_kids > 0 && (degraded_kids + bad_kids) == 0)
874 state = VDEV_STATE_HEALTHY;
875 else if ((good_kids + degraded_kids) > 0)
876 state = VDEV_STATE_DEGRADED;
878 ret = print_state(0, spa->spa_name, state);
881 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
882 ret = vdev_status(vdev, 1);
893 int first = 1, ret = 0;
895 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
897 ret = pager_printf("\n");
902 ret = spa_status(spa);
910 vdev_probe(vdev_phys_read_t *read, void *read_priv, spa_t **spap)
913 vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch;
915 vdev_t *vdev, *top_vdev, *pool_vdev;
918 const unsigned char *nvlist;
921 uint64_t pool_txg, pool_guid;
923 const char *pool_name;
924 const unsigned char *vdevs;
925 const unsigned char *features;
928 const struct uberblock *up;
931 * Load the vdev label and figure out which
932 * uberblock is most current.
934 memset(&vtmp, 0, sizeof(vtmp));
935 vtmp.v_phys_read = read;
936 vtmp.v_read_priv = read_priv;
937 off = offsetof(vdev_label_t, vl_vdev_phys);
939 BP_SET_LSIZE(&bp, sizeof(vdev_phys_t));
940 BP_SET_PSIZE(&bp, sizeof(vdev_phys_t));
941 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
942 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
943 DVA_SET_OFFSET(BP_IDENTITY(&bp), off);
944 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
945 if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0))
948 if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) {
952 nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4;
954 if (nvlist_find(nvlist,
955 ZPOOL_CONFIG_VERSION,
956 DATA_TYPE_UINT64, 0, &val)) {
960 if (!SPA_VERSION_IS_SUPPORTED(val)) {
961 printf("ZFS: unsupported ZFS version %u (should be %u)\n",
962 (unsigned) val, (unsigned) SPA_VERSION);
966 /* Check ZFS features for read */
967 if (nvlist_find(nvlist,
968 ZPOOL_CONFIG_FEATURES_FOR_READ,
969 DATA_TYPE_NVLIST, 0, &features) == 0
970 && nvlist_check_features_for_read(features) != 0)
973 if (nvlist_find(nvlist,
974 ZPOOL_CONFIG_POOL_STATE,
975 DATA_TYPE_UINT64, 0, &val)) {
979 if (val == POOL_STATE_DESTROYED) {
980 /* We don't boot only from destroyed pools. */
984 if (nvlist_find(nvlist,
985 ZPOOL_CONFIG_POOL_TXG,
986 DATA_TYPE_UINT64, 0, &pool_txg)
987 || nvlist_find(nvlist,
988 ZPOOL_CONFIG_POOL_GUID,
989 DATA_TYPE_UINT64, 0, &pool_guid)
990 || nvlist_find(nvlist,
991 ZPOOL_CONFIG_POOL_NAME,
992 DATA_TYPE_STRING, 0, &pool_name)) {
994 * Cache and spare devices end up here - just ignore
997 /*printf("ZFS: can't find pool details\n");*/
1002 (void) nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64, 0,
1008 * Create the pool if this is the first time we've seen it.
1010 spa = spa_find_by_guid(pool_guid);
1012 spa = spa_create(pool_guid);
1013 spa->spa_name = strdup(pool_name);
1015 if (pool_txg > spa->spa_txg) {
1016 spa->spa_txg = pool_txg;
1022 * Get the vdev tree and create our in-core copy of it.
1023 * If we already have a vdev with this guid, this must
1024 * be some kind of alias (overlapping slices, dangerously dedicated
1027 if (nvlist_find(nvlist,
1029 DATA_TYPE_UINT64, 0, &guid)) {
1032 vdev = vdev_find(guid);
1033 if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */
1036 if (nvlist_find(nvlist,
1037 ZPOOL_CONFIG_VDEV_TREE,
1038 DATA_TYPE_NVLIST, 0, &vdevs)) {
1042 rc = vdev_init_from_nvlist(vdevs, NULL, &top_vdev, is_newer);
1047 * Add the toplevel vdev to the pool if its not already there.
