int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
int err = 0;
dmu_tx_t *tx;
nvlist_t *nvarg;
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
return (SET_ERROR(EINVAL));
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
VERIFY0(nvlist_dup(nvl, &nvarg, KM_PUSHPAGE));
if (spa_history_zone() != NULL) {
fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
spa_history_zone());
}
fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
/* Kick this off asynchronously; errors are ignored. */
dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
nvarg, 0, tx);
dmu_tx_commit(tx);
/* spa_history_log_sync will free nvl */
return (err);
}
boolean_t
dmu_objset_userused_enabled(objset_impl_t *os)
{
return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE &&
used_cbs[os->os_phys->os_type] &&
os->os_userused_dnode);
}
int
dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset)
{
dsl_dir_t *dd;
int error;
nvpair_t *whopair = NULL;
int blocks_modified = 0;
error = dsl_dir_open(ddname, FTAG, &dd, NULL);
if (error)
return (error);
if (spa_version(dmu_objset_spa(dd->dd_pool->dp_meta_objset)) <
SPA_VERSION_DELEGATED_PERMS) {
dsl_dir_close(dd, FTAG);
return (ENOTSUP);
}
while ((whopair = nvlist_next_nvpair(nvp, whopair)))
blocks_modified++;
error = dsl_sync_task_do(dd->dd_pool, NULL,
unset ? dsl_deleg_unset_sync : dsl_deleg_set_sync,
dd, nvp, blocks_modified);
dsl_dir_close(dd, FTAG);
return (error);
}
int
dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp)
{
dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus;
dnode_t *dn;
int err;
DB_DNODE_ENTER(db);
dn = DB_DNODE(db);
if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) {
err = EINVAL;
} else {
rw_enter(&dn->dn_struct_rwlock, RW_READER);
if (!dn->dn_have_spill) {
err = ENOENT;
} else {
err = dmu_spill_hold_by_dnode(dn,
DB_RF_HAVESTRUCT | DB_RF_CANFAIL, tag, dbp);
}
rw_exit(&dn->dn_struct_rwlock);
}
DB_DNODE_EXIT(db);
return (err);
}
boolean_t
dmu_objset_userused_enabled(objset_t *os)
{
return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE &&
used_cbs[os->os_phys->os_type] != NULL &&
DMU_USERUSED_DNODE(os) != NULL);
}
int
dsl_destroy_snapshot_check_impl(dsl_dataset_t *ds, boolean_t defer)
{
if (!ds->ds_is_snapshot)
return (SET_ERROR(EINVAL));
if (dsl_dataset_long_held(ds))
return (SET_ERROR(EBUSY));
/*
* Only allow deferred destroy on pools that support it.
* NOTE: deferred destroy is only supported on snapshots.
*/
if (defer) {
if (spa_version(ds->ds_dir->dd_pool->dp_spa) <
SPA_VERSION_USERREFS)
return (SET_ERROR(ENOTSUP));
return (0);
}
/*
* If this snapshot has an elevated user reference count,
* we can't destroy it yet.
*/
if (ds->ds_userrefs > 0)
return (SET_ERROR(EBUSY));
/*
* Can't delete a branch point.
*/
if (dsl_dataset_phys(ds)->ds_num_children > 1)
return (SET_ERROR(EEXIST));
return (0);
}
void
dsl_prop_check_prediction(dsl_dir_t *dd, dsl_prop_setarg_t *psa)
{
zfs_prop_t prop = zfs_name_to_prop(psa->psa_name);
uint64_t intval;
char setpoint[MAXNAMELEN];
uint64_t version = spa_version(dd->dd_pool->dp_spa);
int err;
if (version < SPA_VERSION_RECVD_PROPS) {
switch (prop) {
case ZFS_PROP_QUOTA:
case ZFS_PROP_RESERVATION:
return;
default:
break;
}
}
err = dsl_prop_get_dd(dd, psa->psa_name, 8, 1, &intval,
setpoint, B_FALSE);
if (err == 0 && intval != psa->psa_effective_value) {
cmn_err(CE_PANIC, "%s property, source: %x, "
"predicted effective value: %llu, "
"actual effective value: %llu (setpoint: %s)",
psa->psa_name, psa->psa_source,
(unsigned long long)psa->psa_effective_value,
(unsigned long long)intval, setpoint);
}
}
void
dsl_destroy_snapshot_sync_impl(dsl_dataset_t *ds, boolean_t defer, dmu_tx_t *tx)
{
#ifdef ZFS_DEBUG
int err;
#endif
int after_branch_point = FALSE;
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
dsl_dataset_t *ds_prev = NULL;
uint64_t obj, old_unique, used = 0, comp = 0, uncomp = 0;
dsl_dataset_t *ds_next, *ds_head, *hds;
ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
ASSERT3U(ds->ds_phys->ds_bp.blk_birth, <=, tx->tx_txg);
ASSERT(refcount_is_zero(&ds->ds_longholds));
if (defer &&
(ds->ds_userrefs > 0 || ds->ds_phys->ds_num_children > 1)) {
ASSERT(spa_version(dp->dp_spa) >= SPA_VERSION_USERREFS);
dmu_buf_will_dirty(ds->ds_dbuf, tx);
ds->ds_phys->ds_flags |= DS_FLAG_DEFER_DESTROY;
spa_history_log_internal_ds(ds, "defer_destroy", tx, "");
