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)
332 psize = BP_GET_PSIZE(bp);
337 /*printf("ZFS: reading %d bytes at 0x%llx to %p\n", psize, offset, buf);*/
338 rc = vdev->v_phys_read(vdev, vdev->v_read_priv, offset, buf, psize);
341 if (bp && zio_checksum_error(bp, buf))
348 vdev_disk_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
349 off_t offset, size_t bytes)
352 return (vdev_read_phys(vdev, bp, buf,
353 offset + VDEV_LABEL_START_SIZE, bytes));
358 vdev_mirror_read(vdev_t *vdev, const blkptr_t *bp, void *buf,
359 off_t offset, size_t bytes)
365 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
366 if (kid->v_state != VDEV_STATE_HEALTHY)
368 rc = kid->v_read(kid, bp, buf, offset, bytes);
377 vdev_find(uint64_t guid)
381 STAILQ_FOREACH(vdev, &zfs_vdevs, v_alllink)
382 if (vdev->v_guid == guid)
389 vdev_create(uint64_t guid, vdev_read_t *read)
393 vdev = malloc(sizeof(vdev_t));
394 memset(vdev, 0, sizeof(vdev_t));
395 STAILQ_INIT(&vdev->v_children);
397 vdev->v_state = VDEV_STATE_OFFLINE;
399 vdev->v_phys_read = 0;
400 vdev->v_read_priv = 0;
401 STAILQ_INSERT_TAIL(&zfs_vdevs, vdev, v_alllink);
407 vdev_init_from_nvlist(const unsigned char *nvlist, vdev_t **vdevp)
410 uint64_t guid, id, ashift, nparity;
414 const unsigned char *kids;
417 if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID,
418 DATA_TYPE_UINT64, 0, &guid)
419 || nvlist_find(nvlist, ZPOOL_CONFIG_ID,
420 DATA_TYPE_UINT64, 0, &id)
421 || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE,
422 DATA_TYPE_STRING, 0, &type)) {
423 printf("ZFS: can't find vdev details\n");
428 * Assume that if we've seen this vdev tree before, this one
431 vdev = vdev_find(guid);
438 if (strcmp(type, VDEV_TYPE_MIRROR)
439 && strcmp(type, VDEV_TYPE_DISK)
440 && strcmp(type, VDEV_TYPE_RAIDZ)) {
441 printf("ZFS: can only boot from disk, mirror or raidz vdevs\n");
445 if (!strcmp(type, VDEV_TYPE_MIRROR))
446 vdev = vdev_create(guid, vdev_mirror_read);
447 else if (!strcmp(type, VDEV_TYPE_RAIDZ))
448 vdev = vdev_create(guid, vdev_raidz_read);
450 vdev = vdev_create(guid, vdev_disk_read);
453 if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT,
454 DATA_TYPE_UINT64, 0, &ashift) == 0)
455 vdev->v_ashift = ashift;
458 if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY,
459 DATA_TYPE_UINT64, 0, &nparity) == 0)
460 vdev->v_nparity = nparity;
463 if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH,
464 DATA_TYPE_STRING, 0, &path) == 0) {
472 vdev->v_name = strdup(path);
474 if (!strcmp(type, "raidz")) {
475 if (vdev->v_nparity == 1)
476 vdev->v_name = "raidz1";
478 vdev->v_name = "raidz2";
480 vdev->v_name = strdup(type);
483 rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN,
484 DATA_TYPE_NVLIST_ARRAY, &nkids, &kids);
486 * Its ok if we don't have any kids.
489 vdev->v_nchildren = nkids;
490 for (i = 0; i < nkids; i++) {
491 rc = vdev_init_from_nvlist(kids, &kid);
494 STAILQ_INSERT_TAIL(&vdev->v_children, kid, v_childlink);
495 kids = nvlist_next(kids);
498 vdev->v_nchildren = 0;
507 vdev_set_state(vdev_t *vdev)
514 * A mirror or raidz is healthy if all its kids are healthy. A
515 * mirror is degraded if any of its kids is healthy; a raidz
516 * is degraded if at most nparity kids are offline.
