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.
38 * List of all vdevs, chained through v_alllink.
40 static vdev_list_t zfs_vdevs;
43 * List of all pools, chained through spa_link.
45 static spa_list_t zfs_pools;
47 static uint64_t zfs_crc64_table[256];
48 static const dnode_phys_t *dnode_cache_obj = 0;
49 static uint64_t dnode_cache_bn;
50 static char *dnode_cache_buf;
51 static char *zap_scratch;
52 static char *zfs_temp_buf, *zfs_temp_end, *zfs_temp_ptr;
54 #define TEMP_SIZE (1024 * 1024)
56 static int zio_read(spa_t *spa, const blkptr_t *bp, void *buf);
61 STAILQ_INIT(&zfs_vdevs);
62 STAILQ_INIT(&zfs_pools);
64 zfs_temp_buf = malloc(TEMP_SIZE);
65 zfs_temp_end = zfs_temp_buf + TEMP_SIZE;
66 zfs_temp_ptr = zfs_temp_buf;
67 dnode_cache_buf = malloc(SPA_MAXBLOCKSIZE);
68 zap_scratch = malloc(SPA_MAXBLOCKSIZE);
74 zfs_alloc_temp(size_t sz)
78 if (zfs_temp_ptr + sz > zfs_temp_end) {
79 printf("ZFS: out of temporary buffer space\n");
92 zfs_temp_ptr = zfs_temp_buf;
96 xdr_int(const unsigned char **xdr, int *ip)
98 *ip = ((*xdr)[0] << 24)
107 xdr_u_int(const unsigned char **xdr, u_int *ip)
109 *ip = ((*xdr)[0] << 24)
118 xdr_uint64_t(const unsigned char **xdr, uint64_t *lp)
124 *lp = (((uint64_t) hi) << 32) | lo;
129 nvlist_find(const unsigned char *nvlist, const char *name, int type,
130 int* elementsp, void *valuep)
132 const unsigned char *p, *pair;
134 int encoded_size, decoded_size;
141 xdr_int(&p, &encoded_size);
142 xdr_int(&p, &decoded_size);
143 while (encoded_size && decoded_size) {
144 int namelen, pairtype, elements;
145 const char *pairname;
147 xdr_int(&p, &namelen);
148 pairname = (const char*) p;
149 p += roundup(namelen, 4);
150 xdr_int(&p, &pairtype);
152 if (!memcmp(name, pairname, namelen) && type == pairtype) {
153 xdr_int(&p, &elements);
155 *elementsp = elements;
156 if (type == DATA_TYPE_UINT64) {
157 xdr_uint64_t(&p, (uint64_t *) valuep);
159 } else if (type == DATA_TYPE_STRING) {
162 (*(const char**) valuep) = (const char*) p;
164 } else if (type == DATA_TYPE_NVLIST
165 || type == DATA_TYPE_NVLIST_ARRAY) {
166 (*(const unsigned char**) valuep) =
167 (const unsigned char*) p;
174 * Not the pair we are looking for, skip to the next one.
176 p = pair + encoded_size;
180 xdr_int(&p, &encoded_size);
181 xdr_int(&p, &decoded_size);
188 * Return the next nvlist in an nvlist array.