return;
}
ASSERT3U(ds->ds_phys->ds_num_children, <=, 1);
/* We need to log before removing it from the namespace. */
spa_history_log_internal_ds(ds, "destroy", tx, "");
dsl_scan_ds_destroyed(ds, tx);
obj = ds->ds_object;
if (ds->ds_phys->ds_prev_snap_obj != 0) {
ASSERT3P(ds->ds_prev, ==, NULL);
VERIFY0(dsl_dataset_hold_obj(dp,
ds->ds_phys->ds_prev_snap_obj, FTAG, &ds_prev));
after_branch_point =
(ds_prev->ds_phys->ds_next_snap_obj != obj);
dmu_buf_will_dirty(ds_prev->ds_dbuf, tx);
if (after_branch_point &&
ds_prev->ds_phys->ds_next_clones_obj != 0) {
dsl_dataset_remove_from_next_clones(ds_prev, obj, tx);
if (ds->ds_phys->ds_next_snap_obj != 0) {
VERIFY0(zap_add_int(mos,
ds_prev->ds_phys->ds_next_clones_obj,
ds->ds_phys->ds_next_snap_obj, tx));
}
}
if (!after_branch_point) {
ds_prev->ds_phys->ds_next_snap_obj =
ds->ds_phys->ds_next_snap_obj;
}
}
void
spa_history_log_version(spa_t *spa, const char *operation)
{
spa_history_log_internal(spa, operation, NULL,
"pool version %llu; software version %llu/%d; uts %s %s %s %s",
(u_longlong_t)spa_version(spa), SPA_VERSION, ZPL_VERSION,
utsname.nodename, utsname.release, utsname.version,
utsname.machine);
}
static void
zhack_spa_open(const char *target, boolean_t readonly, void *tag, spa_t **spa)
{
int err;
import_pool(target, readonly);
zfeature_checks_disable = B_TRUE;
err = spa_open(target, spa, tag);
zfeature_checks_disable = B_FALSE;
if (err != 0)
fatal("cannot open '%s': %s", target, strerror(err));
if (spa_version(*spa) < SPA_VERSION_FEATURES) {
fatal("'%s' has version %d, features not enabled", target,
(int)spa_version(*spa));
}
}
void
spa_history_log_version(spa_t *spa, const char *operation, dmu_tx_t *tx)
{
utsname_t *u = utsname();
spa_history_log_internal(spa, operation, tx,
"pool version %llu; software version %llu/%llu; uts %s %s %s %s",
(u_longlong_t)spa_version(spa), SPA_VERSION, ZPL_VERSION,
u->nodename, u->release, u->version, u->machine);
}
uint64_t
bplist_create(objset_t *mos, int blocksize, dmu_tx_t *tx)
{
int size;
size = spa_version(dmu_objset_spa(mos)) < SPA_VERSION_BPLIST_ACCOUNT ?
BPLIST_SIZE_V0 : sizeof (bplist_phys_t);
return (dmu_object_alloc(mos, DMU_OT_BPLIST, blocksize,
DMU_OT_BPLIST_HDR, size, tx));
}
int
zfs_set_version(zfs_sb_t *zsb, uint64_t newvers)
{
int error;
objset_t *os = zsb->z_os;
dmu_tx_t *tx;
if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
return (SET_ERROR(EINVAL));
if (newvers < zsb->z_version)
return (SET_ERROR(EINVAL));
if (zfs_spa_version_map(newvers) >
spa_version(dmu_objset_spa(zsb->z_os)))
return (SET_ERROR(ENOTSUP));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR);
if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
ZFS_SA_ATTRS);
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL);
}
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
8, 1, &newvers, tx);
if (error) {
dmu_tx_commit(tx);
return (error);
}
if (newvers >= ZPL_VERSION_SA && !zsb->z_use_sa) {
uint64_t sa_obj;
ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=,
SPA_VERSION_SA);
sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
DMU_OT_NONE, 0, tx);
error = zap_add(os, MASTER_NODE_OBJ,
ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
ASSERT0(error);
VERIFY(0 == sa_set_sa_object(os, sa_obj));
sa_register_update_callback(os, zfs_sa_upgrade);
}
static int
dsl_dataset_user_hold_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_hold_arg_t *dduha = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvpair_t *pair;
if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS)
return (SET_ERROR(ENOTSUP));
if (!dmu_tx_is_syncing(tx))
return (0);
for (pair = nvlist_next_nvpair(dduha->dduha_holds, NULL);
pair != NULL; pair = nvlist_next_nvpair(dduha->dduha_holds, pair)) {
dsl_dataset_t *ds;
int error = 0;
char *htag, *name;
/* must be a snapshot */
name = nvpair_name(pair);
if (strchr(name, '@') == NULL)
error = SET_ERROR(EINVAL);
if (error == 0)
error = nvpair_value_string(pair, &htag);
if (error == 0)
error = dsl_dataset_hold(dp, name, FTAG, &ds);
if (error == 0) {
error = dsl_dataset_user_hold_check_one(ds, htag,
dduha->dduha_minor != 0, tx);
dsl_dataset_rele(ds, FTAG);
}
if (error == 0) {
fnvlist_add_string(dduha->dduha_chkholds, name, htag);
} else {
/*
* We register ENOENT errors so they can be correctly
* reported if needed, such as when all holds fail.