518 if (STAILQ_FIRST(&vdev->v_children)) {
521 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
522 if (kid->v_state == VDEV_STATE_HEALTHY)
528 vdev->v_state = VDEV_STATE_HEALTHY;
530 if (vdev->v_read == vdev_mirror_read) {
532 vdev->v_state = VDEV_STATE_DEGRADED;
534 vdev->v_state = VDEV_STATE_OFFLINE;
536 } else if (vdev->v_read == vdev_raidz_read) {
537 if (bad_kids > vdev->v_nparity) {
538 vdev->v_state = VDEV_STATE_OFFLINE;
540 vdev->v_state = VDEV_STATE_DEGRADED;
548 spa_find_by_guid(uint64_t guid)
552 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
553 if (spa->spa_guid == guid)
562 spa_find_by_name(const char *name)
566 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
567 if (!strcmp(spa->spa_name, name))
576 spa_create(uint64_t guid)
580 spa = malloc(sizeof(spa_t));
581 memset(spa, 0, sizeof(spa_t));
582 STAILQ_INIT(&spa->spa_vdevs);
583 spa->spa_guid = guid;
584 STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link);
590 state_name(vdev_state_t state)
592 static const char* names[] = {
605 #define pager_printf printf
610 pager_printf(const char *fmt, ...)
616 vsprintf(line, fmt, args);
623 #define STATUS_FORMAT " %-16s %-10s\n"
626 print_state(int indent, const char *name, vdev_state_t state)
632 for (i = 0; i < indent; i++)
635 pager_printf(STATUS_FORMAT, buf, state_name(state));
640 vdev_status(vdev_t *vdev, int indent)
643 print_state(indent, vdev->v_name, vdev->v_state);
645 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
646 vdev_status(kid, indent + 1);
651 spa_status(spa_t *spa)
654 int good_kids, bad_kids, degraded_kids;
657 pager_printf(" pool: %s\n", spa->spa_name);
658 pager_printf("config:\n\n");
659 pager_printf(STATUS_FORMAT, "NAME", "STATE");
664 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
665 if (vdev->v_state == VDEV_STATE_HEALTHY)
667 else if (vdev->v_state == VDEV_STATE_DEGRADED)
673 state = VDEV_STATE_CLOSED;
674 if (good_kids > 0 && (degraded_kids + bad_kids) == 0)
675 state = VDEV_STATE_HEALTHY;
676 else if ((good_kids + degraded_kids) > 0)
677 state = VDEV_STATE_DEGRADED;
679 print_state(0, spa->spa_name, state);
680 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
681 vdev_status(vdev, 1);
691 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
700 vdev_probe(vdev_phys_read_t *read, void *read_priv, spa_t **spap)
703 vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch;
705 vdev_t *vdev, *top_vdev, *pool_vdev;
708 const unsigned char *nvlist;
711 uint64_t pool_txg, pool_guid;
712 const char *pool_name;
713 const unsigned char *vdevs;
716 const struct uberblock *up;
719 * Load the vdev label and figure out which
720 * uberblock is most current.
722 memset(&vtmp, 0, sizeof(vtmp));
723 vtmp.v_phys_read = read;
724 vtmp.v_read_priv = read_priv;
725 off = offsetof(vdev_label_t, vl_vdev_phys);
727 BP_SET_LSIZE(&bp, sizeof(vdev_phys_t));
728 BP_SET_PSIZE(&bp, sizeof(vdev_phys_t));
729 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
730 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
731 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
732 if (vdev_read_phys(&vtmp, &bp, vdev_label, off, 0))
735 if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR) {
739 nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4;
741 if (nvlist_find(nvlist,
742 ZPOOL_CONFIG_VERSION,
743 DATA_TYPE_UINT64, 0, &val)) {
747 if (val > SPA_VERSION) {
748 printf("ZFS: unsupported ZFS version %u (should be %u)\n",
749 (unsigned) val, (unsigned) SPA_VERSION);
753 if (nvlist_find(nvlist,
754 ZPOOL_CONFIG_POOL_STATE,
755 DATA_TYPE_UINT64, 0, &val)) {
760 if (val != POOL_STATE_ACTIVE) {
762 * Don't print a message here. If we happen to reboot
763 * while where is an exported pool around, we don't
764 * need a cascade of confusing messages during boot.