1049 STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink)
1050 if (top_vdev == pool_vdev)
1052 if (!pool_vdev && top_vdev) {
1053 top_vdev->spa = spa;
1054 STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink);
1058 * We should already have created an incomplete vdev for this
1059 * vdev. Find it and initialise it with our read proc.
1061 vdev = vdev_find(guid);
1063 vdev->v_phys_read = read;
1064 vdev->v_read_priv = read_priv;
1065 vdev->v_state = VDEV_STATE_HEALTHY;
1067 printf("ZFS: inconsistent nvlist contents\n");
1072 * Re-evaluate top-level vdev state.
1074 vdev_set_state(top_vdev);
1077 * Ok, we are happy with the pool so far. Lets find
1078 * the best uberblock and then we can actually access
1079 * the contents of the pool.
1081 upbuf = zfs_alloc(VDEV_UBERBLOCK_SIZE(vdev));
1082 up = (const struct uberblock *)upbuf;
1084 i < VDEV_UBERBLOCK_COUNT(vdev);
1086 off = VDEV_UBERBLOCK_OFFSET(vdev, i);
1088 DVA_SET_OFFSET(&bp.blk_dva[0], off);
1089 BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1090 BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1091 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
1092 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
1093 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
1095 if (vdev_read_phys(vdev, &bp, upbuf, off, 0))
1098 if (up->ub_magic != UBERBLOCK_MAGIC)
1100 if (up->ub_txg < spa->spa_txg)
1102 if (up->ub_txg > spa->spa_uberblock.ub_txg) {
1103 spa->spa_uberblock = *up;
1104 } else if (up->ub_txg == spa->spa_uberblock.ub_txg) {
1105 if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp)
1106 spa->spa_uberblock = *up;
1109 zfs_free(upbuf, VDEV_UBERBLOCK_SIZE(vdev));
1122 for (v = 0; v < 32; v++)
1129 zio_read_gang(const spa_t *spa, const blkptr_t *bp, void *buf)
1132 zio_gbh_phys_t zio_gb;
1136 /* Artificial BP for gang block header. */
1138 BP_SET_PSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1139 BP_SET_LSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1140 BP_SET_CHECKSUM(&gbh_bp, ZIO_CHECKSUM_GANG_HEADER);
1141 BP_SET_COMPRESS(&gbh_bp, ZIO_COMPRESS_OFF);
1142 for (i = 0; i < SPA_DVAS_PER_BP; i++)
1143 DVA_SET_GANG(&gbh_bp.blk_dva[i], 0);
1145 /* Read gang header block using the artificial BP. */
1146 if (zio_read(spa, &gbh_bp, &zio_gb))
1150 for (i = 0; i < SPA_GBH_NBLKPTRS; i++) {
1151 blkptr_t *gbp = &zio_gb.zg_blkptr[i];
1153 if (BP_IS_HOLE(gbp))
1155 if (zio_read(spa, gbp, pbuf))
1157 pbuf += BP_GET_PSIZE(gbp);
1160 if (zio_checksum_verify(spa, bp, buf))
1166 zio_read(const spa_t *spa, const blkptr_t *bp, void *buf)
1168 int cpfunc = BP_GET_COMPRESS(bp);
1169 uint64_t align, size;
1174 * Process data embedded in block pointer
1176 if (BP_IS_EMBEDDED(bp)) {
1177 ASSERT(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
1179 size = BPE_GET_PSIZE(bp);
1180 ASSERT(size <= BPE_PAYLOAD_SIZE);
1182 if (cpfunc != ZIO_COMPRESS_OFF)
1183 pbuf = zfs_alloc(size);
1187 decode_embedded_bp_compressed(bp, pbuf);
1190 if (cpfunc != ZIO_COMPRESS_OFF) {
1191 error = zio_decompress_data(cpfunc, pbuf,
1192 size, buf, BP_GET_LSIZE(bp));
1193 zfs_free(pbuf, size);
1196 printf("ZFS: i/o error - unable to decompress block pointer data, error %d\n",
1203 for (i = 0; i < SPA_DVAS_PER_BP; i++) {
1204 const dva_t *dva = &bp->blk_dva[i];
1209 if (!dva->dva_word[0] && !dva->dva_word[1])
1212 vdevid = DVA_GET_VDEV(dva);