190 static const unsigned char *
191 nvlist_next(const unsigned char *nvlist)
193 const unsigned char *p, *pair;
195 int encoded_size, decoded_size;
202 xdr_int(&p, &encoded_size);
203 xdr_int(&p, &decoded_size);
204 while (encoded_size && decoded_size) {
205 p = pair + encoded_size;
208 xdr_int(&p, &encoded_size);
209 xdr_int(&p, &decoded_size);
217 static const unsigned char *
218 nvlist_print(const unsigned char *nvlist, unsigned int indent)
220 static const char* typenames[] = {
231 "DATA_TYPE_BYTE_ARRAY",
232 "DATA_TYPE_INT16_ARRAY",
233 "DATA_TYPE_UINT16_ARRAY",
234 "DATA_TYPE_INT32_ARRAY",
235 "DATA_TYPE_UINT32_ARRAY",
236 "DATA_TYPE_INT64_ARRAY",
237 "DATA_TYPE_UINT64_ARRAY",
238 "DATA_TYPE_STRING_ARRAY",
241 "DATA_TYPE_NVLIST_ARRAY",
242 "DATA_TYPE_BOOLEAN_VALUE",
245 "DATA_TYPE_BOOLEAN_ARRAY",
246 "DATA_TYPE_INT8_ARRAY",
247 "DATA_TYPE_UINT8_ARRAY"
251 const unsigned char *p, *pair;
253 int encoded_size, decoded_size;
260 xdr_int(&p, &encoded_size);
261 xdr_int(&p, &decoded_size);
262 while (encoded_size && decoded_size) {
263 int namelen, pairtype, elements;
264 const char *pairname;
266 xdr_int(&p, &namelen);
267 pairname = (const char*) p;
268 p += roundup(namelen, 4);
269 xdr_int(&p, &pairtype);
271 for (i = 0; i < indent; i++)
273 printf("%s %s", typenames[pairtype], pairname);
275 xdr_int(&p, &elements);
277 case DATA_TYPE_UINT64: {
279 xdr_uint64_t(&p, &val);
280 printf(" = 0x%llx\n", val);
284 case DATA_TYPE_STRING: {
287 printf(" = \"%s\"\n", p);
291 case DATA_TYPE_NVLIST:
293 nvlist_print(p, indent + 1);
296 case DATA_TYPE_NVLIST_ARRAY:
297 for (j = 0; j < elements; j++) {
299 p = nvlist_print(p, indent + 1);
300 if (j != elements - 1) {
301 for (i = 0; i < indent; i++)
303 printf("%s %s", typenames[pairtype], pairname);
312 p = pair + encoded_size;
315 xdr_int(&p, &encoded_size);
316 xdr_int(&p, &decoded_size);
325 vdev_read_phys(vdev_t *vdev, const blkptr_t *bp, void *buf,
326 off_t offset, size_t size)
331 if (!vdev->v_phys_read)
335 psize = BP_GET_PSIZE(bp);
340 /*printf("ZFS: reading %d bytes at 0x%llx to %p\n", psize, offset, buf);*/
341 rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize);
344 if (bp && zio_checksum_error(bp, buf))
351 vdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
352 off_t offset, size_t bytes)
355 return (vdev_read_phys(vdev, bp, buf,
356 offset + VDEV_LABEL_START_SIZE, bytes));
361 vdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
362 off_t offset, size_t bytes)
368 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
369 if (kid->v_state != VDEV_STATE_HEALTHY)
371 rc = kid->v_read(kid, bp, buf, offset, bytes);
380 vdev_find(uint64_t guid)
384 STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink)
385 if (vdev->v_guid == guid)
392 vdev_create(uint64_t guid, vdev_read_t *read)
396 vdev = malloc(sizeof(vdev_t));
397 memset(vdev, 0, sizeof(vdev_t));
398 STAILQ_INIT(&vdev->v_children);
400 vdev->v_state = VDEV_STATE_OFFLINE;
402 vdev->v_phys_read = 0;
403 vdev->v_read_priv = 0;
404 STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink);
410 vdev_init_from_nvlist(const unsigned char *nvlist, vdev_t **vdevp, int is_newer)
413 uint64_t guid, id, ashift, nparity;
417 const unsigned char *kids;
418 int nkids, i, is_new;
419 uint64_t is_offline, is_faulted, is_degraded, is_removed;
421 if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID,
422 DATA_TYPE_UINT64, 0, &guid)
423 || nvlist_find(nvlist, ZPOOL_CONFIG_ID,
424 DATA_TYPE_UINT64, 0, &id)
425 || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE,
426 DATA_TYPE_STRING, 0, &type)) {