766 /*printf("ZFS: pool is not active\n");*/
771 if (nvlist_find(nvlist,
772 ZPOOL_CONFIG_POOL_TXG,
773 DATA_TYPE_UINT64, 0, &pool_txg)
774 || nvlist_find(nvlist,
775 ZPOOL_CONFIG_POOL_GUID,
776 DATA_TYPE_UINT64, 0, &pool_guid)
777 || nvlist_find(nvlist,
778 ZPOOL_CONFIG_POOL_NAME,
779 DATA_TYPE_STRING, 0, &pool_name)) {
781 * Cache and spare devices end up here - just ignore
784 /*printf("ZFS: can't find pool details\n");*/
789 * Create the pool if this is the first time we've seen it.
791 spa = spa_find_by_guid(pool_guid);
793 spa = spa_create(pool_guid);
794 spa->spa_name = strdup(pool_name);
796 if (pool_txg > spa->spa_txg)
797 spa->spa_txg = pool_txg;
800 * Get the vdev tree and create our in-core copy of it.
801 * If we already have a healthy vdev with this guid, this must
802 * be some kind of alias (overlapping slices, dangerously dedicated
805 if (nvlist_find(nvlist,
807 DATA_TYPE_UINT64, 0, &guid)) {
810 vdev = vdev_find(guid);
811 if (vdev && vdev->v_state == VDEV_STATE_HEALTHY) {
815 if (nvlist_find(nvlist,
816 ZPOOL_CONFIG_VDEV_TREE,
817 DATA_TYPE_NVLIST, 0, &vdevs)) {
820 rc = vdev_init_from_nvlist(vdevs, &top_vdev);
825 * Add the toplevel vdev to the pool if its not already there.
827 STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink)
828 if (top_vdev == pool_vdev)
830 if (!pool_vdev && top_vdev)
831 STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink);
834 * We should already have created an incomplete vdev for this
835 * vdev. Find it and initialise it with our read proc.
837 vdev = vdev_find(guid);
839 vdev->v_phys_read = read;
840 vdev->v_read_priv = read_priv;
841 vdev->v_state = VDEV_STATE_HEALTHY;
843 printf("ZFS: inconsistent nvlist contents\n");
848 * Re-evaluate top-level vdev state.
850 vdev_set_state(top_vdev);
853 * Ok, we are happy with the pool so far. Lets find
854 * the best uberblock and then we can actually access
855 * the contents of the pool.
858 i < VDEV_UBERBLOCK_RING >> UBERBLOCK_SHIFT;
860 off = offsetof(vdev_label_t, vl_uberblock);
861 off += i << UBERBLOCK_SHIFT;
863 DVA_SET_OFFSET(&bp.blk_dva[0], off);
864 BP_SET_LSIZE(&bp, 1 << UBERBLOCK_SHIFT);
865 BP_SET_PSIZE(&bp, 1 << UBERBLOCK_SHIFT);
866 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
867 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
868 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
869 if (vdev_read_phys(vdev, &bp, upbuf, off, 0))
872 up = (const struct uberblock *) upbuf;
873 if (up->ub_magic != UBERBLOCK_MAGIC)
875 if (up->ub_txg < spa->spa_txg)
877 if (up->ub_txg > spa->spa_uberblock.ub_txg) {
878 spa->spa_uberblock = *up;
879 } else if (up->ub_txg == spa->spa_uberblock.ub_txg) {
880 if (up->ub_timestamp > spa->spa_uberblock.ub_timestamp)
881 spa->spa_uberblock = *up;
895 for (v = 0; v < 32; v++)
902 zio_read_gang(spa_t *spa, const blkptr_t *bp, const dva_t *dva, void *buf)
904 zio_gbh_phys_t zio_gb;
910 vdevid = DVA_GET_VDEV(dva);
911 offset = DVA_GET_OFFSET(dva);
912 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink)
913 if (vdev->v_id == vdevid)
915 if (!vdev || !vdev->v_read)
917 if (vdev->v_read(vdev, bp, &zio_gb, offset, SPA_GANGBLOCKSIZE))
920 for (i = 0; i < SPA_GBH_NBLKPTRS; i++) {
921 if (zio_read(spa, &zio_gb.zg_blkptr[i], buf))
929 zio_read(spa_t *spa, const blkptr_t *bp, void *buf)
931 int cpfunc = BP_GET_COMPRESS(bp);
932 size_t lsize = BP_GET_LSIZE(bp);
933 size_t psize = BP_GET_PSIZE(bp);
938 if (cpfunc != ZIO_COMPRESS_OFF)
939 pbuf = zfs_alloc_temp(psize);
943 for (i = 0; i < SPA_DVAS_PER_BP; i++) {
944 const dva_t *dva = &bp->blk_dva[i];
949 if (!dva->dva_word[0] && !dva->dva_word[1])
952 if (DVA_GET_GANG(dva)) {
953 printf("ZFS: gang block detected!\n");
954 if (zio_read_gang(spa, bp, dva, buf))
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) {
965 if (vdev->v_read(vdev, bp, pbuf, offset, psize))
968 if (cpfunc != ZIO_COMPRESS_OFF) {
969 if (zio_decompress_data(cpfunc, pbuf, psize,
977 printf("ZFS: i/o error - all block copies unavailable\n");
983 dnode_read(spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen)
985 int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
986 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
987 int nlevels = dnode->dn_nlevels;
991 * Note: bsize may not be a power of two here so we need to do an
992 * actual divide rather than a bitshift.
