static int
dmu_objset_open_ds_os(dsl_dataset_t *ds, objset_t *os, dmu_objset_type_t type)
{
objset_impl_t *osi;
mutex_enter(&ds->ds_opening_lock);
osi = dsl_dataset_get_user_ptr(ds);
if (osi == NULL) {
int err;
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, &ds->ds_phys->ds_bp, &osi);
if (err) {
mutex_exit(&ds->ds_opening_lock);
return (err);
}
}
mutex_exit(&ds->ds_opening_lock);
os->os = osi;
os->os_mode = DS_MODE_NOHOLD;
if (type != DMU_OST_ANY && type != os->os->os_phys->os_type)
return (EINVAL);
return (0);
}
int
dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
dsl_dir_t *dd;
dsl_dataset_t *ds;
uint64_t obj;
rw_enter(&dp->dp_config_rwlock, RW_WRITER);
err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
&dp->dp_meta_objset);
if (err)
goto out;
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_open_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(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_close(dd, dp);
if (err)
goto out;
}
if (spa_version(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;
VERIFY3U(0, ==, bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
int
dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp)
{
int err = 0;
/*
* We shouldn't be doing anything with dsl_dataset_t's unless the
* pool_config lock is held, or the dataset is long-held.
*/
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) ||
dsl_dataset_long_held(ds));
mutex_enter(&ds->ds_opening_lock);
if (ds->ds_objset == NULL) {
objset_t *os;
rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG);
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, dsl_dataset_get_blkptr(ds), &os);
rrw_exit(&ds->ds_bp_rwlock, FTAG);
if (err == 0) {
mutex_enter(&ds->ds_lock);
ASSERT(ds->ds_objset == NULL);
ds->ds_objset = os;
mutex_exit(&ds->ds_lock);
}
}
*osp = ds->ds_objset;
mutex_exit(&ds->ds_opening_lock);
return (err);
}
/* called from dsl for meta-objset */
objset_t *
dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
dmu_objset_type_t type, dmu_tx_t *tx)
{
objset_t *os;
dnode_t *mdn;
ASSERT(dmu_tx_is_syncing(tx));
if (ds != NULL)
VERIFY0(dmu_objset_from_ds(ds, &os));
else
VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os));
mdn = DMU_META_DNODE(os);
dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT,
DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx);
/*
* We don't want to have to increase the meta-dnode's nlevels
* later, because then we could do it in quescing context while
* we are also accessing it in open context.
*
* This precaution is not necessary for the MOS (ds == NULL),
* because the MOS is only updated in syncing context.
* This is most fortunate: the MOS is the only objset that
* needs to be synced multiple times as spa_sync() iterates
* to convergence, so minimizing its dn_nlevels matters.
*/
if (ds != NULL) {
int levels = 1;
/*
* Determine the number of levels necessary for the meta-dnode
* to contain DN_MAX_OBJECT dnodes.
*/
while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift +
(levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
DN_MAX_OBJECT * sizeof (dnode_phys_t))
levels++;
mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =
mdn->dn_nlevels = levels;
}
ASSERT(type != DMU_OST_NONE);
ASSERT(type != DMU_OST_ANY);
ASSERT(type < DMU_OST_NUMTYPES);
os->os_phys->os_type = type;
if (dmu_objset_userused_enabled(os)) {
os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
os->os_flags = os->os_phys->os_flags;
}
dsl_dataset_dirty(ds, tx);
return (os);
}
int
dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp)
{
int err = 0;
mutex_enter(&ds->ds_opening_lock);
*osp = ds->ds_objset;
if (*osp == NULL) {
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, dsl_dataset_get_blkptr(ds), osp);
}
mutex_exit(&ds->ds_opening_lock);
return (err);
}
int
dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
&dp->dp_meta_objset);
if (err != 0)
dsl_pool_close(dp);
else
*dpp = dp;
return (err);
}
int
dmu_objset_from_ds_NEW(dsl_dataset_t *ds, objset_t **osp)
{
mutex_enter(&ds->ds_opening_lock);
if (ds->ds_objset == NULL) {
int err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, dsl_dataset_get_blkptr(ds), osp);
if (err) {
mutex_exit(&ds->ds_opening_lock);
return (err);
}
ASSERT(ds->ds_objset == *osp);
}
*osp = ds->ds_objset;
mutex_exit(&ds->ds_opening_lock);
return (0);
}
int
dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp)
{
int err = 0;
mutex_enter(&ds->ds_opening_lock);
if (ds->ds_objset == NULL) {
objset_t *os;
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, dsl_dataset_get_blkptr(ds), &os);
if (err == 0) {
mutex_enter(&ds->ds_lock);
ASSERT(ds->ds_objset == NULL);
ds->ds_objset = os;
mutex_exit(&ds->ds_lock);
}
}
*osp = ds->ds_objset;
mutex_exit(&ds->ds_opening_lock);
return (err);
}
int
dsl_pool_init(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
/*
* Initialize the caller's dsl_pool_t structure before we actually open
* the meta objset. This is done because a self-healing write zio may
* be issued as part of dmu_objset_open_impl() and the spa needs its
* dsl_pool_t initialized in order to handle the write.
