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",
63 "org.zfsonlinux:large_dnode",
68 * List of all pools, chained through spa_link.
70 static spa_list_t zfs_pools;
72 static const dnode_phys_t *dnode_cache_obj;
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 %zu 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;
506 vdev->v_read = _read;
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, DATA_TYPE_UINT64,
529 || nvlist_find(nvlist, ZPOOL_CONFIG_ID, DATA_TYPE_UINT64, NULL, &id)
530 || nvlist_find(nvlist, ZPOOL_CONFIG_TYPE, DATA_TYPE_STRING,
532 printf("ZFS: can't find vdev details\n");
536 if (strcmp(type, VDEV_TYPE_MIRROR)
537 && strcmp(type, VDEV_TYPE_DISK)
539 && strcmp(type, VDEV_TYPE_FILE)
541 && strcmp(type, VDEV_TYPE_RAIDZ)
542 && strcmp(type, VDEV_TYPE_REPLACING)) {
543 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
547 is_offline = is_removed = is_faulted = is_degraded = isnt_present = 0;
549 nvlist_find(nvlist, ZPOOL_CONFIG_OFFLINE, DATA_TYPE_UINT64, NULL,
551 nvlist_find(nvlist, ZPOOL_CONFIG_REMOVED, DATA_TYPE_UINT64, NULL,
553 nvlist_find(nvlist, ZPOOL_CONFIG_FAULTED, DATA_TYPE_UINT64, NULL,
555 nvlist_find(nvlist, ZPOOL_CONFIG_DEGRADED, DATA_TYPE_UINT64, NULL,
557 nvlist_find(nvlist, ZPOOL_CONFIG_NOT_PRESENT, DATA_TYPE_UINT64, NULL,
560 vdev = vdev_find(guid);
564 if (!strcmp(type, VDEV_TYPE_MIRROR))
565 vdev = vdev_create(guid, vdev_mirror_read);
566 else if (!strcmp(type, VDEV_TYPE_RAIDZ))
567 vdev = vdev_create(guid, vdev_raidz_read);
568 else if (!strcmp(type, VDEV_TYPE_REPLACING))
569 vdev = vdev_create(guid, vdev_replacing_read);
571 vdev = vdev_create(guid, vdev_disk_read);
574 vdev->v_top = pvdev != NULL ? pvdev : vdev;
575 if (nvlist_find(nvlist, ZPOOL_CONFIG_ASHIFT,
576 DATA_TYPE_UINT64, NULL, &ashift) == 0) {
577 vdev->v_ashift = ashift;
581 if (nvlist_find(nvlist, ZPOOL_CONFIG_NPARITY,
582 DATA_TYPE_UINT64, NULL, &nparity) == 0) {
583 vdev->v_nparity = nparity;
587 if (nvlist_find(nvlist, ZPOOL_CONFIG_PATH,
588 DATA_TYPE_STRING, NULL, &path) == 0) {
589 if (strncmp(path, "/dev/", 5) == 0)
591 vdev->v_name = strdup(path);
593 if (!strcmp(type, "raidz")) {
594 if (vdev->v_nparity == 1)
595 vdev->v_name = "raidz1";
596 else if (vdev->v_nparity == 2)
597 vdev->v_name = "raidz2";
598 else if (vdev->v_nparity == 3)
599 vdev->v_name = "raidz3";
601 printf("ZFS: can only boot from disk, mirror, raidz1, raidz2 and raidz3 vdevs\n");
605 vdev->v_name = strdup(type);
612 if (is_new || is_newer) {
614 * This is either new vdev or we've already seen this vdev,
615 * but from an older vdev label, so let's refresh its state
616 * from the newer label.
619 vdev->v_state = VDEV_STATE_OFFLINE;
621 vdev->v_state = VDEV_STATE_REMOVED;
623 vdev->v_state = VDEV_STATE_FAULTED;
624 else if (is_degraded)
625 vdev->v_state = VDEV_STATE_DEGRADED;
626 else if (isnt_present)
627 vdev->v_state = VDEV_STATE_CANT_OPEN;
630 rc = nvlist_find(nvlist, ZPOOL_CONFIG_CHILDREN, DATA_TYPE_NVLIST_ARRAY,
633 * Its ok if we don't have any kids.
636 vdev->v_nchildren = nkids;
637 for (i = 0; i < nkids; i++) {
638 rc = vdev_init_from_nvlist(kids, vdev, &kid, is_newer);
642 STAILQ_INSERT_TAIL(&vdev->v_children, kid,
644 kids = nvlist_next(kids);
647 vdev->v_nchildren = 0;
656 vdev_set_state(vdev_t *vdev)
663 * A mirror or raidz is healthy if all its kids are healthy. A
664 * mirror is degraded if any of its kids is healthy; a raidz
665 * is degraded if at most nparity kids are offline.
667 if (STAILQ_FIRST(&vdev->v_children)) {
670 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
671 if (kid->v_state == VDEV_STATE_HEALTHY)
677 vdev->v_state = VDEV_STATE_HEALTHY;
679 if (vdev->v_read == vdev_mirror_read) {
681 vdev->v_state = VDEV_STATE_DEGRADED;
683 vdev->v_state = VDEV_STATE_OFFLINE;
685 } else if (vdev->v_read == vdev_raidz_read) {
686 if (bad_kids > vdev->v_nparity) {
687 vdev->v_state = VDEV_STATE_OFFLINE;
689 vdev->v_state = VDEV_STATE_DEGRADED;
697 spa_find_by_guid(uint64_t guid)
701 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
702 if (spa->spa_guid == guid)
709 spa_find_by_name(const char *name)
713 STAILQ_FOREACH(spa, &zfs_pools, spa_link)
714 if (!strcmp(spa->spa_name, name))
722 spa_get_primary(void)
725 return (STAILQ_FIRST(&zfs_pools));
729 spa_get_primary_vdev(const spa_t *spa)
735 spa = spa_get_primary();
738 vdev = STAILQ_FIRST(&spa->spa_vdevs);
741 for (kid = STAILQ_FIRST(&vdev->v_children); kid != NULL;
742 kid = STAILQ_FIRST(&vdev->v_children))
749 spa_create(uint64_t guid, const char *name)
753 if ((spa = malloc(sizeof(spa_t))) == NULL)
755 memset(spa, 0, sizeof(spa_t));
756 if ((spa->spa_name = strdup(name)) == NULL) {
760 STAILQ_INIT(&spa->spa_vdevs);
761 spa->spa_guid = guid;
762 STAILQ_INSERT_TAIL(&zfs_pools, spa, spa_link);
768 state_name(vdev_state_t state)
770 static const char* names[] = {
785 #define pager_printf printf
790 pager_printf(const char *fmt, ...)