427 printf("ZFS: can't find vdev details\n");
431 if (strcmp(type, VDEV_TYPE_MIRROR)
432 && strcmp(type, VDEV_TYPE_DISK)
433 && strcmp(type, VDEV_TYPE_RAIDZ)) {
434 printf("ZFS: can only boot from disk, mirror or raidz vdevs\n");
438 is_offline = is_removed = is_faulted = is_degraded = 0;
440 nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, 0,
442 nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, 0,
444 nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, 0,
446 nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, 0,
449 vdev = vdev_find(guid);
453 if (!strcmp(type, VDEV_TYPE_MIRROR))
454 vdev = vdev_create(guid, vdev_mirror_read);
455 else if (!strcmp(type, VDEV_TYPE_RAIDZ))
456 vdev = vdev_create(guid, vdev_raidz_read);
458 vdev = vdev_create(guid, vdev_disk_read);
461 if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT,
462 DATA_TYPE_UINT64, 0, &ashift) == 0)
463 vdev->v_ashift = ashift;
466 if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY,
467 DATA_TYPE_UINT64, 0, &nparity) == 0)
468 vdev->v_nparity = nparity;
471 if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH,
472 DATA_TYPE_STRING, 0, &path) == 0) {
480 vdev->v_name = strdup(path);
482 if (!strcmp(type, "raidz")) {
483 if (vdev->v_nparity == 1)
484 vdev->v_name = "raidz1";
486 vdev->v_name = "raidz2";
488 vdev->v_name = strdup(type);
493 vdev->v_state = VDEV_STATE_OFFLINE;
495 vdev->v_state = VDEV_STATE_REMOVED;
497 vdev->v_state = VDEV_STATE_FAULTED;
498 else if (is_degraded)
499 vdev->v_state = VDEV_STATE_DEGRADED;
501 vdev->v_state = VDEV_STATE_HEALTHY;
507 * We've already seen this vdev, but from an older
508 * vdev label, so let's refresh its state from the
512 vdev->v_state = VDEV_STATE_OFFLINE;
514 vdev->v_state = VDEV_STATE_REMOVED;
516 vdev->v_state = VDEV_STATE_FAULTED;
517 else if (is_degraded)
518 vdev->v_state = VDEV_STATE_DEGRADED;
520 vdev->v_state = VDEV_STATE_HEALTHY;
524 rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN,
525 DATA_TYPE_NVLIST_ARRAY, &nkids, &kids);
527 * Its ok if we don't have any kids.
530 vdev->v_nchildren = nkids;
531 for (i = 0; i < nkids; i++) {
532 rc = vdev_init_from_nvlist(kids, &kid, is_newer);
536 STAILQ_INSERT_TAIL(&vdev->v_children, kid,
538 kids = nvlist_next(kids);
541 vdev->v_nchildren = 0;
550 vdev_set_state(vdev_t *vdev)
557 * A mirror or raidz is healthy if all its kids are healthy. A
558 * mirror is degraded if any of its kids is healthy; a raidz
559 * is degraded if at most nparity kids are offline.
561 if (STAILQ_FIRST(&vdev->v_children)) {
564 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
565 if (kid->v_state == VDEV_STATE_HEALTHY)
571 vdev->v_state = VDEV_STATE_HEALTHY;
573 if (vdev->v_read == vdev_mirror_read) {
575 vdev->v_state = VDEV_STATE_DEGRADED;
577 vdev->v_state = VDEV_STATE_OFFLINE;
579 } else if (vdev->v_read == vdev_raidz_read) {
580 if (bad_kids > vdev->v_nparity) {
581 vdev->v_state = VDEV_STATE_OFFLINE;
583 vdev->v_state = VDEV_STATE_DEGRADED;
591 spa_find_by_guid(uint64_t guid)
595 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
596 if (spa->spa_guid == guid)
605 spa_find_by_name(const char *name)
609 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
610 if (!strcmp(spa->spa_name, name))
619 spa_create(uint64_t guid)
623 spa = malloc(sizeof(spa_t));
624 memset(spa, 0, sizeof(spa_t));
625 STAILQ_INIT(&spa->spa_vdevs);
626 spa->spa_guid = guid;
627 STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link);
633 state_name(vdev_state_t state)
635 static const char* names[] = {
650 #define pager_printf printf
655 pager_printf(const char *fmt, ...)