*/
fnvlist_add_int32(dduha->dduha_errlist, name, error);
if (error != ENOENT)
return (error);
}
}
return (0);
}
void
spa_history_log_version(spa_t *spa, const char *operation)
{
#ifdef _KERNEL
uint64_t current_vers = spa_version(spa);
spa_history_log_internal(spa, operation, NULL,
"pool version %llu; software version %llu/%d; uts %s %s %s %s",
(u_longlong_t)current_vers, SPA_VERSION, ZPL_VERSION,
utsname.nodename, utsname.release, utsname.version,
utsname.machine);
cmn_err(CE_CONT, "!%s version %llu pool %s using %llu", operation,
(u_longlong_t)current_vers, spa_name(spa), SPA_VERSION);
#endif
}
/*
* Generate the pool's configuration based on the current in-core state.
* We infer whether to generate a complete config or just one top-level config
* based on whether vd is the root vdev.
*/
nvlist_t *
spa_config_generate(spa_t *spa, vdev_t *vd, uint64_t txg, int getstats)
{
nvlist_t *config, *nvroot;
vdev_t *rvd = spa->spa_root_vdev;
ASSERT(spa_config_held(spa, RW_READER));
if (vd == NULL)
vd = rvd;
/*
* If txg is -1, report the current value of spa->spa_config_txg.
*/
if (txg == -1ULL)
txg = spa->spa_config_txg;
VERIFY(nvlist_alloc(&config, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_VERSION,
spa_version(spa)) == 0);
VERIFY(nvlist_add_string(config, ZPOOL_CONFIG_POOL_NAME,
spa_name(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_STATE,
spa_state(spa)) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_TXG,
txg) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_POOL_GUID,
spa_guid(spa)) == 0);
if (vd != rvd) {
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_TOP_GUID,
vd->vdev_top->vdev_guid) == 0);
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_GUID,
vd->vdev_guid) == 0);
if (vd->vdev_isspare)
VERIFY(nvlist_add_uint64(config, ZPOOL_CONFIG_IS_SPARE,
1ULL) == 0);
vd = vd->vdev_top; /* label contains top config */
}
nvroot = vdev_config_generate(spa, vd, getstats, B_FALSE);
VERIFY(nvlist_add_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, nvroot) == 0);
nvlist_free(nvroot);
return (config);
}
static int
dsl_deleg_check(void *arg, dmu_tx_t *tx)
{
dsl_deleg_arg_t *dda = arg;
dsl_dir_t *dd;
int error;
if (spa_version(dmu_tx_pool(tx)->dp_spa) <
SPA_VERSION_DELEGATED_PERMS) {
return (SET_ERROR(ENOTSUP));
}
error = dsl_dir_hold(dmu_tx_pool(tx), dda->dda_name, FTAG, &dd, NULL);
if (error == 0)
dsl_dir_rele(dd, FTAG);
return (error);
}
void
spa_history_log_version(spa_t *spa, history_internal_events_t event)
{
#ifdef _KERNEL
uint64_t current_vers = spa_version(spa);
if (current_vers >= SPA_VERSION_ZPOOL_HISTORY) {
spa_history_log_internal(event, spa, NULL,
"pool spa %llu; zfs spa %llu; zpl %d; uts %s %s %s %s",
(u_longlong_t)current_vers, SPA_VERSION, ZPL_VERSION,
utsname.nodename, utsname.release, utsname.version,
utsname.machine);
}
cmn_err(CE_CONT, "!%s version %llu pool %s using %llu",
event == LOG_POOL_IMPORT ? "imported" :
event == LOG_POOL_CREATE ? "created" : "accessed",
(u_longlong_t)current_vers, spa_name(spa), SPA_VERSION);
#endif
}
/*
* set all create time permission on new dataset.