796 vsprintf(line, fmt, args);
799 return (pager_output(line));
804 #define STATUS_FORMAT " %s %s\n"
807 print_state(int indent, const char *name, vdev_state_t state)
813 for (i = 0; i < indent; i++)
817 return (pager_printf(STATUS_FORMAT, buf, state_name(state)));
821 vdev_status(vdev_t *vdev, int indent)
825 ret = print_state(indent, vdev->v_name, vdev->v_state);
829 STAILQ_FOREACH(kid, &vdev->v_children, v_childlink) {
830 ret = vdev_status(kid, indent + 1);
838 spa_status(spa_t *spa)
840 static char bootfs[ZFS_MAXNAMELEN];
843 int good_kids, bad_kids, degraded_kids, ret;
846 ret = pager_printf(" pool: %s\n", spa->spa_name);
850 if (zfs_get_root(spa, &rootid) == 0 &&
851 zfs_rlookup(spa, rootid, bootfs) == 0) {
852 if (bootfs[0] == '\0')
853 ret = pager_printf("bootfs: %s\n", spa->spa_name);
855 ret = pager_printf("bootfs: %s/%s\n", spa->spa_name,
860 ret = pager_printf("config:\n\n");
863 ret = pager_printf(STATUS_FORMAT, "NAME", "STATE");
870 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
871 if (vdev->v_state == VDEV_STATE_HEALTHY)
873 else if (vdev->v_state == VDEV_STATE_DEGRADED)
879 state = VDEV_STATE_CLOSED;
880 if (good_kids > 0 && (degraded_kids + bad_kids) == 0)
881 state = VDEV_STATE_HEALTHY;
882 else if ((good_kids + degraded_kids) > 0)
883 state = VDEV_STATE_DEGRADED;
885 ret = print_state(0, spa->spa_name, state);
888 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
889 ret = vdev_status(vdev, 1);
900 int first = 1, ret = 0;
902 STAILQ_FOREACH(spa, &zfs_pools, spa_link) {
904 ret = pager_printf("\n");
909 ret = spa_status(spa);
917 vdev_label_offset(uint64_t psize, int l, uint64_t offset)
919 uint64_t label_offset;
921 if (l < VDEV_LABELS / 2)
924 label_offset = psize - VDEV_LABELS * sizeof (vdev_label_t);
926 return (offset + l * sizeof (vdev_label_t) + label_offset);
930 vdev_probe(vdev_phys_read_t *_read, void *read_priv, spa_t **spap)
933 vdev_phys_t *vdev_label = (vdev_phys_t *) zap_scratch;
934 vdev_phys_t *tmp_label;
936 vdev_t *vdev, *top_vdev, *pool_vdev;
939 const unsigned char *nvlist = NULL;
942 uint64_t best_txg = 0;
943 uint64_t pool_txg, pool_guid;
945 const char *pool_name;
946 const unsigned char *vdevs;
947 const unsigned char *features;
948 int i, l, rc, is_newer;
950 const struct uberblock *up;
953 * Load the vdev label and figure out which
954 * uberblock is most current.
956 memset(&vtmp, 0, sizeof(vtmp));
957 vtmp.v_phys_read = _read;
958 vtmp.v_read_priv = read_priv;
959 psize = P2ALIGN(ldi_get_size(read_priv),
960 (uint64_t)sizeof (vdev_label_t));
962 /* Test for minimum pool size. */
963 if (psize < SPA_MINDEVSIZE)
966 tmp_label = zfs_alloc(sizeof(vdev_phys_t));
968 for (l = 0; l < VDEV_LABELS; l++) {
969 off = vdev_label_offset(psize, l,
970 offsetof(vdev_label_t, vl_vdev_phys));
973 BP_SET_LSIZE(&bp, sizeof(vdev_phys_t));
974 BP_SET_PSIZE(&bp, sizeof(vdev_phys_t));
975 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
976 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
977 DVA_SET_OFFSET(BP_IDENTITY(&bp), off);
978 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
980 if (vdev_read_phys(&vtmp, &bp, tmp_label, off, 0))
983 if (tmp_label->vp_nvlist[0] != NV_ENCODE_XDR)
986 nvlist = (const unsigned char *) tmp_label->vp_nvlist + 4;
987 if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG,
988 DATA_TYPE_UINT64, NULL, &pool_txg) != 0)
991 if (best_txg <= pool_txg) {
993 memcpy(vdev_label, tmp_label, sizeof (vdev_phys_t));
997 zfs_free(tmp_label, sizeof (vdev_phys_t));
1002 if (vdev_label->vp_nvlist[0] != NV_ENCODE_XDR)
1005 nvlist = (const unsigned char *) vdev_label->vp_nvlist + 4;
1007 if (nvlist_find(nvlist, ZPOOL_CONFIG_VERSION, DATA_TYPE_UINT64,
1012 if (!SPA_VERSION_IS_SUPPORTED(val)) {
1013 printf("ZFS: unsupported ZFS version %u (should be %u)\n",
1014 (unsigned) val, (unsigned) SPA_VERSION);
1018 /* Check ZFS features for read */
1019 if (nvlist_find(nvlist, ZPOOL_CONFIG_FEATURES_FOR_READ,
1020 DATA_TYPE_NVLIST, NULL, &features) == 0 &&
1021 nvlist_check_features_for_read(features) != 0) {
1025 if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_STATE, DATA_TYPE_UINT64,
1030 if (val == POOL_STATE_DESTROYED) {
1031 /* We don't boot only from destroyed pools. */
1035 if (nvlist_find(nvlist, ZPOOL_CONFIG_POOL_TXG, DATA_TYPE_UINT64,
1036 NULL, &pool_txg) != 0 ||
1037 nvlist_find(nvlist, ZPOOL_CONFIG_POOL_GUID, DATA_TYPE_UINT64,
1038 NULL, &pool_guid) != 0 ||
1039 nvlist_find(nvlist, ZPOOL_CONFIG_POOL_NAME, DATA_TYPE_STRING,
1040 NULL, &pool_name) != 0) {
1042 * Cache and spare devices end up here - just ignore
1045 /*printf("ZFS: can't find pool details\n");*/
1049 if (nvlist_find(nvlist, ZPOOL_CONFIG_IS_LOG, DATA_TYPE_UINT64,
1050 NULL, &val) == 0 && val != 0) {
1055 * Create the pool if this is the first time we've seen it.