661 vsprintf(line, fmt, args);
668 #define STATUS_FORMAT " %-16s %-10s\n"
671 print_state(int indent, const char *name, vdev_state_t state)
677 for (i = 0; i < indent; i++)
680 pager_printf(STATUS_FORMAT, buf, state_name(state));
685 vdev_status(vdev_t *vdev, int indent)
688 print_state(indent, vdev->v_name, vdev->v_state);
690 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
691 vdev_status(kid, indent + 1);
696 spa_status(spa_t *spa)
699 int good_kids, bad_kids, degraded_kids;
702 pager_printf(" pool: %s\n", spa->spa_name);
703 pager_printf("config:\n\n");
704 pager_printf(STATUS_FORMAT, "NAME", "STATE");
709 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
710 if (vdev->v_state == VDEV_STATE_HEALTHY)
712 else if (vdev->v_state == VDEV_STATE_DEGRADED)
718 state = VDEV_STATE_CLOSED;
719 if (good_kids > 0 && (degraded_kids + bad_kids) == 0)
720 state = VDEV_STATE_HEALTHY;
721 else if ((good_kids + degraded_kids) > 0)
722 state = VDEV_STATE_DEGRADED;
724 print_state(0, spa->spa_name, state);
725 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
726 vdev_status(vdev, 1);
736 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
745 vdev_probe(vdev_phys_read_t *read, void *read_priv, spa_t **spap)
748 vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch;
750 vdev_t *vdev, *top_vdev, *pool_vdev;
753 const unsigned char *nvlist;
756 uint64_t pool_txg, pool_guid;
757 const char *pool_name;
758 const unsigned char *vdevs;
761 const struct uberblock *up;
764 * Load the vdev label and figure out which
765 * uberblock is most current.
767 memset(&vtmp, 0, sizeof(vtmp));
768 vtmp.v_phys_read = read;
769 vtmp.v_read_priv = read_priv;
770 off = offsetof(vdev_label_t, vl_vdev_phys);
772 BP_SET_LSIZE(&bp, sizeof(vdev_phys_t));
773 BP_SET_PSIZE(&bp, sizeof(vdev_phys_t));
774 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
775 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
776 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
777 if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0))
780 if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) {
784 nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4;
786 if (nvlist_find(nvlist,
787 ZPOOL_CONFIG_VERSION,
788 DATA_TYPE_UINT64, 0, &val)) {
792 if (val > SPA_VERSION) {
793 printf("ZFS: unsupported ZFS version %u (should be %u)\n",
794 (unsigned) val, (unsigned) SPA_VERSION);
798 if (nvlist_find(nvlist,
799 ZPOOL_CONFIG_POOL_STATE,
800 DATA_TYPE_UINT64, 0, &val)) {
805 if (val != POOL_STATE_ACTIVE) {
807 * Don't print a message here. If we happen to reboot
808 * while where is an exported pool around, we don't
809 * need a cascade of confusing messages during boot.
811 /*printf("ZFS: pool is not active\n");*/
816 if (nvlist_find(nvlist,
817 ZPOOL_CONFIG_POOL_TXG,
818 DATA_TYPE_UINT64, 0, &pool_txg)
819 || nvlist_find(nvlist,
820 ZPOOL_CONFIG_POOL_GUID,
821 DATA_TYPE_UINT64, 0, &pool_guid)
822 || nvlist_find(nvlist,
823 ZPOOL_CONFIG_POOL_NAME,
824 DATA_TYPE_STRING, 0, &pool_name)) {
826 * Cache and spare devices end up here - just ignore
829 /*printf("ZFS: can't find pool details\n");*/
834 * Create the pool if this is the first time we've seen it.
836 spa = spa_find_by_guid(pool_guid);
838 spa = spa_create(pool_guid);
839 spa->spa_name = strdup(pool_name);
841 if (pool_txg > spa->spa_txg) {
842 spa->spa_txg = pool_txg;
848 * Get the vdev tree and create our in-core copy of it.