*/
void
dsl_deleg_set_create_perms(dsl_dir_t *sdd, dmu_tx_t *tx, cred_t *cr)
{
dsl_dir_t *dd;
uint64_t uid = crgetuid(cr);
if (spa_version(dmu_objset_spa(sdd->dd_pool->dp_meta_objset)) <
SPA_VERSION_DELEGATED_PERMS)
return;
for (dd = sdd->dd_parent; dd != NULL; dd = dd->dd_parent) {
uint64_t pzapobj = dd->dd_phys->dd_deleg_zapobj;
if (pzapobj == 0)
continue;
copy_create_perms(sdd, pzapobj, B_FALSE, uid, tx);
copy_create_perms(sdd, pzapobj, B_TRUE, uid, tx);
}
}
static int
dsl_dataset_user_hold_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_hold_arg_t *dduha = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvpair_t *pair;
int rv = 0;
if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS)
return (ENOTSUP);
for (pair = nvlist_next_nvpair(dduha->dduha_holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(dduha->dduha_holds, pair)) {
int error = 0;
dsl_dataset_t *ds;
char *htag;
/* must be a snapshot */
if (strchr(nvpair_name(pair), '@') == NULL)
error = EINVAL;
if (error == 0)
error = nvpair_value_string(pair, &htag);
if (error == 0) {
error = dsl_dataset_hold(dp,
nvpair_name(pair), FTAG, &ds);
}
if (error == 0) {
error = dsl_dataset_user_hold_check_one(ds, htag,
dduha->dduha_minor != 0, tx);
dsl_dataset_rele(ds, FTAG);
}
if (error != 0) {
rv = error;
fnvlist_add_int32(dduha->dduha_errlist,
nvpair_name(pair), error);
}
}
return (rv);
}
/*
* zfs_init_fs - Initialize the zfsvfs struct and the file system
* incore "master" object. Verify version compatibility.
*/
int
zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp, cred_t *cr)
{
extern int zfsfstype;
objset_t *os = zfsvfs->z_os;
int i, error;
dmu_object_info_t doi;
uint64_t fsid_guid;
uint64_t zval;
*zpp = NULL;
/*
* XXX - hack to auto-create the pool root filesystem at
* the first attempted mount.
*/
if (dmu_object_info(os, MASTER_NODE_OBJ, &doi) == ENOENT) {
dmu_tx_t *tx = dmu_tx_create(os);
uint64_t zpl_version;
nvlist_t *zprops;
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* master */
dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* del queue */
dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); /* root node */
error = dmu_tx_assign(tx, TXG_WAIT);
ASSERT3U(error, ==, 0);
if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
zpl_version = ZPL_VERSION;
else
zpl_version = ZPL_VERSION_FUID - 1;
VERIFY(nvlist_alloc(&zprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(zprops,
zfs_prop_to_name(ZFS_PROP_VERSION), zpl_version) == 0);
zfs_create_fs(os, cr, zprops, tx);
nvlist_free(zprops);
dmu_tx_commit(tx);
}
/* ARGSUSED */
static int
zcp_synctask_snapshot(lua_State *state, boolean_t sync, nvlist_t *err_details)
{
int err;
dsl_dataset_snapshot_arg_t ddsa = { 0 };
const char *dsname = lua_tostring(state, 1);
zcp_run_info_t *ri = zcp_run_info(state);
/*
* On old pools, the ZIL must not be active when a snapshot is created,
* but we can't suspend the ZIL because we're already in syncing
* context.
*/
if (spa_version(ri->zri_pool->dp_spa) < SPA_VERSION_FAST_SNAP) {
return (ENOTSUP);
}
/*
* We only allow for a single snapshot rather than a list, so the
* error list output is unnecessary.
*/
ddsa.ddsa_errors = NULL;
ddsa.ddsa_props = NULL;
ddsa.ddsa_cr = ri->zri_cred;
ddsa.ddsa_snaps = fnvlist_alloc();
fnvlist_add_boolean(ddsa.ddsa_snaps, dsname);
zcp_cleanup_handler_t *zch = zcp_register_cleanup(state,
(zcp_cleanup_t *)&fnvlist_free, ddsa.ddsa_snaps);
err = zcp_sync_task(state, dsl_dataset_snapshot_check,
dsl_dataset_snapshot_sync, &ddsa, sync, dsname);
zcp_deregister_cleanup(state, zch);
fnvlist_free(ddsa.ddsa_snaps);
return (err);
}
int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
int err = 0;
dmu_tx_t *tx;
nvlist_t *nvarg, *in_nvl = NULL;
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
return (SET_ERROR(EINVAL));
err = nvlist_lookup_nvlist(nvl, ZPOOL_HIST_INPUT_NVL, &in_nvl);
if (err == 0) {
(void) nvlist_remove_all(in_nvl, ZPOOL_HIDDEN_ARGS);
}
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP));
if (spa_history_zone() != NULL) {
fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
spa_history_zone());
}
fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
/* Kick this off asynchronously; errors are ignored. */
dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
dmu_tx_commit(tx);
/* spa_history_log_sync will free nvl */
return (err);
}
uint64_t
dsl_dir_create_sync(dsl_pool_t *dp, dsl_dir_t *pds, const char *name,
dmu_tx_t *tx)
{
objset_t *mos = dp->dp_meta_objset;
uint64_t ddobj;
dsl_dir_phys_t *ddphys;
dmu_buf_t *dbuf;
ddobj = dmu_object_alloc(mos, DMU_OT_DSL_DIR, 0,
DMU_OT_DSL_DIR, sizeof (dsl_dir_phys_t), tx);
if (pds) {
VERIFY(0 == zap_add(mos, pds->dd_phys->dd_child_dir_zapobj,
name, sizeof (uint64_t), 1, &ddobj, tx));
} else {
/* it's the root dir */
VERIFY(0 == zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1, &ddobj, tx));
}
VERIFY(0 == dmu_bonus_hold(mos, ddobj, FTAG, &dbuf));
dmu_buf_will_dirty(dbuf, tx);
ddphys = dbuf->db_data;
ddphys->dd_creation_time = gethrestime_sec();
if (pds)
ddphys->dd_parent_obj = pds->dd_object;
ddphys->dd_props_zapobj = zap_create(mos,
DMU_OT_DSL_PROPS, DMU_OT_NONE, 0, tx);
ddphys->dd_child_dir_zapobj = zap_create(mos,
DMU_OT_DSL_DIR_CHILD_MAP, DMU_OT_NONE, 0, tx);
if (spa_version(dp->dp_spa) >= SPA_VERSION_USED_BREAKDOWN)
ddphys->dd_flags |= DD_FLAG_USED_BREAKDOWN;
dmu_buf_rele(dbuf, FTAG);
return (ddobj);
}
/*
* Predict the effective value of the given special property if it were set with
* the given value and source. This is not a general purpose function. It exists
* only to handle the special requirements of the quota and reservation
* properties. The fact that these properties are non-inheritable greatly
* simplifies the prediction logic.