*/
*dpp = dp;
err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
&dp->dp_meta_objset);
if (err != 0) {
dsl_pool_close(dp);
*dpp = NULL;
}
return (err);
}
/* called from zpl */
int
dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
objset_t **osp)
{
dsl_dataset_t *ds;
int err;
objset_t *os;
objset_impl_t *osi;
os = kmem_alloc(sizeof (objset_t), KM_SLEEP);
err = dsl_dataset_open(name, mode, os, &ds);
if (err) {
kmem_free(os, sizeof (objset_t));
return (err);
}
osi = dsl_dataset_get_user_ptr(ds);
if (osi == NULL) {
err = dmu_objset_open_impl(dsl_dataset_get_spa(ds),
ds, &ds->ds_phys->ds_bp, &osi);
if (err) {
dsl_dataset_close(ds, mode, os);
kmem_free(os, sizeof (objset_t));
return (err);
}
}
os->os = osi;
os->os_mode = mode;
if (type != DMU_OST_ANY && type != os->os->os_phys->os_type) {
dmu_objset_close(os);
return (EINVAL);
}
*osp = os;
return (0);
}
/* called from dsl for meta-objset */
objset_t *
dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
dmu_objset_type_t type, dsl_crypto_ctx_t *dcc, dmu_tx_t *tx)
{
objset_t *os;
dnode_t *mdn;
ASSERT(dmu_tx_is_syncing(tx));
if (ds != NULL)
VERIFY(0 == dmu_objset_from_ds(ds, &os));
else
VERIFY(0 == dmu_objset_open_impl(spa, NULL, bp, &os));
mdn = DMU_META_DNODE(os);
dnode_allocate(mdn, DMU_OT_DNODE, 1 << DNODE_BLOCK_SHIFT,
DN_MAX_INDBLKSHIFT, DMU_OT_NONE, 0, tx);
/*
* We don't want to have to increase the meta-dnode's nlevels
* later, because then we could do it in quescing context while
* we are also accessing it in open context.
*
* This precaution is not necessary for the MOS (ds == NULL),
* because the MOS is only updated in syncing context.
* This is most fortunate: the MOS is the only objset that
* needs to be synced multiple times as spa_sync() iterates
* to convergence, so minimizing its dn_nlevels matters.
*/
if (ds != NULL) {
int levels = 1;
/*
* Determine the number of levels necessary for the meta-dnode
* to contain DN_MAX_OBJECT dnodes.
*/
while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift +
(levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
DN_MAX_OBJECT * sizeof (dnode_phys_t))
levels++;
mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =
mdn->dn_nlevels = levels;
}
ASSERT(type != DMU_OST_NONE);
ASSERT(type != DMU_OST_ANY);
ASSERT(type < DMU_OST_NUMTYPES);
/*
* Note: although we should not be dirtying the objset outside of
* sync context, the os_type is used directly from the os_phys
* in dsl_scan so it may not be save to put os_type in the in-core
* objset_t and set it in the phys in sync context (as with os_flags)
*/
dmu_objset_dirty(os);
os->os_phys->os_type = type;
if (dmu_objset_userused_enabled(os)) {
os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
// FIXME - removed in newer.
// os->os_flags = os->os_phys->os_flags;
}
if (dcc != NULL && dcc->dcc_crypt != ZIO_CRYPT_INHERIT) {
os->os_crypt = zio_crypt_select(dcc->dcc_crypt,
ZIO_CRYPT_ON_VALUE);
}
dsl_dataset_dirty(ds, tx);
return (os);
}