1057 spa = spa_find_by_guid(pool_guid);
1059 spa = spa_create(pool_guid, pool_name);
1063 if (pool_txg > spa->spa_txg) {
1064 spa->spa_txg = pool_txg;
1071 * Get the vdev tree and create our in-core copy of it.
1072 * If we already have a vdev with this guid, this must
1073 * be some kind of alias (overlapping slices, dangerously dedicated
1076 if (nvlist_find(nvlist, ZPOOL_CONFIG_GUID, DATA_TYPE_UINT64,
1077 NULL, &guid) != 0) {
1080 vdev = vdev_find(guid);
1081 if (vdev && vdev->v_phys_read) /* Has this vdev already been inited? */
1084 if (nvlist_find(nvlist, ZPOOL_CONFIG_VDEV_TREE, DATA_TYPE_NVLIST,
1089 rc = vdev_init_from_nvlist(vdevs, NULL, &top_vdev, is_newer);
1094 * Add the toplevel vdev to the pool if its not already there.
1096 STAILQ_FOREACH(pool_vdev, &spa->spa_vdevs, v_childlink)
1097 if (top_vdev == pool_vdev)
1099 if (!pool_vdev && top_vdev) {
1100 top_vdev->spa = spa;
1101 STAILQ_INSERT_TAIL(&spa->spa_vdevs, top_vdev, v_childlink);
1105 * We should already have created an incomplete vdev for this
1106 * vdev. Find it and initialise it with our read proc.
1108 vdev = vdev_find(guid);
1110 vdev->v_phys_read = _read;
1111 vdev->v_read_priv = read_priv;
1112 vdev->v_state = VDEV_STATE_HEALTHY;
1114 printf("ZFS: inconsistent nvlist contents\n");
1119 * Re-evaluate top-level vdev state.
1121 vdev_set_state(top_vdev);
1124 * Ok, we are happy with the pool so far. Lets find
1125 * the best uberblock and then we can actually access
1126 * the contents of the pool.
1128 upbuf = zfs_alloc(VDEV_UBERBLOCK_SIZE(vdev));
1129 up = (const struct uberblock *)upbuf;
1130 for (l = 0; l < VDEV_LABELS; l++) {
1131 for (i = 0; i < VDEV_UBERBLOCK_COUNT(vdev); i++) {
1132 off = vdev_label_offset(psize, l,
1133 VDEV_UBERBLOCK_OFFSET(vdev, i));
1135 DVA_SET_OFFSET(&bp.blk_dva[0], off);
1136 BP_SET_LSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1137 BP_SET_PSIZE(&bp, VDEV_UBERBLOCK_SIZE(vdev));
1138 BP_SET_CHECKSUM(&bp, ZIO_CHECKSUM_LABEL);
1139 BP_SET_COMPRESS(&bp, ZIO_COMPRESS_OFF);
1140 ZIO_SET_CHECKSUM(&bp.blk_cksum, off, 0, 0, 0);
1142 if (vdev_read_phys(vdev, &bp, upbuf, off, 0))
1145 if (up->ub_magic != UBERBLOCK_MAGIC)
1147 if (up->ub_txg < spa->spa_txg)
1149 if (up->ub_txg > spa->spa_uberblock.ub_txg ||
1150 (up->ub_txg == spa->spa_uberblock.ub_txg &&
1152 spa->spa_uberblock.ub_timestamp)) {
1153 spa->spa_uberblock = *up;
1157 zfs_free(upbuf, VDEV_UBERBLOCK_SIZE(vdev));
1170 for (v = 0; v < 32; v++)
1177 zio_read_gang(const spa_t *spa, const blkptr_t *bp, void *buf)
1180 zio_gbh_phys_t zio_gb;
1184 /* Artificial BP for gang block header. */
1186 BP_SET_PSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1187 BP_SET_LSIZE(&gbh_bp, SPA_GANGBLOCKSIZE);
1188 BP_SET_CHECKSUM(&gbh_bp, ZIO_CHECKSUM_GANG_HEADER);
1189 BP_SET_COMPRESS(&gbh_bp, ZIO_COMPRESS_OFF);
1190 for (i = 0; i < SPA_DVAS_PER_BP; i++)
1191 DVA_SET_GANG(&gbh_bp.blk_dva[i], 0);
1193 /* Read gang header block using the artificial BP. */
1194 if (zio_read(spa, &gbh_bp, &zio_gb))
1198 for (i = 0; i < SPA_GBH_NBLKPTRS; i++) {
1199 blkptr_t *gbp = &zio_gb.zg_blkptr[i];
1201 if (BP_IS_HOLE(gbp))
1203 if (zio_read(spa, gbp, pbuf))
1205 pbuf += BP_GET_PSIZE(gbp);
1208 if (zio_checksum_verify(spa, bp, buf))
1214 zio_read(const spa_t *spa, const blkptr_t *bp, void *buf)
1216 int cpfunc = BP_GET_COMPRESS(bp);
1217 uint64_t align, size;
1222 * Process data embedded in block pointer
1224 if (BP_IS_EMBEDDED(bp)) {
1225 ASSERT(BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA);
1227 size = BPE_GET_PSIZE(bp);
1228 ASSERT(size <= BPE_PAYLOAD_SIZE);
1230 if (cpfunc != ZIO_COMPRESS_OFF)
1231 pbuf = zfs_alloc(size);
1235 decode_embedded_bp_compressed(bp, pbuf);
1238 if (cpfunc != ZIO_COMPRESS_OFF) {
1239 error = zio_decompress_data(cpfunc, pbuf,
1240 size, buf, BP_GET_LSIZE(bp));
1241 zfs_free(pbuf, size);
1244 printf("ZFS: i/o error - unable to decompress block pointer data, error %d\n",
1251 for (i = 0; i < SPA_DVAS_PER_BP; i++) {
1252 const dva_t *dva = &bp->blk_dva[i];
1257 if (!dva->dva_word[0] && !dva->dva_word[1])
1260 vdevid = DVA_GET_VDEV(dva);
1261 offset = DVA_GET_OFFSET(dva);
1262 STAILQ_FOREACH(vdev, &spa->spa_vdevs, v_childlink) {
1263 if (vdev->v_id == vdevid)
1266 if (!vdev || !vdev->v_read)
1269 size = BP_GET_PSIZE(bp);
1270 if (vdev->v_read == vdev_raidz_read) {
1271 align = 1ULL << vdev->v_top->v_ashift;
1272 if (P2PHASE(size, align) != 0)
1273 size = P2ROUNDUP(size, align);