*
* Returns 0 on success, a positive error code on failure, or -1 if called with
* a property not handled by this function.
*/
int
dsl_prop_predict(dsl_dir_t *dd, const char *propname,
zprop_source_t source, uint64_t value, uint64_t *newvalp)
{
zfs_prop_t prop = zfs_name_to_prop(propname);
objset_t *mos;
uint64_t zapobj;
uint64_t version;
char *recvdstr;
int err = 0;
switch (prop) {
case ZFS_PROP_QUOTA:
case ZFS_PROP_RESERVATION:
case ZFS_PROP_REFQUOTA:
case ZFS_PROP_REFRESERVATION:
break;
default:
return (-1);
}
mos = dd->dd_pool->dp_meta_objset;
zapobj = dsl_dir_phys(dd)->dd_props_zapobj;
recvdstr = kmem_asprintf("%s%s", propname, ZPROP_RECVD_SUFFIX);
version = spa_version(dd->dd_pool->dp_spa);
if (version < SPA_VERSION_RECVD_PROPS) {
if (source & ZPROP_SRC_NONE)
source = ZPROP_SRC_NONE;
else if (source & ZPROP_SRC_RECEIVED)
source = ZPROP_SRC_LOCAL;
}
switch ((int)source) {
case ZPROP_SRC_NONE:
/* Revert to the received value, if any. */
err = zap_lookup(mos, zapobj, recvdstr, 8, 1, newvalp);
if (err == ENOENT)
*newvalp = 0;
break;
case ZPROP_SRC_LOCAL:
*newvalp = value;
break;
case ZPROP_SRC_RECEIVED:
/*
* If there's no local setting, then the new received value will
* be the effective value.
*/
err = zap_lookup(mos, zapobj, propname, 8, 1, newvalp);
if (err == ENOENT)
*newvalp = value;
break;
case (ZPROP_SRC_NONE | ZPROP_SRC_RECEIVED):
/*
* We're clearing the received value, so the local setting (if
* it exists) remains the effective value.
*/
err = zap_lookup(mos, zapobj, propname, 8, 1, newvalp);
if (err == ENOENT)
*newvalp = 0;
break;
default:
panic("unexpected property source: %d", source);
}
strfree(recvdstr);
if (err == ENOENT)
return (0);
return (err);
}
int
dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
objset_t **osp)
{
objset_t *os;
int i, err = 0;
ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
os = kmem_zalloc(sizeof (objset_t), KM_PUSHPAGE);
os->os_dsl_dataset = ds;
os->os_spa = spa;
os->os_rootbp = bp;
if (!BP_IS_HOLE(os->os_rootbp)) {
uint32_t aflags = ARC_WAIT;
zbookmark_t zb;
SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
if (DMU_OS_IS_L2CACHEABLE(os))
aflags |= ARC_L2CACHE;
dprintf_bp(os->os_rootbp, "reading %s", "");
/*
* XXX when bprewrite scrub can change the bp,
* and this is called from dmu_objset_open_ds_os, the bp
* could change, and we'll need a lock.
