/* 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_userspace_upgrade(objset_t *os)
{
uint64_t obj;
int err = 0;
if (dmu_objset_userspace_present(os))
return (0);
if (!dmu_objset_userused_enabled(os->os))
return (ENOTSUP);
if (dmu_objset_is_snapshot(os))
return (EINVAL);
/*
* We simply need to mark every object dirty, so that it will be
* synced out and now accounted. If this is called
* concurrently, or if we already did some work before crashing,
* that's fine, since we track each object's accounted state
* independently.
*/
for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) {
dmu_tx_t *tx;
dmu_buf_t *db;
int objerr;
if (issig(JUSTLOOKING) && issig(FORREAL))
return (EINTR);
objerr = dmu_bonus_hold(os, obj, FTAG, &db);
if (objerr)
continue;
tx = dmu_tx_create(os);
dmu_tx_hold_bonus(tx, obj);
objerr = dmu_tx_assign(tx, TXG_WAIT);
if (objerr) {
dmu_tx_abort(tx);
continue;
}
dmu_buf_will_dirty(db, tx);
dmu_buf_rele(db, FTAG);
dmu_tx_commit(tx);
}
os->os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE;
txg_wait_synced(dmu_objset_pool(os), 0);
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);
}
/* called from dsl */
void
dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_t zb;
zio_prop_t zp;
zio_t *zio;
list_t *list;
list_t *newlist = NULL;
dbuf_dirty_record_t *dr;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
SET_BOOKMARK(&zb, os->os_dsl_dataset ?
os->os_dsl_dataset->ds_object : DMU_META_OBJSET,
ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
arc_release(os->os_phys_buf, &os->os_phys_buf);
dmu_write_policy(os, NULL, 0, 0, &zp);
zio = arc_write(pio, os->os_spa, tx->tx_txg,
os->os_rootbp, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), &zp,
dmu_objset_write_ready, dmu_objset_write_done, os,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
/*
* Sync special dnodes - the parent IO for the sync is the root block
*/
DMU_META_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_META_DNODE(os), tx);
os->os_phys->os_flags = os->os_flags;
if (DMU_USERUSED_DNODE(os) &&
DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) {
DMU_USERUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_USERUSED_DNODE(os), tx);
DMU_GROUPUSED_DNODE(os)->dn_zio = zio;
dnode_sync(DMU_GROUPUSED_DNODE(os), tx);
}
txgoff = tx->tx_txg & TXG_MASK;
if (dmu_objset_userused_enabled(os)) {
newlist = &os->os_synced_dnodes;
/*
* We must create the list here because it uses the
* dn_dirty_link[] of this txg.
*/
list_create(newlist, sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[txgoff]));
}
dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx);
dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx);
list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff];
while (dr = list_head(list)) {
ASSERT0(dr->dr_dbuf->db_level);
list_remove(list, dr);
if (dr->dr_zio)
zio_nowait(dr->dr_zio);
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
/* called from dsl */
void
dmu_objset_sync(objset_impl_t *os, zio_t *pio, dmu_tx_t *tx)
{
int txgoff;
zbookmark_t zb;
writeprops_t wp = { 0 };
zio_t *zio;
list_t *list;
list_t *newlist = NULL;
dbuf_dirty_record_t *dr;
dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg);
ASSERT(dmu_tx_is_syncing(tx));
/* XXX the write_done callback should really give us the tx... */
os->os_synctx = tx;
if (os->os_dsl_dataset == NULL) {
/*
* This is the MOS. If we have upgraded,
* spa_max_replication() could change, so reset
* os_copies here.