1275 if (size != BP_GET_PSIZE(bp) || cpfunc != ZIO_COMPRESS_OFF)
1276 pbuf = zfs_alloc(size);
1280 if (DVA_GET_GANG(dva))
1281 error = zio_read_gang(spa, bp, pbuf);
1283 error = vdev->v_read(vdev, bp, pbuf, offset, size);
1285 if (cpfunc != ZIO_COMPRESS_OFF)
1286 error = zio_decompress_data(cpfunc, pbuf,
1287 BP_GET_PSIZE(bp), buf, BP_GET_LSIZE(bp));
1288 else if (size != BP_GET_PSIZE(bp))
1289 bcopy(pbuf, buf, BP_GET_PSIZE(bp));
1292 zfs_free(pbuf, size);
1297 printf("ZFS: i/o error - all block copies unavailable\n");
1302 dnode_read(const spa_t *spa, const dnode_phys_t *dnode, off_t offset, void *buf, size_t buflen)
1304 int ibshift = dnode->dn_indblkshift - SPA_BLKPTRSHIFT;
1305 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1306 int nlevels = dnode->dn_nlevels;
1309 if (bsize > SPA_MAXBLOCKSIZE) {
1310 printf("ZFS: I/O error - blocks larger than %llu are not "
1311 "supported\n", SPA_MAXBLOCKSIZE);
1316 * Note: bsize may not be a power of two here so we need to do an
1317 * actual divide rather than a bitshift.
1319 while (buflen > 0) {
1320 uint64_t bn = offset / bsize;
1321 int boff = offset % bsize;
1323 const blkptr_t *indbp;
1326 if (bn > dnode->dn_maxblkid)
1329 if (dnode == dnode_cache_obj && bn == dnode_cache_bn)
1332 indbp = dnode->dn_blkptr;
1333 for (i = 0; i < nlevels; i++) {
1335 * Copy the bp from the indirect array so that
1336 * we can re-use the scratch buffer for multi-level
1339 ibn = bn >> ((nlevels - i - 1) * ibshift);
1340 ibn &= ((1 << ibshift) - 1);
1342 if (BP_IS_HOLE(&bp)) {
1343 memset(dnode_cache_buf, 0, bsize);
1346 rc = zio_read(spa, &bp, dnode_cache_buf);
1349 indbp = (const blkptr_t *) dnode_cache_buf;
1351 dnode_cache_obj = dnode;
1352 dnode_cache_bn = bn;
1356 * The buffer contains our data block. Copy what we
1357 * need from it and loop.
1360 if (i > buflen) i = buflen;
1361 memcpy(buf, &dnode_cache_buf[boff], i);
1362 buf = ((char*) buf) + i;
1371 * Lookup a value in a microzap directory. Assumes that the zap
1372 * scratch buffer contains the directory contents.
1375 mzap_lookup(const dnode_phys_t *dnode, const char *name, uint64_t *value)
1377 const mzap_phys_t *mz;
1378 const mzap_ent_phys_t *mze;
1383 * Microzap objects use exactly one block. Read the whole
1386 size = dnode->dn_datablkszsec * 512;
1388 mz = (const mzap_phys_t *) zap_scratch;
1389 chunks = size / MZAP_ENT_LEN - 1;
1391 for (i = 0; i < chunks; i++) {
1392 mze = &mz->mz_chunk[i];
1393 if (!strcmp(mze->mze_name, name)) {
1394 *value = mze->mze_value;
1403 * Compare a name with a zap leaf entry. Return non-zero if the name
1407 fzap_name_equal(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, const char *name)
1410 const zap_leaf_chunk_t *nc;
1413 namelen = zc->l_entry.le_name_numints;
1415 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1417 while (namelen > 0) {
1420 if (len > ZAP_LEAF_ARRAY_BYTES)
1421 len = ZAP_LEAF_ARRAY_BYTES;
1422 if (memcmp(p, nc->l_array.la_array, len))
1426 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1433 * Extract a uint64_t value from a zap leaf entry.
1436 fzap_leaf_value(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc)
1438 const zap_leaf_chunk_t *vc;
1443 vc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_value_chunk);
1444 for (i = 0, value = 0, p = vc->l_array.la_array; i < 8; i++) {
1445 value = (value << 8) | p[i];
1452 stv(int len, void *addr, uint64_t value)
1456 *(uint8_t *)addr = value;
1459 *(uint16_t *)addr = value;
1462 *(uint32_t *)addr = value;
1465 *(uint64_t *)addr = value;
1471 * Extract a array from a zap leaf entry.
1474 fzap_leaf_array(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc,
1475 uint64_t integer_size, uint64_t num_integers, void *buf)
1477 uint64_t array_int_len = zc->l_entry.le_value_intlen;
1479 uint64_t *u64 = buf;
1481 int len = MIN(zc->l_entry.le_value_numints, num_integers);
1482 int chunk = zc->l_entry.le_value_chunk;
1485 if (integer_size == 8 && len == 1) {
1486 *u64 = fzap_leaf_value(zl, zc);
1491 struct zap_leaf_array *la = &ZAP_LEAF_CHUNK(zl, chunk).l_array;
1494 ASSERT3U(chunk, <, ZAP_LEAF_NUMCHUNKS(zl));
1495 for (i = 0; i < ZAP_LEAF_ARRAY_BYTES && len > 0; i++) {
1496 value = (value << 8) | la->la_array[i];
1498 if (byten == array_int_len) {
1499 stv(integer_size, p, value);
1507 chunk = la->la_next;
1512 * Lookup a value in a fatzap directory. Assumes that the zap scratch
1513 * buffer contains the directory header.