*/
err = dsl_read_nolock(NULL, spa, os->os_rootbp,
arc_getbuf_func, &os->os_phys_buf,
ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb);
if (err) {
kmem_free(os, sizeof (objset_t));
/* convert checksum errors into IO errors */
if (err == ECKSUM)
err = EIO;
return (err);
}
/* Increase the blocksize if we are permitted. */
if (spa_version(spa) >= SPA_VERSION_USERSPACE &&
arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) {
arc_buf_t *buf = arc_buf_alloc(spa,
sizeof (objset_phys_t), &os->os_phys_buf,
ARC_BUFC_METADATA);
bzero(buf->b_data, sizeof (objset_phys_t));
bcopy(os->os_phys_buf->b_data, buf->b_data,
arc_buf_size(os->os_phys_buf));
(void) arc_buf_remove_ref(os->os_phys_buf,
&os->os_phys_buf);
os->os_phys_buf = buf;
}
os->os_phys = os->os_phys_buf->b_data;
os->os_flags = os->os_phys->os_flags;
} else {
int size = spa_version(spa) >= SPA_VERSION_USERSPACE ?
sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE;
os->os_phys_buf = arc_buf_alloc(spa, size,
&os->os_phys_buf, ARC_BUFC_METADATA);
os->os_phys = os->os_phys_buf->b_data;
bzero(os->os_phys, size);
}
/*
* Note: the changed_cb will be called once before the register
* func returns, thus changing the checksum/compression from the
* default (fletcher2/off). Snapshots don't need to know about
* checksum/compression/copies. But they do need to know about
* encryption so that clones from the snaphost inherit the
* same encryption property regardless of where in the namespace
* they get created.
*/
if (ds) {
err = dsl_prop_register(ds, "primarycache",
primary_cache_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "secondarycache",
secondary_cache_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "encryption",
crypt_changed_cb, os);
if (!dsl_dataset_is_snapshot(ds)) {
if (err == 0)
err = dsl_prop_register(ds, "checksum",
checksum_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "compression",
compression_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "copies",
copies_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "dedup",
dedup_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "logbias",
logbias_changed_cb, os);
if (err == 0)
err = dsl_prop_register(ds, "sync",
sync_changed_cb, os);
}
if (err) {
VERIFY(arc_buf_remove_ref(os->os_phys_buf,
&os->os_phys_buf) == 1);
kmem_free(os, sizeof (objset_t));
return (err);
}
} else if (ds == NULL) {
/*
* It's the meta-objset.
* Encryption is off for ZFS metadata but on for ZPL metadata
* and file/zvol contents.
*/
os->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
os->os_compress = ZIO_COMPRESS_LZJB;
os->os_copies = spa_max_replication(spa);
os->os_dedup_checksum = ZIO_CHECKSUM_OFF;
os->os_dedup_verify = 0;
os->os_logbias = 0;
os->os_sync = 0;
os->os_primary_cache = ZFS_CACHE_ALL;
os->os_secondary_cache = ZFS_CACHE_ALL;
os->os_crypt = ZIO_CRYPT_OFF;
}
if (ds == NULL || !dsl_dataset_is_snapshot(ds))
os->os_zil_header = os->os_phys->os_zil_header;
os->os_zil = zil_alloc(os, &os->os_zil_header);
for (i = 0; i < TXG_SIZE; i++) {
list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
list_create(&os->os_free_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
}
list_create(&os->os_dnodes, sizeof (dnode_t),
offsetof(dnode_t, dn_link));
list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_link));
mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
DMU_META_DNODE(os) = dnode_special_open(os,
&os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT,
&os->os_meta_dnode);
if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) {
DMU_USERUSED_DNODE(os) = dnode_special_open(os,
&os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT,
&os->os_userused_dnode);
DMU_GROUPUSED_DNODE(os) = dnode_special_open(os,
&os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT,
&os->os_groupused_dnode);
}
/*
* We should be the only thread trying to do this because we
* have ds_opening_lock
*/
if (ds) {
mutex_enter(&ds->ds_lock);
ASSERT(ds->ds_objset == NULL);
ds->ds_objset = os;
mutex_exit(&ds->ds_lock);
}
*osp = os;
return (0);
}
dsl_pool_t *
dsl_pool_create(spa_t *spa, nvlist_t *zplprops, uint64_t txg)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
objset_t *os;
dsl_dataset_t *ds;
uint64_t obj;
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
/* create and open the MOS (meta-objset) */
dp->dp_meta_objset = dmu_objset_create_impl(spa,
NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx);
/* create the pool directory */
err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
ASSERT0(err);