1213 offset = DVA_GET_OFFSET(dva);
1214 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
1215 if (vdev->v_id == vdevid)
1218 if (!vdev || !vdev->v_read)
1221 size = BP_GET_PSIZE(bp);
1222 if (vdev->v_read == vdev_raidz_read) {
1223 align = 1ULL << vdev->v_top->v_ashift;
1224 if (P2PHASE(size, align) != 0)
1225 size = P2ROUNDUP(size, align);
1227 if (size != BP_GET_PSIZE(bp) || cpfunc != ZIO_COMPRESS_OFF)
1228 pbuf = zfs_alloc(size);
1232 if (DVA_GET_GANG(dva))
1233 error = zio_read_gang(spa, bp, pbuf);
1235 error = vdev->v_read(vdev, bp, pbuf, offset, size);
1237 if (cpfunc != ZIO_COMPRESS_OFF)
1238 error = zio_decompress_data(cpfunc, pbuf,
1239 BP_GET_PSIZE(bp), buf, BP_GET_LSIZE(bp));
1240 else if (size != BP_GET_PSIZE(bp))
1241 bcopy(pbuf, buf, BP_GET_PSIZE(bp));
1244 zfs_free(pbuf, size);
1249 printf("ZFS: i/o error - all block copies unavailable\n");
1254 dnode_read(const spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen)
1256 int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
1257 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1258 int nlevels = dnode->dn_nlevels;
1261 if (bsize > SPA_MAXBLOCKSIZE) {
1262 printf("ZFS: I/O error - blocks larger than %llu are not "
1263 "supported\n", SPA_MAXBLOCKSIZE);
1268 * Note: bsize may not be a power of two here so we need to do an
1269 * actual divide rather than a bitshift.
1271 while (buflen > 0) {
1272 uint64_t bn = offset / bsize;
1273 int boff = offset % bsize;
1275 const blkptr_t *indbp;
1278 if (bn > dnode->dn_maxblkid)
1281 if (dnode == dnode_cache_obj && bn == dnode_cache_bn)
1284 indbp = dnode->dn_blkptr;
1285 for (i = 0; i < nlevels; i++) {
1287 * Copy the bp from the indirect array so that
1288 * we can re-use the scratch buffer for multi-level
1291 ibn = bn >> ((nlevels - i - 1) * ibshift);
1292 ibn &= ((1 << ibshift) - 1);
1294 if (BP_IS_HOLE(&bp)) {
1295 memset(dnode_cache_buf, 0, bsize);
1298 rc = zio_read(spa, &bp, dnode_cache_buf);
1301 indbp = (const blkptr_t *) dnode_cache_buf;
1303 dnode_cache_obj = dnode;
1304 dnode_cache_bn = bn;
1308 * The buffer contains our data block. Copy what we
1309 * need from it and loop.
1312 if (i > buflen) i = buflen;
1313 memcpy(buf, &dnode_cache_buf[boff], i);
1314 buf = ((char*) buf) + i;
1323 * Lookup a value in a microzap directory. Assumes that the zap
1324 * scratch buffer contains the directory contents.
1327 mzap_lookup(const dnode_phys_t *dnode, const char *name, uint64_t *value)
1329 const mzap_phys_t *mz;
1330 const mzap_ent_phys_t *mze;
1335 * Microzap objects use exactly one block. Read the whole
1338 size = dnode->dn_datablkszsec * 512;
1340 mz = (const mzap_phys_t *) zap_scratch;
1341 chunks = size / MZAP_ENT_LEN - 1;
1343 for (i = 0; i < chunks; i++) {
1344 mze = &mz->mz_chunk[i];
1345 if (!strcmp(mze->mze_name, name)) {
1346 *value = mze->mze_value;
1355 * Compare a name with a zap leaf entry. Return non-zero if the name
1359 fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name)
1362 const zap_leaf_chunk_t *nc;
1365 namelen = zc->l_entry.le_name_numints;
1367 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1369 while (namelen > 0) {
1372 if (len > ZAP_LEAF_ARRAY_BYTES)
1373 len = ZAP_LEAF_ARRAY_BYTES;
1374 if (memcmp(p, nc->l_array.la_array, len))
1378 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1385 * Extract a uint64_t value from a zap leaf entry.