849 * If we already have a vdev with this guid, this must
850 * be some kind of alias (overlapping slices, dangerously dedicated
853 if (nvlist_find(nvlist,
855 DATA_TYPE_UINT64, 0, &guid)) {
858 vdev = vdev_find(guid);
859 if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */
862 if (nvlist_find(nvlist,
863 ZPOOL_CONFIG_VDEV_TREE,
864 DATA_TYPE_NVLIST, 0, &vdevs)) {
868 rc = vdev_init_from_nvlist(vdevs, &top_vdev, is_newer);
873 * Add the toplevel vdev to the pool if its not already there.
875 STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink)
876 if (top_vdev == pool_vdev)
878 if (!pool_vdev && top_vdev)
879 STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink);
882 * We should already have created an incomplete vdev for this
883 * vdev. Find it and initialise it with our read proc.
885 vdev = vdev_find(guid);
887 vdev->v_phys_read = read;
888 vdev->v_read_priv = read_priv;
890 printf("ZFS: inconsistent nvlist contents\n");
895 * Re-evaluate top-level vdev state.
897 vdev_set_state(top_vdev);
900 * Ok, we are happy with the pool so far. Lets find
901 * the best uberblock and then we can actually access
902 * the contents of the pool.
905 i < VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT;
907 off = offsetof(vdev_label_t, vl_uberblock);
908 off += i << UBERBLOCK_SHIFT;
910 DVA_SET_OFFSET(&bp.blk_dva[0], off);
911 BP_SET_LSIZE(&bp, 1 << UBERBLOCK_SHIFT);
912 BP_SET_PSIZE(&bp, 1 << UBERBLOCK_SHIFT);
913 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
914 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
915 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
916 if (vdev_read_phys(vdev, &bp, upbuf, off, 0))
919 up = (const struct uberblock *) upbuf;
920 if (up->ub_magic != UBERBLOCK_MAGIC)
922 if (up->ub_txg < spa->spa_txg)
924 if (up->ub_txg > spa->spa_uberblock.ub_txg) {
925 spa->spa_uberblock = *up;
926 } else if (up->ub_txg == spa->spa_uberblock.ub_txg) {
927 if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp)
928 spa->spa_uberblock = *up;
942 for (v = 0; v < 32; v++)
949 zio_read_gang(spa_t *spa, const blkptr_t *bp, const dva_t *dva, void *buf)
951 zio_gbh_phys_t zio_gb;
957 vdevid = DVA_GET_VDEV(dva);
958 offset = DVA_GET_OFFSET(dva);
959 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink)
960 if (vdev->v_id == vdevid)
962 if (!vdev || !vdev->v_read)
964 if (vdev->v_read(vdev, NULL, &zio_gb, offset, SPA_GANGBLOCKSIZE))
967 for (i = 0; i < SPA_GBH_NBLKPTRS; i++) {
968 blkptr_t *gbp = &zio_gb.zg_blkptr[i];
972 if (zio_read(spa, gbp, buf))
974 buf = (char*)buf + BP_GET_PSIZE(gbp);
981 zio_read(spa_t *spa, const blkptr_t *bp, void *buf)
983 int cpfunc = BP_GET_COMPRESS(bp);
984 size_t lsize = BP_GET_LSIZE(bp);
985 size_t psize = BP_GET_PSIZE(bp);
990 if (cpfunc != ZIO_COMPRESS_OFF)
991 pbuf = zfs_alloc_temp(psize);
995 for (i = 0; i < SPA_DVAS_PER_BP; i++) {
996 const dva_t *dva = &bp->blk_dva[i];
1001 if (!dva->dva_word[0] && !dva->dva_word[1])
1004 if (DVA_GET_GANG(dva)) {
1005 if (zio_read_gang(spa, bp, dva, buf))
1008 vdevid = DVA_GET_VDEV(dva);
1009 offset = DVA_GET_OFFSET(dva);
1010 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink)
1011 if (vdev->v_id == vdevid)
1013 if (!vdev || !vdev->v_read) {
1016 if (vdev->v_read(vdev, bp, pbuf, offset, psize))
1019 if (cpfunc != ZIO_COMPRESS_OFF) {
1020 if (zio_decompress_data(cpfunc, pbuf, psize,
1028 printf("ZFS: i/o error - all block copies unavailable\n");
1034 dnode_read(spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen)
1036 int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
1037 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1038 int nlevels = dnode->dn_nlevels;
1042 * Note: bsize may not be a power of two here so we need to do an
1043 * actual divide rather than a bitshift.