1516 fzap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name,
1517 uint64_t integer_size, uint64_t num_integers, void *value)
1519 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1520 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1526 if (zh.zap_magic != ZAP_MAGIC)
1529 z.zap_block_shift = ilog2(bsize);
1530 z.zap_phys = (zap_phys_t *) zap_scratch;
1533 * Figure out where the pointer table is and read it in if necessary.
1535 if (zh.zap_ptrtbl.zt_blk) {
1536 rc = dnode_read(spa, dnode, zh.zap_ptrtbl.zt_blk * bsize,
1537 zap_scratch, bsize);
1540 ptrtbl = (uint64_t *) zap_scratch;
1542 ptrtbl = &ZAP_EMBEDDED_PTRTBL_ENT(&z, 0);
1545 hash = zap_hash(zh.zap_salt, name);
1548 zl.l_bs = z.zap_block_shift;
1550 off_t off = ptrtbl[hash >> (64 - zh.zap_ptrtbl.zt_shift)] << zl.l_bs;
1551 zap_leaf_chunk_t *zc;
1553 rc = dnode_read(spa, dnode, off, zap_scratch, bsize);
1557 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1560 * Make sure this chunk matches our hash.
1562 if (zl.l_phys->l_hdr.lh_prefix_len > 0
1563 && zl.l_phys->l_hdr.lh_prefix
1564 != hash >> (64 - zl.l_phys->l_hdr.lh_prefix_len))
1568 * Hash within the chunk to find our entry.
1570 int shift = (64 - ZAP_LEAF_HASH_SHIFT(&zl) - zl.l_phys->l_hdr.lh_prefix_len);
1571 int h = (hash >> shift) & ((1 << ZAP_LEAF_HASH_SHIFT(&zl)) - 1);
1572 h = zl.l_phys->l_hash[h];
1575 zc = &ZAP_LEAF_CHUNK(&zl, h);
1576 while (zc->l_entry.le_hash != hash) {
1577 if (zc->l_entry.le_next == 0xffff) {
1581 zc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_next);
1583 if (fzap_name_equal(&zl, zc, name)) {
1584 if (zc->l_entry.le_value_intlen * zc->l_entry.le_value_numints >
1585 integer_size * num_integers)
1587 fzap_leaf_array(&zl, zc, integer_size, num_integers, value);
1595 * Lookup a name in a zap object and return its value as a uint64_t.
1598 zap_lookup(const spa_t *spa, const dnode_phys_t *dnode, const char *name,
1599 uint64_t integer_size, uint64_t num_integers, void *value)
1603 size_t size = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1605 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1609 zap_type = *(uint64_t *) zap_scratch;
1610 if (zap_type == ZBT_MICRO)
1611 return mzap_lookup(dnode, name, value);
1612 else if (zap_type == ZBT_HEADER) {
1613 return fzap_lookup(spa, dnode, name, integer_size,
1614 num_integers, value);
1616 printf("ZFS: invalid zap_type=%d\n", (int)zap_type);
1621 * List a microzap directory. Assumes that the zap scratch buffer contains
1622 * the directory contents.
1625 mzap_list(const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1627 const mzap_phys_t *mz;
1628 const mzap_ent_phys_t *mze;
1633 * Microzap objects use exactly one block. Read the whole
1636 size = dnode->dn_datablkszsec * 512;
1637 mz = (const mzap_phys_t *) zap_scratch;
1638 chunks = size / MZAP_ENT_LEN - 1;
1640 for (i = 0; i < chunks; i++) {
1641 mze = &mz->mz_chunk[i];
1642 if (mze->mze_name[0]) {
1643 rc = callback(mze->mze_name, mze->mze_value);
1653 * List a fatzap directory. Assumes that the zap scratch buffer contains
1654 * the directory header.
1657 fzap_list(const spa_t *spa, const dnode_phys_t *dnode, int (*callback)(const char *, uint64_t))
1659 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1660 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1664 if (zh.zap_magic != ZAP_MAGIC)
1667 z.zap_block_shift = ilog2(bsize);
1668 z.zap_phys = (zap_phys_t *) zap_scratch;
1671 * This assumes that the leaf blocks start at block 1. The
1672 * documentation isn't exactly clear on this.
1675 zl.l_bs = z.zap_block_shift;
1676 for (i = 0; i < zh.zap_num_leafs; i++) {
1677 off_t off = (i + 1) << zl.l_bs;
1681 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1684 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1686 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1687 zap_leaf_chunk_t *zc, *nc;
1690 zc = &ZAP_LEAF_CHUNK(&zl, j);
1691 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1693 namelen = zc->l_entry.le_name_numints;
1694 if (namelen > sizeof(name))
1695 namelen = sizeof(name);
1698 * Paste the name back together.
1700 nc = &ZAP_LEAF_CHUNK(&zl, zc->l_entry.le_name_chunk);
1702 while (namelen > 0) {
1705 if (len > ZAP_LEAF_ARRAY_BYTES)
1706 len = ZAP_LEAF_ARRAY_BYTES;
1707 memcpy(p, nc->l_array.la_array, len);
1710 nc = &ZAP_LEAF_CHUNK(&zl, nc->l_array.la_next);
1714 * Assume the first eight bytes of the value are
1717 value = fzap_leaf_value(&zl, zc);
1719 //printf("%s 0x%jx\n", name, (uintmax_t)value);
1720 rc = callback((const char *)name, value);
1729 static int zfs_printf(const char *name, uint64_t value __unused)
1732 printf("%s\n", name);
1738 * List a zap directory.