1388 fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc)
1390 const zap_leaf_chunk_t *vc;
1395 vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk);
1396 for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) {
1397 value = (value << 8) | p[i];
1404 stv(int len, void *addr, uint64_t value)
1408 *(uint8_t *)addr = value;
1411 *(uint16_t *)addr = value;
1414 *(uint32_t *)addr = value;
1417 *(uint64_t *)addr = value;
1423 * Extract a array from a zap leaf entry.
1426 fzap_leaf_array(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc,
1427 uint64_t integer_size, uint64_t num_integers, void *buf)
1429 uint64_t array_int_len = zc->l_entry.le_value_intlen;
1431 uint64_t *u64 = buf;
1433 int len = MIN(zc->l_entry.le_value_numints, num_integers);
1434 int chunk = zc->l_entry.le_value_chunk;
1437 if (integer_size == 8 && len == 1) {
1438 *u64 = fzap_leaf_value(zl, zc);
1443 struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(zl, chunk).l_array;
1446 ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(zl));
1447 for (i = 0; i < ZAP_LEAF_ARRAY_BYTES && len > 0; i++) {
1448 value = (value << 8) | la->la_array[i];
1450 if (byten == array_int_len) {
1451 stv(integer_size, p, value);
1459 chunk = la->la_next;
1464 * Lookup a value in a fatzap directory. Assumes that the zap scratch
1465 * buffer contains the directory header.
1468 fzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name,
1469 uint64_t integer_size, uint64_t num_integers, void *value)
1471 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1472 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1478 if (zh.zap_magic != ZAP_MAGIC)
1481 z.zap_block_shift = ilog2(bsize);
1482 z.zap_phys = (zap_phys_t *) zap_scratch;
1485 * Figure out where the pointer table is and read it in if necessary.
1487 if (zh.zap_ptrtbl.zt_blk) {
1488 rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize,
1489 zap_scratch, bsize);
1492 ptrtbl = (uint64_t *) zap_scratch;
1494 ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0);
1497 hash = zap_hash(zh.zap_salt, name);
1500 zl.l_bs = z.zap_block_shift;
1502 off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs;
1503 zap_leaf_chunk_t *zc;
1505 rc = dnode_read(spa, dnode, off, zap_scratch, bsize);
1509 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1512 * Make sure this chunk matches our hash.
1514 if (zl.l_phys->l_hdr.lh_prefix_len > 0
1515 && zl.l_phys->l_hdr.lh_prefix
1516 != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len))
1520 * Hash within the chunk to find our entry.
1522 int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len);
1523 int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1);
1524 h = zl.l_phys->l_hash[h];
1527 zc = &ZAP_LEAF_CHUNK(&zl, h);
1528 while (zc->l_entry.le_hash != hash) {
1529 if (zc->l_entry.le_next == 0xffff) {
1533 zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next);
1535 if (fzap_name_equal(&zl, zc, name)) {
1536 if (zc->l_entry.le_value_intlen * zc->l_entry.le_value_numints >
1537 integer_size * num_integers)
1539 fzap_leaf_array(&zl, zc, integer_size, num_integers, value);
1547 * Lookup a name in a zap object and return its value as a uint64_t.
1550 zap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name,
1551 uint64_t integer_size, uint64_t num_integers, void *value)
1555 size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1557 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1561 zap_type = *(uint64_t *) zap_scratch;
1562 if (zap_type == ZBT_MICRO)
1563 return mzap_lookup(dnode, name, value);
1564 else if (zap_type == ZBT_HEADER) {
1565 return fzap_lookup(spa, dnode, name, integer_size,
1566 num_integers, value);
1568 printf("ZFS: invalid zap_type=%d\n", (int)zap_type);
1573 * List a microzap directory. Assumes that the zap scratch buffer contains
1574 * the directory contents.