1045 while (buflen > 0) {
1046 uint64_t bn = offset / bsize;
1047 int boff = offset % bsize;
1049 const blkptr_t *indbp;
1052 if (bn > dnode->dn_maxblkid)
1055 if (dnode == dnode_cache_obj && bn == dnode_cache_bn)
1058 indbp = dnode->dn_blkptr;
1059 for (i = 0; i < nlevels; i++) {
1061 * Copy the bp from the indirect array so that
1062 * we can re-use the scratch buffer for multi-level
1065 ibn = bn >> ((nlevels - i - 1) * ibshift);
1066 ibn &= ((1 << ibshift) - 1);
1068 rc = zio_read(spa, &bp, dnode_cache_buf);
1071 indbp = (const blkptr_t *) dnode_cache_buf;
1073 dnode_cache_obj = dnode;
1074 dnode_cache_bn = bn;
1078 * The buffer contains our data block. Copy what we
1079 * need from it and loop.
1082 if (i > buflen) i = buflen;
1083 memcpy(buf, &dnode_cache_buf[boff], i);
1084 buf = ((char*) buf) + i;
1093 * Lookup a value in a microzap directory. Assumes that the zap
1094 * scratch buffer contains the directory contents.
1097 mzap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1099 const mzap_phys_t *mz;
1100 const mzap_ent_phys_t *mze;
1105 * Microzap objects use exactly one block. Read the whole
1108 size = dnode->dn_datablkszsec * 512;
1110 mz = (const mzap_phys_t *) zap_scratch;
1111 chunks = size / MZAP_ENT_LEN - 1;
1113 for (i = 0; i < chunks; i++) {
1114 mze = &mz->mz_chunk[i];
1115 if (!strcmp(mze->mze_name, name)) {
1116 *value = mze->mze_value;
1125 * Compare a name with a zap leaf entry. Return non-zero if the name
1129 fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name)
1132 const zap_leaf_chunk_t *nc;
1135 namelen = zc->l_entry.le_name_length;
1137 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1139 while (namelen > 0) {
1142 if (len > ZAP_LEAF_ARRAY_BYTES)
1143 len = ZAP_LEAF_ARRAY_BYTES;
1144 if (memcmp(p, nc->l_array.la_array, len))
1148 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1155 * Extract a uint64_t value from a zap leaf entry.
1158 fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc)
1160 const zap_leaf_chunk_t *vc;
1165 vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk);
1166 for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) {
1167 value = (value << 8) | p[i];
1174 * Lookup a value in a fatzap directory. Assumes that the zap scratch
1175 * buffer contains the directory header.
1178 fzap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1180 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1181 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1187 if (zh.zap_magic != ZAP_MAGIC)
1190 z.zap_block_shift = ilog2(bsize);
1191 z.zap_phys = (zap_phys_t *) zap_scratch;
1194 * Figure out where the pointer table is and read it in if necessary.
1196 if (zh.zap_ptrtbl.zt_blk) {
1197 rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize,
1198 zap_scratch, bsize);
1201 ptrtbl = (uint64_t *) zap_scratch;
1203 ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0);
1206 hash = zap_hash(zh.zap_salt, name);
1209 zl.l_bs = z.zap_block_shift;
1211 off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs;
1212 zap_leaf_chunk_t *zc;
1214 rc = dnode_read(spa, dnode, off, zap_scratch, bsize);
1218 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1221 * Make sure this chunk matches our hash.