1741 zap_list(const spa_t *spa, const dnode_phys_t *dnode)
1744 size_t size = dnode->dn_datablkszsec * 512;
1746 if (dnode_read(spa, dnode, 0, zap_scratch, size))
1749 zap_type = *(uint64_t *) zap_scratch;
1750 if (zap_type == ZBT_MICRO)
1751 return mzap_list(dnode, zfs_printf);
1753 return fzap_list(spa, dnode, zfs_printf);
1757 objset_get_dnode(const spa_t *spa, const objset_phys_t *os, uint64_t objnum, dnode_phys_t *dnode)
1761 offset = objnum * sizeof(dnode_phys_t);
1762 return dnode_read(spa, &os->os_meta_dnode, offset,
1763 dnode, sizeof(dnode_phys_t));
1767 mzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1769 const mzap_phys_t *mz;
1770 const mzap_ent_phys_t *mze;
1775 * Microzap objects use exactly one block. Read the whole
1778 size = dnode->dn_datablkszsec * 512;
1780 mz = (const mzap_phys_t *) zap_scratch;
1781 chunks = size / MZAP_ENT_LEN - 1;
1783 for (i = 0; i < chunks; i++) {
1784 mze = &mz->mz_chunk[i];
1785 if (value == mze->mze_value) {
1786 strcpy(name, mze->mze_name);
1795 fzap_name_copy(const zap_leaf_t *zl, const zap_leaf_chunk_t *zc, char *name)
1798 const zap_leaf_chunk_t *nc;
1801 namelen = zc->l_entry.le_name_numints;
1803 nc = &ZAP_LEAF_CHUNK(zl, zc->l_entry.le_name_chunk);
1805 while (namelen > 0) {
1808 if (len > ZAP_LEAF_ARRAY_BYTES)
1809 len = ZAP_LEAF_ARRAY_BYTES;
1810 memcpy(p, nc->l_array.la_array, len);
1813 nc = &ZAP_LEAF_CHUNK(zl, nc->l_array.la_next);
1820 fzap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1822 int bsize = dnode->dn_datablkszsec << SPA_MINBLOCKSHIFT;
1823 zap_phys_t zh = *(zap_phys_t *) zap_scratch;
1827 if (zh.zap_magic != ZAP_MAGIC)
1830 z.zap_block_shift = ilog2(bsize);
1831 z.zap_phys = (zap_phys_t *) zap_scratch;
1834 * This assumes that the leaf blocks start at block 1. The
1835 * documentation isn't exactly clear on this.
1838 zl.l_bs = z.zap_block_shift;
1839 for (i = 0; i < zh.zap_num_leafs; i++) {
1840 off_t off = (i + 1) << zl.l_bs;
1842 if (dnode_read(spa, dnode, off, zap_scratch, bsize))
1845 zl.l_phys = (zap_leaf_phys_t *) zap_scratch;
1847 for (j = 0; j < ZAP_LEAF_NUMCHUNKS(&zl); j++) {
1848 zap_leaf_chunk_t *zc;
1850 zc = &ZAP_LEAF_CHUNK(&zl, j);
1851 if (zc->l_entry.le_type != ZAP_CHUNK_ENTRY)
1853 if (zc->l_entry.le_value_intlen != 8 ||
1854 zc->l_entry.le_value_numints != 1)
1857 if (fzap_leaf_value(&zl, zc) == value) {
1858 fzap_name_copy(&zl, zc, name);
1868 zap_rlookup(const spa_t *spa, const dnode_phys_t *dnode, char *name, uint64_t value)
1872 size_t size = dnode->dn_datablkszsec * 512;
1874 rc = dnode_read(spa, dnode, 0, zap_scratch, size);
1878 zap_type = *(uint64_t *) zap_scratch;
1879 if (zap_type == ZBT_MICRO)
1880 return mzap_rlookup(spa, dnode, name, value);
1882 return fzap_rlookup(spa, dnode, name, value);
1886 zfs_rlookup(const spa_t *spa, uint64_t objnum, char *result)
1889 char component[256];
1890 uint64_t dir_obj, parent_obj, child_dir_zapobj;
1891 dnode_phys_t child_dir_zap, dataset, dir, parent;
1893 dsl_dataset_phys_t *ds;
1897 p = &name[sizeof(name) - 1];
1900 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
1901 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
1904 ds = (dsl_dataset_phys_t *)&dataset.dn_bonus;
1905 dir_obj = ds->ds_dir_obj;
1908 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir) != 0)
1910 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1912 /* Actual loop condition. */
1913 parent_obj = dd->dd_parent_obj;
1914 if (parent_obj == 0)
1917 if (objset_get_dnode(spa, &spa->spa_mos, parent_obj, &parent) != 0)
1919 dd = (dsl_dir_phys_t *)&parent.dn_bonus;
1920 child_dir_zapobj = dd->dd_child_dir_zapobj;
1921 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1923 if (zap_rlookup(spa, &child_dir_zap, component, dir_obj) != 0)
1926 len = strlen(component);
1928 memcpy(p, component, len);
1932 /* Actual loop iteration. */
1933 dir_obj = parent_obj;
1944 zfs_lookup_dataset(const spa_t *spa, const char *name, uint64_t *objnum)
1947 uint64_t dir_obj, child_dir_zapobj;
1948 dnode_phys_t child_dir_zap, dir;
1952 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir))
1954 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof (dir_obj),
1960 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir))
1962 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
1966 /* Actual loop condition #1. */
1972 memcpy(element, p, q - p);
1973 element[q - p] = '\0';
1980 child_dir_zapobj = dd->dd_child_dir_zapobj;
1981 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0)
1984 /* Actual loop condition #2. */
1985 if (zap_lookup(spa, &child_dir_zap, element, sizeof (dir_obj),
1990 *objnum = dd->dd_head_dataset_obj;
1996 zfs_list_dataset(const spa_t *spa, uint64_t objnum/*, int pos, char *entry*/)