1577 mzap_list(const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1579 const mzap_phys_t *mz;
1580 const mzap_ent_phys_t *mze;
1585 * Microzap objects use exactly one block. Read the whole
1588 size = dnode->dn_datablkszsec * 512;
1589 mz = (const mzap_phys_t *) zap_scratch;
1590 chunks = size / MZAP_ENT_LEN - 1;
1592 for (i = 0; i < chunks; i++) {
1593 mze = &mz->mz_chunk[i];
1594 if (mze->mze_name[0]) {
1595 rc = callback(mze->mze_name, mze->mze_value);
1605 * List a fatzap directory. Assumes that the zap scratch buffer contains
1606 * the directory header.
1609 fzap_list(const spa_t *spa, const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1611 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1612 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1616 if (zh.zap_magic != ZAP_MAGIC)
1619 z.zap_block_shift = ilog2(bsize);
1620 z.zap_phys = (zap_phys_t *) zap_scratch;
1623 * This assumes that the leaf blocks start at block 1. The
1624 * documentation isn't exactly clear on this.
1627 zl.l_bs = z.zap_block_shift;
1628 for (i = 0; i < zh.zap_num_leafs; i++) {
1629 off_t off = (i + 1) << zl.l_bs;
1633 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1636 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1638 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1639 zap_leaf_chunk_t *zc, *nc;
1642 zc = &ZAP_LEAF_CHUNK(&zl, j);
1643 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1645 namelen = zc->l_entry.le_name_numints;
1646 if (namelen > sizeof(name))
1647 namelen = sizeof(name);
1650 * Paste the name back together.
1652 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
1654 while (namelen > 0) {
1657 if (len > ZAP_LEAF_ARRAY_BYTES)
1658 len = ZAP_LEAF_ARRAY_BYTES;
1659 memcpy(p, nc->l_array.la_array, len);
1662 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
1666 * Assume the first eight bytes of the value are
1669 value = fzap_leaf_value(&zl, zc);
1671 //printf("%s 0x%jx\n", name, (uintmax_t)value);
1672 rc = callback((const char *)name, value);
1681 static int zfs_printf(const char *name, uint64_t value __unused)
1684 printf("%s\n", name);
1690 * List a zap directory.
1693 zap_list(const spa_t *spa, const dnode_phys_t *dnode)
1696 size_t size = dnode->dn_datablkszsec * 512;
1698 if (dnode_read(spa, dnode, 0, zap_scratch, size))
1701 zap_type = *(uint64_t *) zap_scratch;
1702 if (zap_type == ZBT_MICRO)
1703 return mzap_list(dnode, zfs_printf);
1705 return fzap_list(spa, dnode, zfs_printf);
1709 objset_get_dnode(const spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode)
1713 offset = objnum * sizeof(dnode_phys_t);
1714 return dnode_read(spa, &os->os_meta_dnode, offset,
1715 dnode, sizeof(dnode_phys_t));
1719 mzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1721 const mzap_phys_t *mz;
1722 const mzap_ent_phys_t *mze;
1727 * Microzap objects use exactly one block. Read the whole
1730 size = dnode->dn_datablkszsec * 512;
1732 mz = (const mzap_phys_t *) zap_scratch;
1733 chunks = size / MZAP_ENT_LEN - 1;
1735 for (i = 0; i < chunks; i++) {
1736 mze = &mz->mz_chunk[i];
1737 if (value == mze->mze_value) {
1738 strcpy(name, mze->mze_name);
1747 fzap_name_copy(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, char *name)
1750 const zap_leaf_chunk_t *nc;
1753 namelen = zc->l_entry.le_name_numints;
1755 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1757 while (namelen > 0) {
1760 if (len > ZAP_LEAF_ARRAY_BYTES)
1761 len = ZAP_LEAF_ARRAY_BYTES;
1762 memcpy(p, nc->l_array.la_array, len);
1765 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1772 fzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1774 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1775 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1779 if (zh.zap_magic != ZAP_MAGIC)
1782 z.zap_block_shift = ilog2(bsize);
1783 z.zap_phys = (zap_phys_t *) zap_scratch;
1786 * This assumes that the leaf blocks start at block 1. The
1787 * documentation isn't exactly clear on this.