*/
os->os_copies = spa_max_replication(os->os_spa);
}
/*
* Create the root block IO
*/
zb.zb_objset = os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : 0;
zb.zb_object = 0;
zb.zb_level = -1; /* for block ordering; it's level 0 on disk */
zb.zb_blkid = 0;
wp.wp_type = DMU_OT_OBJSET;
wp.wp_level = 0; /* on-disk BP level; see above */
wp.wp_copies = os->os_copies;
wp.wp_oschecksum = os->os_checksum;
wp.wp_oscompress = os->os_compress;
if (BP_IS_OLDER(os->os_rootbp, tx->tx_txg)) {
(void) dsl_dataset_block_kill(os->os_dsl_dataset,
os->os_rootbp, pio, tx);
}
arc_release(os->os_phys_buf, &os->os_phys_buf);
zio = arc_write(pio, os->os_spa, &wp, DMU_OS_IS_L2CACHEABLE(os),
tx->tx_txg, os->os_rootbp, os->os_phys_buf, ready, NULL, os,
ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);
/*
* Sync special dnodes - the parent IO for the sync is the root block
*/
os->os_meta_dnode->dn_zio = zio;
dnode_sync(os->os_meta_dnode, tx);
os->os_phys->os_flags = os->os_flags;
if (os->os_userused_dnode &&
os->os_userused_dnode->dn_type != DMU_OT_NONE) {
os->os_userused_dnode->dn_zio = zio;
dnode_sync(os->os_userused_dnode, tx);
os->os_groupused_dnode->dn_zio = zio;
dnode_sync(os->os_groupused_dnode, tx);
}
txgoff = tx->tx_txg & TXG_MASK;
if (dmu_objset_userused_enabled(os)) {
newlist = &os->os_synced_dnodes;
/*
* We must create the list here because it uses the
* dn_dirty_link[] of this txg.
*/
list_create(newlist, sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[txgoff]));
}
dmu_objset_sync_dnodes(&os->os_free_dnodes[txgoff], newlist, tx);
dmu_objset_sync_dnodes(&os->os_dirty_dnodes[txgoff], newlist, tx);
list = &os->os_meta_dnode->dn_dirty_records[txgoff];
while ((dr = list_head(list))) {
ASSERT(dr->dr_dbuf->db_level == 0);
list_remove(list, dr);
if (dr->dr_zio)
zio_nowait(dr->dr_zio);
}
/*
* Free intent log blocks up to this tx.
*/
zil_sync(os->os_zil, tx);
os->os_phys->os_zil_header = os->os_zil_header;
zio_nowait(zio);
}
void
dmu_objset_do_userquota_callbacks(objset_impl_t *os, dmu_tx_t *tx)
{
dnode_t *dn;
list_t *list = &os->os_synced_dnodes;
ASSERTV(static const char zerobuf[DN_MAX_BONUSLEN] = {0});
ASSERT(list_head(list) == NULL || dmu_objset_userused_enabled(os));
while ((dn = list_head(list))) {
dmu_object_type_t bonustype;
ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object));
ASSERT(dn->dn_oldphys);
ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE ||
dn->dn_phys->dn_flags &
DNODE_FLAG_USERUSED_ACCOUNTED);
/* Allocate the user/groupused objects if necessary. */
if (os->os_userused_dnode->dn_type == DMU_OT_NONE) {
VERIFY(0 == zap_create_claim(&os->os,
DMU_USERUSED_OBJECT,
DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
VERIFY(0 == zap_create_claim(&os->os,
DMU_GROUPUSED_OBJECT,
DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx));
}
/*
* If the object was not previously
* accounted, pretend that it was free.
*/
if (!(dn->dn_oldphys->dn_flags &
DNODE_FLAG_USERUSED_ACCOUNTED)) {
bzero(dn->dn_oldphys, sizeof (dnode_phys_t));
}
/*
* If the object was freed, use the previous bonustype.
*/
bonustype = dn->dn_phys->dn_bonustype ?
dn->dn_phys->dn_bonustype : dn->dn_oldphys->dn_bonustype;
ASSERT(dn->dn_phys->dn_type != 0 ||
(bcmp(DN_BONUS(dn->dn_phys), zerobuf,
DN_MAX_BONUSLEN) == 0 &&
DN_USED_BYTES(dn->dn_phys) == 0));
ASSERT(dn->dn_oldphys->dn_type != 0 ||
(bcmp(DN_BONUS(dn->dn_oldphys), zerobuf,
DN_MAX_BONUSLEN) == 0 &&
DN_USED_BYTES(dn->dn_oldphys) == 0));
used_cbs[os->os_phys->os_type](&os->os, bonustype,
DN_BONUS(dn->dn_oldphys), DN_BONUS(dn->dn_phys),
DN_USED_BYTES(dn->dn_oldphys),
DN_USED_BYTES(dn->dn_phys), tx);
/*
* The mutex is needed here for interlock with dnode_allocate.
*/
mutex_enter(&dn->dn_mtx);
zio_buf_free(dn->dn_oldphys, sizeof (dnode_phys_t));
dn->dn_oldphys = NULL;
mutex_exit(&dn->dn_mtx);
list_remove(list, dn);
dnode_rele(dn, list);
}
}