1998 uint64_t dir_obj, child_dir_zapobj;
1999 dnode_phys_t child_dir_zap, dir, dataset;
2000 dsl_dataset_phys_t *ds;
2003 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
2004 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
2007 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
2008 dir_obj = ds->ds_dir_obj;
2010 if (objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir)) {
2011 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
2014 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
2016 child_dir_zapobj = dd->dd_child_dir_zapobj;
2017 if (objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap) != 0) {
2018 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
2022 return (zap_list(spa, &child_dir_zap) != 0);
2026 zfs_callback_dataset(const spa_t *spa, uint64_t objnum, int (*callback)(const char *, uint64_t))
2028 uint64_t dir_obj, child_dir_zapobj, zap_type;
2029 dnode_phys_t child_dir_zap, dir, dataset;
2030 dsl_dataset_phys_t *ds;
2034 err = objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset);
2036 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
2039 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
2040 dir_obj = ds->ds_dir_obj;
2042 err = objset_get_dnode(spa, &spa->spa_mos, dir_obj, &dir);
2044 printf("ZFS: can't find dirobj %ju\n", (uintmax_t)dir_obj);
2047 dd = (dsl_dir_phys_t *)&dir.dn_bonus;
2049 child_dir_zapobj = dd->dd_child_dir_zapobj;
2050 err = objset_get_dnode(spa, &spa->spa_mos, child_dir_zapobj, &child_dir_zap);
2052 printf("ZFS: can't find child zap %ju\n", (uintmax_t)dir_obj);
2056 err = dnode_read(spa, &child_dir_zap, 0, zap_scratch, child_dir_zap.dn_datablkszsec * 512);
2060 zap_type = *(uint64_t *) zap_scratch;
2061 if (zap_type == ZBT_MICRO)
2062 return mzap_list(&child_dir_zap, callback);
2064 return fzap_list(spa, &child_dir_zap, callback);
2069 * Find the object set given the object number of its dataset object
2070 * and return its details in *objset
2073 zfs_mount_dataset(const spa_t *spa, uint64_t objnum, objset_phys_t *objset)
2075 dnode_phys_t dataset;
2076 dsl_dataset_phys_t *ds;
2078 if (objset_get_dnode(spa, &spa->spa_mos, objnum, &dataset)) {
2079 printf("ZFS: can't find dataset %ju\n", (uintmax_t)objnum);
2083 ds = (dsl_dataset_phys_t *) &dataset.dn_bonus;
2084 if (zio_read(spa, &ds->ds_bp, objset)) {
2085 printf("ZFS: can't read object set for dataset %ju\n",
2094 * Find the object set pointed to by the BOOTFS property or the root
2095 * dataset if there is none and return its details in *objset
2098 zfs_get_root(const spa_t *spa, uint64_t *objid)
2100 dnode_phys_t dir, propdir;
2101 uint64_t props, bootfs, root;
2106 * Start with the MOS directory object.
2108 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT, &dir)) {
2109 printf("ZFS: can't read MOS object directory\n");
2114 * Lookup the pool_props and see if we can find a bootfs.
2116 if (zap_lookup(spa, &dir, DMU_POOL_PROPS, sizeof (props), 1, &props) == 0
2117 && objset_get_dnode(spa, &spa->spa_mos, props, &propdir) == 0
2118 && zap_lookup(spa, &propdir, "bootfs", sizeof (bootfs), 1, &bootfs) == 0
2125 * Lookup the root dataset directory
2127 if (zap_lookup(spa, &dir, DMU_POOL_ROOT_DATASET, sizeof (root), 1, &root)
2128 || objset_get_dnode(spa, &spa->spa_mos, root, &dir)) {
2129 printf("ZFS: can't find root dsl_dir\n");
2134 * Use the information from the dataset directory's bonus buffer
2135 * to find the dataset object and from that the object set itself.
2137 dsl_dir_phys_t *dd = (dsl_dir_phys_t *) &dir.dn_bonus;
2138 *objid = dd->dd_head_dataset_obj;
2143 zfs_mount(const spa_t *spa, uint64_t rootobj, struct zfsmount *mount)
2149 * Find the root object set if not explicitly provided
2151 if (rootobj == 0 && zfs_get_root(spa, &rootobj)) {
2152 printf("ZFS: can't find root filesystem\n");
2156 if (zfs_mount_dataset(spa, rootobj, &mount->objset)) {
2157 printf("ZFS: can't open root filesystem\n");
2161 mount->rootobj = rootobj;
2167 * callback function for feature name checks.
2170 check_feature(const char *name, uint64_t value)
2176 if (name[0] == '\0')
2179 for (i = 0; features_for_read[i] != NULL; i++) {
2180 if (strcmp(name, features_for_read[i]) == 0)
2183 printf("ZFS: unsupported feature: %s\n", name);
2188 * Checks whether the MOS features that are active are supported.
2191 check_mos_features(const spa_t *spa)
2194 uint64_t objnum, zap_type;
2198 if ((rc = objset_get_dnode(spa, &spa->spa_mos, DMU_OT_OBJECT_DIRECTORY,
2201 if ((rc = zap_lookup(spa, &dir, DMU_POOL_FEATURES_FOR_READ,
2202 sizeof (objnum), 1, &objnum)) != 0) {
2204 * It is older pool without features. As we have already
2205 * tested the label, just return without raising the error.