1790 zl.l_bs = z.zap_block_shift;
1791 for (i = 0; i < zh.zap_num_leafs; i++) {
1792 off_t off = (i + 1) << zl.l_bs;
1794 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1797 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1799 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1800 zap_leaf_chunk_t *zc;
1802 zc = &ZAP_LEAF_CHUNK(&zl, j);
1803 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1805 if (zc->l_entry.le_value_intlen != 8 ||
1806 zc->l_entry.le_value_numints != 1)
1809 if (fzap_leaf_value(&zl, zc) == value) {
1810 fzap_name_copy(&zl, zc, name);
1820 zap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1824 size_t size = dnode->dn_datablkszsec * 512;
1826 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1830 zap_type = *(uint64_t *) zap_scratch;
1831 if (zap_type == ZBT_MICRO)
1832 return mzap_rlookup(spa, dnode, name, value);
1834 return fzap_rlookup(spa, dnode, name, value);
1838 zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result)
1841 char component[256];
1842 uint64_t dir_obj, parent_obj, child_dir_zapobj;
1843 dnode_phys_t child_dir_zap, dataset, dir, parent;
1845 dsl_dataset_phys_t *ds;
1849 p = &name[sizeof(name) - 1];
1852 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1853 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1856 ds = (dsl_dataset_phys_t *)&dataset.dn_bonus;
1857 dir_obj = ds->ds_dir_obj;
1860 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir) != 0)
1862 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1864 /* Actual loop condition. */
1865 parent_obj = dd->dd_parent_obj;
1866 if (parent_obj == 0)
1869 if (objset_get_dnode(spa, &spa->spa_mos, parent_obj, &parent) != 0)
1871 dd = (dsl_dir_phys_t *)&parent.dn_bonus;
1872 child_dir_zapobj = dd->dd_child_dir_zapobj;
1873 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1875 if (zap_rlookup(spa, &child_dir_zap, component, dir_obj) != 0)
1878 len = strlen(component);
1880 memcpy(p, component, len);
1884 /* Actual loop iteration. */
1885 dir_obj = parent_obj;
1896 zfs_lookup_dataset(const spa_t *spa, const char *name, uint64_t *objnum)
1899 uint64_t dir_obj, child_dir_zapobj;
1900 dnode_phys_t child_dir_zap, dir;
1904 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir))
1906 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof (dir_obj),
1912 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir))
1914 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1918 /* Actual loop condition #1. */
1924 memcpy(element, p, q - p);
1925 element[q - p] = '\0';
1932 child_dir_zapobj = dd->dd_child_dir_zapobj;
1933 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1936 /* Actual loop condition #2. */
1937 if (zap_lookup(spa, &child_dir_zap, element, sizeof (dir_obj),
1942 *objnum = dd->dd_head_dataset_obj;
1948 zfs_list_dataset(const spa_t *spa, uint64_t objnum/*, int pos, char *entry*/)
1950 uint64_t dir_obj, child_dir_zapobj;
1951 dnode_phys_t child_dir_zap, dir, dataset;
1952 dsl_dataset_phys_t *ds;
1955 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1956 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1959 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1960 dir_obj = ds->ds_dir_obj;
1962 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) {
1963 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
1966 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1968 child_dir_zapobj = dd->dd_child_dir_zapobj;
1969 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) {
1970 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
1974 return (zap_list(spa, &child_dir_zap) != 0);
1978 zfs_callback_dataset(const spa_t *spa, uint64_t objnum, int (*callback)(const char *, uint64_t))
1980 uint64_t dir_obj, child_dir_zapobj, zap_type;
1981 dnode_phys_t child_dir_zap, dir, dataset;
1982 dsl_dataset_phys_t *ds;
1986 err = objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset);
1988 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1991 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1992 dir_obj = ds->ds_dir_obj;
1994 err = objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir);
1996 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
1999 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
2001 child_dir_zapobj = dd->dd_child_dir_zapobj;
2002 err = objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap);
2004 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
2008 err = dnode_read(spa, &child_dir_zap, 0, zap_scratch, child_dir_zap.dn_datablkszsec * 512);
2012 zap_type = *(uint64_t *) zap_scratch;
2013 if (zap_type == ZBT_MICRO)
2014 return mzap_list(&child_dir_zap, callback);
2016 return fzap_list(spa, &child_dir_zap, callback);
2021 * Find the object set given the object number of its dataset object
2022 * and return its details in *objset
2025 zfs_mount_dataset(const spa_t *spa, uint64_t objnum, objset_phys_t *objset)
2027 dnode_phys_t dataset;
2028 dsl_dataset_phys_t *ds;
2030 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
2031 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
2035 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
2036 if (zio_read(spa, &ds->ds_bp, objset)) {
2037 printf("ZFS: can't read object set for dataset %ju\n",
2046 * Find the object set pointed to by the BOOTFS property or the root
2047 * dataset if there is none and return its details in *objset
2050 zfs_get_root(const spa_t *spa, uint64_t *objid)
2052 dnode_phys_t dir, propdir;
2053 uint64_t props, bootfs, root;
2058 * Start with the MOS directory object.