/* Initialize scan structures */
VERIFY0(dsl_scan_init(dp, txg));
/* create and open the root dir */
dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
VERIFY0(dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
NULL, dp, &dp->dp_root_dir));
/* create and open the meta-objset dir */
(void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
MOS_DIR_NAME, &dp->dp_mos_dir));
if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
/* create and open the free dir */
(void) dsl_dir_create_sync(dp, dp->dp_root_dir,
FREE_DIR_NAME, tx);
VERIFY0(dsl_pool_open_special_dir(dp,
FREE_DIR_NAME, &dp->dp_free_dir));
/* create and open the free_bplist */
obj = bpobj_alloc(dp->dp_meta_objset, SPA_MAXBLOCKSIZE, tx);
VERIFY(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj, tx) == 0);
VERIFY0(bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
dsl_pool_create_origin(dp, tx);
/* create the root dataset */
obj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, 0, tx);
/* create the root objset */
VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG, &ds));
os = dmu_objset_create_impl(dp->dp_spa, ds,
dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
#ifdef _KERNEL
zfs_create_fs(os, kcred, zplprops, tx);
#endif
dsl_dataset_rele(ds, FTAG);
dmu_tx_commit(tx);
rrw_exit(&dp->dp_config_rwlock, FTAG);
return (dp);
}
int
dsl_pool_open(dsl_pool_t *dp)
{
int err;
dsl_dir_t *dd;
dsl_dataset_t *ds;
uint64_t obj;
rrw_enter(&dp->dp_config_rwlock, RW_WRITER, FTAG);
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
&dp->dp_root_dir_obj);
if (err)
goto out;
err = dsl_dir_hold_obj(dp, dp->dp_root_dir_obj,
NULL, dp, &dp->dp_root_dir);
if (err)
goto out;
err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
if (err)
goto out;
if (spa_version(dp->dp_spa) >= SPA_VERSION_ORIGIN) {
err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
if (err)
goto out;
err = dsl_dataset_hold_obj(dp, dd->dd_phys->dd_head_dataset_obj,
FTAG, &ds);
if (err == 0) {
err = dsl_dataset_hold_obj(dp,
ds->ds_phys->ds_prev_snap_obj, dp,
&dp->dp_origin_snap);
dsl_dataset_rele(ds, FTAG);
}
dsl_dir_rele(dd, dp);
if (err)
goto out;
}
if (spa_version(dp->dp_spa) >= SPA_VERSION_DEADLISTS) {
err = dsl_pool_open_special_dir(dp, FREE_DIR_NAME,
&dp->dp_free_dir);
if (err)
goto out;
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj);
if (err)
goto out;
VERIFY0(bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
if (spa_feature_is_active(dp->dp_spa,
&spa_feature_table[SPA_FEATURE_ASYNC_DESTROY])) {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_BPTREE_OBJ, sizeof (uint64_t), 1,
&dp->dp_bptree_obj);
if (err != 0)
goto out;
}
if (spa_feature_is_active(dp->dp_spa,
&spa_feature_table[SPA_FEATURE_EMPTY_BPOBJ])) {
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
&dp->dp_empty_bpobj);
if (err != 0)
goto out;
}
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_TMP_USERREFS, sizeof (uint64_t), 1,
&dp->dp_tmp_userrefs_obj);
if (err == ENOENT)
err = 0;
if (err)
goto out;
err = dsl_scan_init(dp, dp->dp_tx.tx_open_txg);
out:
rrw_exit(&dp->dp_config_rwlock, FTAG);
return (err);
}
static void
dsl_scan_setup_sync(void *arg, dmu_tx_t *tx)
{
dsl_scan_t *scn = dmu_tx_pool(tx)->dp_scan;
pool_scan_func_t *funcp = arg;
dmu_object_type_t ot = 0;
dsl_pool_t *dp = scn->scn_dp;
spa_t *spa = dp->dp_spa;
ASSERT(scn->scn_phys.scn_state != DSS_SCANNING);
ASSERT(*funcp > POOL_SCAN_NONE && *funcp < POOL_SCAN_FUNCS);
bzero(&scn->scn_phys, sizeof (scn->scn_phys));
scn->scn_phys.scn_func = *funcp;
scn->scn_phys.scn_state = DSS_SCANNING;
scn->scn_phys.scn_min_txg = 0;
scn->scn_phys.scn_max_txg = tx->tx_txg;
scn->scn_phys.scn_ddt_class_max = DDT_CLASSES - 1; /* the entire DDT */
scn->scn_phys.scn_start_time = gethrestime_sec();
scn->scn_phys.scn_errors = 0;
scn->scn_phys.scn_to_examine = spa->spa_root_vdev->vdev_stat.vs_alloc;
scn->scn_restart_txg = 0;
spa_scan_stat_init(spa);
if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
scn->scn_phys.scn_ddt_class_max = zfs_scrub_ddt_class_max;
/* rewrite all disk labels */
vdev_config_dirty(spa->spa_root_vdev);
if (vdev_resilver_needed(spa->spa_root_vdev,
&scn->scn_phys.scn_min_txg, &scn->scn_phys.scn_max_txg)) {
spa_event_notify(spa, NULL, ESC_ZFS_RESILVER_START);
} else {
spa_event_notify(spa, NULL, ESC_ZFS_SCRUB_START);
}
spa->spa_scrub_started = B_TRUE;
/*
* If this is an incremental scrub, limit the DDT scrub phase
* to just the auto-ditto class (for correctness); the rest
* of the scrub should go faster using top-down pruning.