2210 if ((rc = objset_get_dnode(spa, &spa->spa_mos, objnum, &dir)) != 0)
2213 if (dir.dn_type != DMU_OTN_ZAP_METADATA)
2216 size = dir.dn_datablkszsec * 512;
2217 if (dnode_read(spa, &dir, 0, zap_scratch, size))
2220 zap_type = *(uint64_t *) zap_scratch;
2221 if (zap_type == ZBT_MICRO)
2222 rc = mzap_list(&dir, check_feature);
2224 rc = fzap_list(spa, &dir, check_feature);
2230 zfs_spa_init(spa_t *spa)
2235 if (zio_read(spa, &spa->spa_uberblock.ub_rootbp, &spa->spa_mos)) {
2236 printf("ZFS: can't read MOS of pool %s\n", spa->spa_name);
2239 if (spa->spa_mos.os_type != DMU_OST_META) {
2240 printf("ZFS: corrupted MOS of pool %s\n", spa->spa_name);
2244 if (objset_get_dnode(spa, &spa->spa_mos, DMU_POOL_DIRECTORY_OBJECT,
2246 printf("ZFS: failed to read pool %s directory object\n",
2250 /* this is allowed to fail, older pools do not have salt */
2251 rc = zap_lookup(spa, &dir, DMU_POOL_CHECKSUM_SALT, 1,
2252 sizeof (spa->spa_cksum_salt.zcs_bytes),
2253 spa->spa_cksum_salt.zcs_bytes);
2255 rc = check_mos_features(spa);
2257 printf("ZFS: pool %s is not supported\n", spa->spa_name);
2264 zfs_dnode_stat(const spa_t *spa, dnode_phys_t *dn, struct stat *sb)
2267 if (dn->dn_bonustype != DMU_OT_SA) {
2268 znode_phys_t *zp = (znode_phys_t *)dn->dn_bonus;
2270 sb->st_mode = zp->zp_mode;
2271 sb->st_uid = zp->zp_uid;
2272 sb->st_gid = zp->zp_gid;
2273 sb->st_size = zp->zp_size;
2275 sa_hdr_phys_t *sahdrp;
2280 if (dn->dn_bonuslen != 0)
2281 sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn);
2283 if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) != 0) {
2284 blkptr_t *bp = DN_SPILL_BLKPTR(dn);
2287 size = BP_GET_LSIZE(bp);
2288 buf = zfs_alloc(size);
2289 error = zio_read(spa, bp, buf);
2291 zfs_free(buf, size);
2299 hdrsize = SA_HDR_SIZE(sahdrp);
2300 sb->st_mode = *(uint64_t *)((char *)sahdrp + hdrsize +
2302 sb->st_uid = *(uint64_t *)((char *)sahdrp + hdrsize +
2304 sb->st_gid = *(uint64_t *)((char *)sahdrp + hdrsize +
2306 sb->st_size = *(uint64_t *)((char *)sahdrp + hdrsize +
2309 zfs_free(buf, size);
2316 zfs_dnode_readlink(const spa_t *spa, dnode_phys_t *dn, char *path, size_t psize)
2320 if (dn->dn_bonustype == DMU_OT_SA) {
2321 sa_hdr_phys_t *sahdrp = NULL;
2327 if (dn->dn_bonuslen != 0)
2328 sahdrp = (sa_hdr_phys_t *)DN_BONUS(dn);
2332 if ((dn->dn_flags & DNODE_FLAG_SPILL_BLKPTR) == 0)
2334 bp = DN_SPILL_BLKPTR(dn);
2336 size = BP_GET_LSIZE(bp);
2337 buf = zfs_alloc(size);
2338 rc = zio_read(spa, bp, buf);
2340 zfs_free(buf, size);
2345 hdrsize = SA_HDR_SIZE(sahdrp);
2346 p = (char *)((uintptr_t)sahdrp + hdrsize + SA_SYMLINK_OFFSET);
2347 memcpy(path, p, psize);
2349 zfs_free(buf, size);
2353 * Second test is purely to silence bogus compiler
2354 * warning about accessing past the end of dn_bonus.
2356 if (psize + sizeof(znode_phys_t) <= dn->dn_bonuslen &&
2357 sizeof(znode_phys_t) <= sizeof(dn->dn_bonus)) {
2358 memcpy(path, &dn->dn_bonus[sizeof(znode_phys_t)], psize);
2360 rc = dnode_read(spa, dn, 0, path, psize);
2367 STAILQ_ENTRY(obj_list) entry;
2371 * Lookup a file and return its dnode.
2374 zfs_lookup(const struct zfsmount *mount, const char *upath, dnode_phys_t *dnode)
2383 int symlinks_followed = 0;
2385 struct obj_list *entry, *tentry;
2386 STAILQ_HEAD(, obj_list) on_cache = STAILQ_HEAD_INITIALIZER(on_cache);
2389 if (mount->objset.os_type != DMU_OST_ZFS) {
2390 printf("ZFS: unexpected object set type %ju\n",
2391 (uintmax_t)mount->objset.os_type);
2395 if ((entry = malloc(sizeof(struct obj_list))) == NULL)
2399 * Get the root directory dnode.
2401 rc = objset_get_dnode(spa, &mount->objset, MASTER_NODE_OBJ, &dn);
2407 rc = zap_lookup(spa, &dn, ZFS_ROOT_OBJ, sizeof (objnum), 1, &objnum);
2412 entry->objnum = objnum;
2413 STAILQ_INSERT_HEAD(&on_cache, entry, entry);
2415 rc = objset_get_dnode(spa, &mount->objset, objnum, &dn);
2421 rc = objset_get_dnode(spa, &mount->objset, objnum, &dn);
2430 while (*q != '\0' && *q != '/')
2434 if (p + 1 == q && p[0] == '.') {
2439 if (p + 2 == q && p[0] == '.' && p[1] == '.') {
2441 if (STAILQ_FIRST(&on_cache) ==
2442 STAILQ_LAST(&on_cache, obj_list, entry)) {
2446 entry = STAILQ_FIRST(&on_cache);
2447 STAILQ_REMOVE_HEAD(&on_cache, entry);
2449 objnum = (STAILQ_FIRST(&on_cache))->objnum;
2452 if (q - p + 1 > sizeof(element)) {
2456 memcpy(element, p, q - p);
2460 if ((rc = zfs_dnode_stat(spa, &dn, &sb)) != 0)
2462 if (!S_ISDIR(sb.st_mode)) {
2467 rc = zap_lookup(spa, &dn, element, sizeof (objnum), 1, &objnum);
2470 objnum = ZFS_DIRENT_OBJ(objnum);
2472 if ((entry = malloc(sizeof(struct obj_list))) == NULL) {
2476 entry->objnum = objnum;
2477 STAILQ_INSERT_HEAD(&on_cache, entry, entry);
2478 rc = objset_get_dnode(spa, &mount->objset, objnum, &dn);
2483 * Check for symlink.
2485 rc = zfs_dnode_stat(spa, &dn, &sb);
2488 if (S_ISLNK(sb.st_mode)) {
2489 if (symlinks_followed > 10) {
2493 symlinks_followed++;
2496 * Read the link value and copy the tail of our
2497 * current path onto the end.
2499 if (sb.st_size + strlen(p) + 1 > sizeof(path)) {
2503 strcpy(&path[sb.st_size], p);
2505 rc = zfs_dnode_readlink(spa, &dn, path, sb.st_size);
2510 * Restart with the new path, starting either at
2511 * the root or at the parent depending whether or
2512 * not the link is relative.
2516 while (STAILQ_FIRST(&on_cache) !=
2517 STAILQ_LAST(&on_cache, obj_list, entry)) {
2518 entry = STAILQ_FIRST(&on_cache);
2519 STAILQ_REMOVE_HEAD(&on_cache, entry);
2523 entry = STAILQ_FIRST(&on_cache);
2524 STAILQ_REMOVE_HEAD(&on_cache, entry);
2527 objnum = (STAILQ_FIRST(&on_cache))->objnum;
2533 STAILQ_FOREACH_SAFE(entry, &on_cache, entry, tentry)