995 uint64_t bn = offset / bsize;
996 int boff = offset % bsize;
998 const blkptr_t *indbp;
1001 if (bn > dnode->dn_maxblkid)
1004 if (dnode == dnode_cache_obj && bn == dnode_cache_bn)
1007 indbp = dnode->dn_blkptr;
1008 for (i = 0; i < nlevels; i++) {
1010 * Copy the bp from the indirect array so that
1011 * we can re-use the scratch buffer for multi-level
1014 ibn = bn >> ((nlevels - i - 1) * ibshift);
1015 ibn &= ((1 << ibshift) - 1);
1017 rc = zio_read(spa, &bp, dnode_cache_buf);
1020 indbp = (const blkptr_t *) dnode_cache_buf;
1022 dnode_cache_obj = dnode;
1023 dnode_cache_bn = bn;
1027 * The buffer contains our data block. Copy what we
1028 * need from it and loop.
1031 if (i > buflen) i = buflen;
1032 memcpy(buf, &dnode_cache_buf[boff], i);
1033 buf = ((char*) buf) + i;
1042 * Lookup a value in a microzap directory. Assumes that the zap
1043 * scratch buffer contains the directory contents.
1046 mzap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1048 const mzap_phys_t *mz;
1049 const mzap_ent_phys_t *mze;
1054 * Microzap objects use exactly one block. Read the whole
1057 size = dnode->dn_datablkszsec * 512;
1059 mz = (const mzap_phys_t *) zap_scratch;
1060 chunks = size / MZAP_ENT_LEN - 1;
1062 for (i = 0; i < chunks; i++) {
1063 mze = &mz->mz_chunk[i];
1064 if (!strcmp(mze->mze_name, name)) {
1065 *value = mze->mze_value;
1074 * Compare a name with a zap leaf entry. Return non-zero if the name
1078 fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name)
1081 const zap_leaf_chunk_t *nc;
1084 namelen = zc->l_entry.le_name_length;
1086 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1088 while (namelen > 0) {
1091 if (len > ZAP_LEAF_ARRAY_BYTES)
1092 len = ZAP_LEAF_ARRAY_BYTES;
1093 if (memcmp(p, nc->l_array.la_array, len))
1097 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1104 * Extract a uint64_t value from a zap leaf entry.
1107 fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc)
1109 const zap_leaf_chunk_t *vc;
1114 vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk);
1115 for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) {
1116 value = (value << 8) | p[i];
1123 * Lookup a value in a fatzap directory. Assumes that the zap scratch
1124 * buffer contains the directory header.
1127 fzap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1129 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1130 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1136 if (zh.zap_magic != ZAP_MAGIC)
1139 z.zap_block_shift = ilog2(bsize);
1140 z.zap_phys = (zap_phys_t *) zap_scratch;
1143 * Figure out where the pointer table is and read it in if necessary.