2060 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) {
2061 printf("ZFS: can't read MOS object directory\n");
2066 * Lookup the pool_props and see if we can find a bootfs.
2068 if (zap_lookup(spa, &dir, DMU_POOL_PROPS, sizeof (props), 1, &props) == 0
2069 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0
2070 && zap_lookup(spa, &propdir, "bootfs", sizeof (bootfs), 1, &bootfs) == 0
2077 * Lookup the root dataset directory
2079 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof (root), 1, &root)
2080 || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) {
2081 printf("ZFS: can't find root dsl_dir\n");
2086 * Use the information from the dataset directory's bonus buffer
2087 * to find the dataset object and from that the object set itself.
2089 dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus;
2090 *objid = dd->dd_head_dataset_obj;
2095 zfs_mount(const spa_t *spa, uint64_t rootobj, struct zfsmount *mount)
2101 * Find the root object set if not explicitly provided
2103 if (rootobj == 0 && zfs_get_root(spa, &rootobj)) {
2104 printf("ZFS: can't find root filesystem\n");
2108 if (zfs_mount_dataset(spa, rootobj, &mount->objset)) {
2109 printf("ZFS: can't open root filesystem\n");
2113 mount->rootobj = rootobj;
2119 * callback function for feature name checks.
2122 check_feature(const char *name, uint64_t value)
2128 if (name[0] == '\0')
2131 for (i = 0; features_for_read[i] != NULL; i++) {
2132 if (strcmp(name, features_for_read[i]) == 0)
2135 printf("ZFS: unsupported feature: %s\n", name);
2140 * Checks whether the MOS features that are active are supported.
2143 check_mos_features(const spa_t *spa)
2146 uint64_t objnum, zap_type;
2150 if ((rc = objset_get_dnode(spa, &spa->spa_mos, DMU_OT_OBJECT_DIRECTORY,
2153 if ((rc = zap_lookup(spa, &dir, DMU_POOL_FEATURES_FOR_READ,
2154 sizeof (objnum), 1, &objnum)) != 0) {
2156 * It is older pool without features. As we have already
2157 * tested the label, just return without raising the error.