1223 if (zl.l_phys->l_hdr.lh_prefix_len > 0
1224 && zl.l_phys->l_hdr.lh_prefix
1225 != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len))
1229 * Hash within the chunk to find our entry.
1231 int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len);
1232 int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1);
1233 h = zl.l_phys->l_hash[h];
1236 zc = &ZAP_LEAF_CHUNK(&zl, h);
1237 while (zc->l_entry.le_hash != hash) {
1238 if (zc->l_entry.le_next == 0xffff) {
1242 zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next);
1244 if (fzap_name_equal(&zl, zc, name)) {
1245 *value = fzap_leaf_value(&zl, zc);
1253 * Lookup a name in a zap object and return its value as a uint64_t.
1256 zap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1260 size_t size = dnode->dn_datablkszsec * 512;
1262 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1266 zap_type = *(uint64_t *) zap_scratch;
1267 if (zap_type == ZBT_MICRO)
1268 return mzap_lookup(spa, dnode, name, value);
1270 return fzap_lookup(spa, dnode, name, value);
1276 * List a microzap directory. Assumes that the zap scratch buffer contains
1277 * the directory contents.
1280 mzap_list(spa_t *spa, const dnode_phys_t *dnode)
1282 const mzap_phys_t *mz;
1283 const mzap_ent_phys_t *mze;
1288 * Microzap objects use exactly one block. Read the whole
1291 size = dnode->dn_datablkszsec * 512;
1292 mz = (const mzap_phys_t *) zap_scratch;
1293 chunks = size / MZAP_ENT_LEN - 1;
1295 for (i = 0; i < chunks; i++) {
1296 mze = &mz->mz_chunk[i];
1297 if (mze->mze_name[0])
1298 //printf("%-32s 0x%llx\n", mze->mze_name, mze->mze_value);
1299 printf("%s\n", mze->mze_name);
1306 * List a fatzap directory. Assumes that the zap scratch buffer contains
1307 * the directory header.
1310 fzap_list(spa_t *spa, const dnode_phys_t *dnode)
1312 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1313 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1317 if (zh.zap_magic != ZAP_MAGIC)
1320 z.zap_block_shift = ilog2(bsize);
1321 z.zap_phys = (zap_phys_t *) zap_scratch;
1324 * This assumes that the leaf blocks start at block 1. The
1325 * documentation isn't exactly clear on this.
1328 zl.l_bs = z.zap_block_shift;
1329 for (i = 0; i < zh.zap_num_leafs; i++) {
1330 off_t off = (i + 1) << zl.l_bs;
1334 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1337 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1339 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1340 zap_leaf_chunk_t *zc, *nc;
1343 zc = &ZAP_LEAF_CHUNK(&zl, j);
1344 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1346 namelen = zc->l_entry.le_name_length;
1347 if (namelen > sizeof(name))
1348 namelen = sizeof(name);
1351 * Paste the name back together.
1353 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
1355 while (namelen > 0) {
1358 if (len > ZAP_LEAF_ARRAY_BYTES)
1359 len = ZAP_LEAF_ARRAY_BYTES;
1360 memcpy(p, nc->l_array.la_array, len);
1363 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
1367 * Assume the first eight bytes of the value are
1370 value = fzap_leaf_value(&zl, zc);
1372 printf("%-32s 0x%llx\n", name, value);
1380 * List a zap directory.