*/
if (scn->scn_phys.scn_min_txg > TXG_INITIAL)
scn->scn_phys.scn_ddt_class_max = DDT_CLASS_DITTO;
}
/* back to the generic stuff */
if (dp->dp_blkstats == NULL) {
dp->dp_blkstats =
kmem_alloc(sizeof (zfs_all_blkstats_t), KM_SLEEP);
}
bzero(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
if (spa_version(spa) < SPA_VERSION_DSL_SCRUB)
ot = DMU_OT_ZAP_OTHER;
scn->scn_phys.scn_queue_obj = zap_create(dp->dp_meta_objset,
ot ? ot : DMU_OT_SCAN_QUEUE, DMU_OT_NONE, 0, tx);
dsl_scan_sync_state(scn, tx);
spa_history_log_internal(spa, "scan setup", tx,
"func=%u mintxg=%llu maxtxg=%llu",
*funcp, scn->scn_phys.scn_min_txg, scn->scn_phys.scn_max_txg);
}
int
dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
objset_t **osp)
{
objset_t *os;
int i, err;
ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
os = kmem_zalloc(sizeof (objset_t), KM_SLEEP);
os->os_dsl_dataset = ds;
os->os_spa = spa;
os->os_rootbp = bp;
if (!BP_IS_HOLE(os->os_rootbp)) {
arc_flags_t aflags = ARC_FLAG_WAIT;
zbookmark_phys_t zb;
SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
if (DMU_OS_IS_L2CACHEABLE(os))
aflags |= ARC_FLAG_L2CACHE;
if (DMU_OS_IS_L2COMPRESSIBLE(os))
aflags |= ARC_FLAG_L2COMPRESS;
dprintf_bp(os->os_rootbp, "reading %s", "");
err = arc_read(NULL, spa, os->os_rootbp,
arc_getbuf_func, &os->os_phys_buf,
ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb);
if (err != 0) {
kmem_free(os, sizeof (objset_t));
/* convert checksum errors into IO errors */
if (err == ECKSUM)
err = SET_ERROR(EIO);
return (err);
}
/* Increase the blocksize if we are permitted. */
if (spa_version(spa) >= SPA_VERSION_USERSPACE &&
arc_buf_size(os->os_phys_buf) < sizeof (objset_phys_t)) {
arc_buf_t *buf = arc_buf_alloc(spa,
sizeof (objset_phys_t), &os->os_phys_buf,
ARC_BUFC_METADATA);
bzero(buf->b_data, sizeof (objset_phys_t));
bcopy(os->os_phys_buf->b_data, buf->b_data,
arc_buf_size(os->os_phys_buf));
(void) arc_buf_remove_ref(os->os_phys_buf,
&os->os_phys_buf);
os->os_phys_buf = buf;
}
os->os_phys = os->os_phys_buf->b_data;
os->os_flags = os->os_phys->os_flags;
} else {
int size = spa_version(spa) >= SPA_VERSION_USERSPACE ?
sizeof (objset_phys_t) : OBJSET_OLD_PHYS_SIZE;
os->os_phys_buf = arc_buf_alloc(spa, size,
&os->os_phys_buf, ARC_BUFC_METADATA);
os->os_phys = os->os_phys_buf->b_data;
bzero(os->os_phys, size);
}
/*
* Note: the changed_cb will be called once before the register
* func returns, thus changing the checksum/compression from the
* default (fletcher2/off). Snapshots don't need to know about
* checksum/compression/copies.
*/
if (ds != NULL) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
primary_cache_changed_cb, os);
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE),
secondary_cache_changed_cb, os);
}
if (!ds->ds_is_snapshot) {
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_CHECKSUM),
checksum_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
compression_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_COPIES),
copies_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_DEDUP),
dedup_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_LOGBIAS),
logbias_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SYNC),
sync_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(
ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
recordsize_changed_cb, os);
}
}
if (err != 0) {
VERIFY(arc_buf_remove_ref(os->os_phys_buf,
&os->os_phys_buf));
kmem_free(os, sizeof (objset_t));
return (err);
}
} else {
/* It's the meta-objset. */
os->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
os->os_compress = ZIO_COMPRESS_ON;
os->os_copies = spa_max_replication(spa);
os->os_dedup_checksum = ZIO_CHECKSUM_OFF;
os->os_dedup_verify = B_FALSE;
os->os_logbias = ZFS_LOGBIAS_LATENCY;
os->os_sync = ZFS_SYNC_STANDARD;
os->os_primary_cache = ZFS_CACHE_ALL;
os->os_secondary_cache = ZFS_CACHE_ALL;
}
if (ds == NULL || !ds->ds_is_snapshot)
os->os_zil_header = os->os_phys->os_zil_header;
os->os_zil = zil_alloc(os, &os->os_zil_header);
for (i = 0; i < TXG_SIZE; i++) {
list_create(&os->os_dirty_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
list_create(&os->os_free_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
}
list_create(&os->os_dnodes, sizeof (dnode_t),
offsetof(dnode_t, dn_link));
list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_link));
mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
dnode_special_open(os, &os->os_phys->os_meta_dnode,
DMU_META_DNODE_OBJECT, &os->os_meta_dnode);
if (arc_buf_size(os->os_phys_buf) >= sizeof (objset_phys_t)) {
dnode_special_open(os, &os->os_phys->os_userused_dnode,
DMU_USERUSED_OBJECT, &os->os_userused_dnode);
dnode_special_open(os, &os->os_phys->os_groupused_dnode,
DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode);
}
*osp = os;
return (0);
}