2162 if ((rc = objset_get_dnode(spa, &spa->spa_mos, objnum, &dir)) != 0)
2165 if (dir.dn_type != DMU_OTN_ZAP_METADATA)
2168 size = dir.dn_datablkszsec * 512;
2169 if (dnode_read(spa, &dir, 0, zap_scratch, size))
2172 zap_type = *(uint64_t *) zap_scratch;
2173 if (zap_type == ZBT_MICRO)
2174 rc = mzap_list(&dir, check_feature);
2176 rc = fzap_list(spa, &dir, check_feature);
2182 zfs_spa_init(spa_t *spa)
2187 if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) {
2188 printf("ZFS: can't read MOS of pool %s\n", spa->spa_name);
2191 if (spa->spa_mos.os_type != DMU_OST_META) {
2192 printf("ZFS: corrupted MOS of pool %s\n", spa->spa_name);
2196 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT,
2198 printf("ZFS: failed to read pool %s directory object\n",
2202 /* this is allowed to fail, older pools do not have salt */
2203 rc = zap_lookup(spa, &dir, DMU_POOL_CHECKSUM_SALT, 1,
2204 sizeof (spa->spa_cksum_salt.zcs_bytes),
2205 spa->spa_cksum_salt.zcs_bytes);
2207 rc = check_mos_features(spa);
2209 printf("ZFS: pool %s is not supported\n", spa->spa_name);
2216 zfs_dnode_stat(const spa_t *spa, dnode_phys_t *dn, struct stat *sb)
2219 if (dn->dn_bonustype != DMU_OT_SA) {
2220 znode_phys_t *zp = (znode_phys_t *)dn->dn_bonus;
2222 sb->st_mode = zp->zp_mode;
2223 sb->st_uid = zp->zp_uid;
2224 sb->st_gid = zp->zp_gid;
2225 sb->st_size = zp->zp_size;
2227 sa_hdr_phys_t *sahdrp;
2232 if (dn->dn_bonuslen != 0)
2233 sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn);
2235 if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) != 0) {
2236 blkptr_t *bp = &dn->dn_spill;
2239 size = BP_GET_LSIZE(bp);
2240 buf = zfs_alloc(size);
2241 error = zio_read(spa, bp, buf);
2243 zfs_free(buf, size);
2251 hdrsize = SA_HDR_SIZE(sahdrp);
2252 sb->st_mode = *(uint64_t *)((char *)sahdrp + hdrsize +
2254 sb->st_uid = *(uint64_t *)((char *)sahdrp + hdrsize +
2256 sb->st_gid = *(uint64_t *)((char *)sahdrp + hdrsize +
2258 sb->st_size = *(uint64_t *)((char *)sahdrp + hdrsize +
2261 zfs_free(buf, size);
2268 * Lookup a file and return its dnode.
2271 zfs_lookup(const struct zfsmount *mount, const char *upath, dnode_phys_t *dnode)
2274 uint64_t objnum, rootnum, parentnum;
2280 int symlinks_followed = 0;
2284 if (mount->objset.os_type != DMU_OST_ZFS) {
2285 printf("ZFS: unexpected object set type %ju\n",
2286 (uintmax_t)mount->objset.os_type);
2291 * Get the root directory dnode.
2293 rc = objset_get_dnode(spa, &mount->objset, MASTER_NODE_OBJ, &dn);
2297 rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, sizeof (rootnum), 1, &rootnum);
2301 rc = objset_get_dnode(spa, &mount->objset, rootnum, &dn);
2314 memcpy(element, p, q - p);
2322 rc = zfs_dnode_stat(spa, &dn, &sb);
2325 if (!S_ISDIR(sb.st_mode))
2329 rc = zap_lookup(spa, &dn, element, sizeof (objnum), 1, &objnum);
2332 objnum = ZFS_DIRENT_OBJ(objnum);
2334 rc = objset_get_dnode(spa, &mount->objset, objnum, &dn);
2339 * Check for symlink.
2341 rc = zfs_dnode_stat(spa, &dn, &sb);
2344 if (S_ISLNK(sb.st_mode)) {
2345 if (symlinks_followed > 10)
2347 symlinks_followed++;
2350 * Read the link value and copy the tail of our
2351 * current path onto the end.
2354 strcpy(&path[sb.st_size], p);
2356 path[sb.st_size] = 0;
2358 * Second test is purely to silence bogus compiler
2359 * warning about accessing past the end of dn_bonus.
2361 if (sb.st_size + sizeof(znode_phys_t) <=
2362 dn.dn_bonuslen && sizeof(znode_phys_t) <=
2363 sizeof(dn.dn_bonus)) {
2364 memcpy(path, &dn.dn_bonus[sizeof(znode_phys_t)],
2367 rc = dnode_read(spa, &dn, 0, path, sb.st_size);
2373 * Restart with the new path, starting either at
2374 * the root or at the parent depending whether or
2375 * not the link is relative.
2382 objset_get_dnode(spa, &mount->objset, objnum, &dn);