1383 zap_list(spa_t *spa, const dnode_phys_t *dnode)
1386 size_t size = dnode->dn_datablkszsec * 512;
1388 if (dnode_read(spa, dnode, 0, zap_scratch, size))
1391 zap_type = *(uint64_t *) zap_scratch;
1392 if (zap_type == ZBT_MICRO)
1393 return mzap_list(spa, dnode);
1395 return fzap_list(spa, dnode);
1401 objset_get_dnode(spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode)
1405 offset = objnum * sizeof(dnode_phys_t);
1406 return dnode_read(spa, &os->os_meta_dnode, offset,
1407 dnode, sizeof(dnode_phys_t));
1411 * Find the object set given the object number of its dataset object
1412 * and return its details in *objset
1415 zfs_mount_dataset(spa_t *spa, uint64_t objnum, objset_phys_t *objset)
1417 dnode_phys_t dataset;
1418 dsl_dataset_phys_t *ds;
1420 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1421 printf("ZFS: can't find dataset %llu\n", objnum);
1425 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1426 if (zio_read(spa, &ds->ds_bp, objset)) {
1427 printf("ZFS: can't read object set for dataset %llu\n", objnum);
1435 * Find the object set pointed to by the BOOTFS property or the root
1436 * dataset if there is none and return its details in *objset
1439 zfs_mount_root(spa_t *spa, objset_phys_t *objset)
1441 dnode_phys_t dir, propdir;
1442 uint64_t props, bootfs, root;
1445 * Start with the MOS directory object.
1447 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) {
1448 printf("ZFS: can't read MOS object directory\n");
1453 * Lookup the pool_props and see if we can find a bootfs.
1455 if (zap_lookup(spa, &dir, DMU_POOL_PROPS, &props) == 0
1456 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0
1457 && zap_lookup(spa, &propdir, "bootfs", &bootfs) == 0
1459 return zfs_mount_dataset(spa, bootfs, objset);
1462 * Lookup the root dataset directory
1464 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &root)
1465 || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) {
1466 printf("ZFS: can't find root dsl_dir\n");
1471 * Use the information from the dataset directory's bonus buffer
1472 * to find the dataset object and from that the object set itself.
1474 dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus;
1475 return zfs_mount_dataset(spa, dd->dd_head_dataset_obj, objset);
1479 zfs_mount_pool(spa_t *spa)
1482 * Find the MOS and work our way in from there.
1484 if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) {
1485 printf("ZFS: can't read MOS\n");
1490 * Find the root object set
1492 if (zfs_mount_root(spa, &spa->spa_root_objset)) {
1493 printf("Can't find root filesystem - giving up\n");
1501 * Lookup a file and return its dnode.
1504 zfs_lookup(spa_t *spa, const char *upath, dnode_phys_t *dnode)
1507 uint64_t objnum, rootnum, parentnum;
1509 const znode_phys_t *zp = (const znode_phys_t *) dn.dn_bonus;
1513 int symlinks_followed = 0;
1515 if (spa->spa_root_objset.os_type != DMU_OST_ZFS) {
1516 printf("ZFS: unexpected object set type %llu\n",
1517 spa->spa_root_objset.os_type);
1522 * Get the root directory dnode.
1524 rc = objset_get_dnode(spa, &spa->spa_root_objset, MASTER_NODE_OBJ, &dn);
1528 rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, &rootnum);
1532 rc = objset_get_dnode(spa, &spa->spa_root_objset, rootnum, &dn);
1545 memcpy(element, p, q - p);
1553 if ((zp->zp_mode >> 12) != 0x4) {
1558 rc = zap_lookup(spa, &dn, element, &objnum);
1561 objnum = ZFS_DIRENT_OBJ(objnum);
1563 rc = objset_get_dnode(spa, &spa->spa_root_objset, objnum, &dn);
1568 * Check for symlink.
1570 if ((zp->zp_mode >> 12) == 0xa) {
1571 if (symlinks_followed > 10)
1573 symlinks_followed++;
1576 * Read the link value and copy the tail of our
1577 * current path onto the end.
1580 strcpy(&path[zp->zp_size], p);
1582 path[zp->zp_size] = 0;
1583 if (zp->zp_size + sizeof(znode_phys_t) <= dn.dn_bonuslen) {
1584 memcpy(path, &dn.dn_bonus[sizeof(znode_phys_t)],
1587 rc = dnode_read(spa, &dn, 0, path, zp->zp_size);
1593 * Restart with the new path, starting either at
1594 * the root or at the parent depending whether or
1595 * not the link is relative.
1602 objset_get_dnode(spa, &spa->spa_root_objset, objnum, &dn);