1145 if (zh.zap_ptrtbl.zt_blk) {
1146 rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize,
1147 zap_scratch, bsize);
1150 ptrtbl = (uint64_t *) zap_scratch;
1152 ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0);
1155 hash = zap_hash(zh.zap_salt, name);
1158 zl.l_bs = z.zap_block_shift;
1160 off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs;
1161 zap_leaf_chunk_t *zc;
1163 rc = dnode_read(spa, dnode, off, zap_scratch, bsize);
1167 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1170 * Make sure this chunk matches our hash.
1172 if (zl.l_phys->l_hdr.lh_prefix_len > 0
1173 && zl.l_phys->l_hdr.lh_prefix
1174 != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len))
1178 * Hash within the chunk to find our entry.
1180 int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len);
1181 int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1);
1182 h = zl.l_phys->l_hash[h];
1185 zc = &ZAP_LEAF_CHUNK(&zl, h);
1186 while (zc->l_entry.le_hash != hash) {
1187 if (zc->l_entry.le_next == 0xffff) {
1191 zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next);
1193 if (fzap_name_equal(&zl, zc, name)) {
1194 *value = fzap_leaf_value(&zl, zc);
1202 * Lookup a name in a zap object and return its value as a uint64_t.
1205 zap_lookup(spa_t *spa, const dnode_phys_t *dnode, const char *name, uint64_t *value)
1209 size_t size = dnode->dn_datablkszsec * 512;
1211 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1215 zap_type = *(uint64_t *) zap_scratch;
1216 if (zap_type == ZBT_MICRO)
1217 return mzap_lookup(spa, dnode, name, value);
1219 return fzap_lookup(spa, dnode, name, value);
1225 * List a microzap directory. Assumes that the zap scratch buffer contains
1226 * the directory contents.
1229 mzap_list(spa_t *spa, const dnode_phys_t *dnode)
1231 const mzap_phys_t *mz;
1232 const mzap_ent_phys_t *mze;
1237 * Microzap objects use exactly one block. Read the whole
1240 size = dnode->dn_datablkszsec * 512;
1241 mz = (const mzap_phys_t *) zap_scratch;
1242 chunks = size / MZAP_ENT_LEN - 1;
1244 for (i = 0; i < chunks; i++) {
1245 mze = &mz->mz_chunk[i];
1246 if (mze->mze_name[0])
1247 //printf("%-32s 0x%llx\n", mze->mze_name, mze->mze_value);
1248 printf("%s\n", mze->mze_name);
1255 * List a fatzap directory. Assumes that the zap scratch buffer contains
1256 * the directory header.
1259 fzap_list(spa_t *spa, const dnode_phys_t *dnode)
1261 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1262 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1266 if (zh.zap_magic != ZAP_MAGIC)
1269 z.zap_block_shift = ilog2(bsize);
1270 z.zap_phys = (zap_phys_t *) zap_scratch;
1273 * This assumes that the leaf blocks start at block 1. The
1274 * documentation isn't exactly clear on this.
1277 zl.l_bs = z.zap_block_shift;
1278 for (i = 0; i < zh.zap_num_leafs; i++) {
1279 off_t off = (i + 1) << zl.l_bs;
1283 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1286 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1288 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1289 zap_leaf_chunk_t *zc, *nc;
1292 zc = &ZAP_LEAF_CHUNK(&zl, j);
1293 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1295 namelen = zc->l_entry.le_name_length;
1296 if (namelen > sizeof(name))
1297 namelen = sizeof(name);
1300 * Paste the name back together.
1302 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
1304 while (namelen > 0) {
1307 if (len > ZAP_LEAF_ARRAY_BYTES)
1308 len = ZAP_LEAF_ARRAY_BYTES;
1309 memcpy(p, nc->l_array.la_array, len);
1312 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
1316 * Assume the first eight bytes of the value are
1319 value = fzap_leaf_value(&zl, zc);
1321 printf("%-32s 0x%llx\n", name, value);
1329 * List a zap directory.
1332 zap_list(spa_t *spa, const dnode_phys_t *dnode)
1335 size_t size = dnode->dn_datablkszsec * 512;
1337 if (dnode_read(spa, dnode, 0, zap_scratch, size))
1340 zap_type = *(uint64_t *) zap_scratch;
1341 if (zap_type == ZBT_MICRO)
1342 return mzap_list(spa, dnode);
1344 return fzap_list(spa, dnode);
1350 objset_get_dnode(spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode)
1354 offset = objnum * sizeof(dnode_phys_t);
1355 return dnode_read(spa, &os->os_meta_dnode, offset,
1356 dnode, sizeof(dnode_phys_t));
1360 * Find the object set given the object number of its dataset object
1361 * and return its details in *objset
1364 zfs_mount_dataset(spa_t *spa, uint64_t objnum, objset_phys_t *objset)
1366 dnode_phys_t dataset;
1367 dsl_dataset_phys_t *ds;
1369 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1370 printf("ZFS: can't find dataset %llu\n", objnum);
1374 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
1375 if (zio_read(spa, &ds->ds_bp, objset)) {
1376 printf("ZFS: can't read object set for dataset %llu\n", objnum);
1384 * Find the object set pointed to by the BOOTFS property or the root
1385 * dataset if there is none and return its details in *objset
1388 zfs_mount_root(spa_t *spa, objset_phys_t *objset)
1390 dnode_phys_t dir, propdir;
1391 uint64_t props, bootfs, root;
1394 * Start with the MOS directory object.
1396 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) {
1397 printf("ZFS: can't read MOS object directory\n");
1402 * Lookup the pool_props and see if we can find a bootfs.
1404 if (zap_lookup(spa, &dir, DMU_POOL_PROPS, &props) == 0
1405 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0
1406 && zap_lookup(spa, &propdir, "bootfs", &bootfs) == 0
1408 return zfs_mount_dataset(spa, bootfs, objset);
1411 * Lookup the root dataset directory
1413 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, &root)
1414 || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) {
1415 printf("ZFS: can't find root dsl_dir\n");
1420 * Use the information from the dataset directory's bonus buffer
1421 * to find the dataset object and from that the object set itself.
1423 dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus;
1424 return zfs_mount_dataset(spa, dd->dd_head_dataset_obj, objset);
1428 zfs_mount_pool(spa_t *spa)
1431 * Find the MOS and work our way in from there.
1433 if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) {
1434 printf("ZFS: can't read MOS\n");
1439 * Find the root object set
1441 if (zfs_mount_root(spa, &spa->spa_root_objset)) {
1442 printf("Can't find root filesystem - giving up\n");
1450 * Lookup a file and return its dnode.
1453 zfs_lookup(spa_t *spa, const char *upath, dnode_phys_t *dnode)
1456 uint64_t objnum, rootnum, parentnum;
1458 const znode_phys_t *zp = (const znode_phys_t *) dn.dn_bonus;
1462 int symlinks_followed = 0;
1464 if (spa->spa_root_objset.os_type != DMU_OST_ZFS) {
1465 printf("ZFS: unexpected object set type %llu\n",
1466 spa->spa_root_objset.os_type);
1471 * Get the root directory dnode.
1473 rc = objset_get_dnode(spa, &spa->spa_root_objset, MASTER_NODE_OBJ, &dn);
1477 rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, &rootnum);
1481 rc = objset_get_dnode(spa, &spa->spa_root_objset, rootnum, &dn);
1494 memcpy(element, p, q - p);
1502 if ((zp->zp_mode >> 12) != 0x4) {
1507 rc = zap_lookup(spa, &dn, element, &objnum);
1510 objnum = ZFS_DIRENT_OBJ(objnum);
1512 rc = objset_get_dnode(spa, &spa->spa_root_objset, objnum, &dn);
1517 * Check for symlink.
1519 if ((zp->zp_mode >> 12) == 0xa) {
1520 if (symlinks_followed > 10)
1522 symlinks_followed++;
1525 * Read the link value and copy the tail of our
1526 * current path onto the end.
1529 strcpy(&path[zp->zp_size], p);
1531 path[zp->zp_size] = 0;
1532 if (zp->zp_size + sizeof(znode_phys_t) <= dn.dn_bonuslen) {
1533 memcpy(path, &dn.dn_bonus[sizeof(znode_phys_t)],
1536 rc = dnode_read(spa, &dn, 0, path, zp->zp_size);
1542 * Restart with the new path, starting either at
1543 * the root or at the parent depending whether or
1544 * not the link is relative.
1551 objset_get_dnode(spa, &spa->spa_root_objset, objnum, &dn);