int
zfs_set_userquota(zfs_sb_t *zsb, zfs_userquota_prop_t type,
const char *domain, uint64_t rid, uint64_t quota)
{
char buf[32];
int err;
dmu_tx_t *tx;
uint64_t *objp;
boolean_t fuid_dirtied;
if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA)
return (SET_ERROR(EINVAL));
if (zsb->z_version < ZPL_VERSION_USERSPACE)
return (SET_ERROR(ENOTSUP));
objp = (type == ZFS_PROP_USERQUOTA) ? &zsb->z_userquota_obj :
&zsb->z_groupquota_obj;
err = id_to_fuidstr(zsb, domain, rid, buf, B_TRUE);
if (err)
return (err);
fuid_dirtied = zsb->z_fuid_dirty;
tx = dmu_tx_create(zsb->z_os);
dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL);
if (*objp == 0) {
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE,
zfs_userquota_prop_prefixes[type]);
}
if (fuid_dirtied)
zfs_fuid_txhold(zsb, tx);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
mutex_enter(&zsb->z_lock);
if (*objp == 0) {
*objp = zap_create(zsb->z_os, DMU_OT_USERGROUP_QUOTA,
DMU_OT_NONE, 0, tx);
VERIFY(0 == zap_add(zsb->z_os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[type], 8, 1, objp, tx));
}
mutex_exit(&zsb->z_lock);
if (quota == 0) {
err = zap_remove(zsb->z_os, *objp, buf, tx);
if (err == ENOENT)
err = 0;
} else {
err = zap_update(zsb->z_os, *objp, buf, 8, 1, "a, tx);
}
ASSERT(err == 0);
if (fuid_dirtied)
zfs_fuid_sync(zsb, tx);
dmu_tx_commit(tx);
return (err);
}
/*
* ZFS_IOC_CREATE callback handles dmu zvol and zap object creation.
*/
void
zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx)
{
zfs_creat_t *zct = arg;
nvlist_t *nvprops = zct->zct_props;
int error;
uint64_t volblocksize, volsize;
VERIFY(nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0);
if (nvlist_lookup_uint64(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0)
volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
/*
* These properties must be removed from the list so the generic
* property setting step won't apply to them.
*/
VERIFY(nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0);
(void) nvlist_remove_all(nvprops,
zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE));
error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx);
ASSERT(error == 0);
}
static void
sa_attr_register_sync(sa_handle_t *hdl, dmu_tx_t *tx)
{
uint64_t attr_value = 0;
sa_os_t *sa = hdl->sa_os->os_sa;
sa_attr_table_t *tb = sa->sa_attr_table;
int i;
mutex_enter(&sa->sa_lock);
if (!sa->sa_need_attr_registration || sa->sa_master_obj == NULL) {
mutex_exit(&sa->sa_lock);
return;
}
if (sa->sa_reg_attr_obj == NULL) {
sa->sa_reg_attr_obj = zap_create(hdl->sa_os,
DMU_OT_SA_ATTR_REGISTRATION, DMU_OT_NONE, 0, tx);
VERIFY(zap_add(hdl->sa_os, sa->sa_master_obj,
SA_REGISTRY, 8, 1, &sa->sa_reg_attr_obj, tx) == 0);
}
for (i = 0; i != sa->sa_num_attrs; i++) {
if (sa->sa_attr_table[i].sa_registered)
continue;
ATTR_ENCODE(attr_value, tb[i].sa_attr, tb[i].sa_length,
tb[i].sa_byteswap);
VERIFY(0 == zap_update(hdl->sa_os, sa->sa_reg_attr_obj,
tb[i].sa_name, 8, 1, &attr_value, tx));
tb[i].sa_registered = B_TRUE;
}
sa->sa_need_attr_registration = B_FALSE;
mutex_exit(&sa->sa_lock);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(uint64_t volsize, objset_t *os)
{
dmu_tx_t *tx;
int error;
uint64_t txg;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_mark_netfree(tx);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (SET_ERROR(error));
}
txg = dmu_tx_get_txg(tx);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
txg_wait_synced(dmu_objset_pool(os), txg);
if (error == 0)
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
return (error);
}
/*
* This function is non-static for zhack; it should otherwise not be used
* outside this file.
*/
void
feature_sync(spa_t *spa, zfeature_info_t *feature, uint64_t refcount,
dmu_tx_t *tx)
{
uint64_t zapobj;
ASSERT(VALID_FEATURE_OR_NONE(feature->fi_feature));
zapobj = (feature->fi_flags & ZFEATURE_FLAG_READONLY_COMPAT) ?
spa->spa_feat_for_write_obj : spa->spa_feat_for_read_obj;
VERIFY0(zap_update(spa->spa_meta_objset, zapobj, feature->fi_guid,
sizeof (uint64_t), 1, &refcount, tx));
/*
* feature_sync is called directly from zhack, allowing the
* creation of arbitrary features whose fi_feature field may
* be greater than SPA_FEATURES. When called from zhack, the
* zfeature_info_t object's fi_feature field will be set to
* SPA_FEATURE_NONE.
*/
if (feature->fi_feature != SPA_FEATURE_NONE) {
uint64_t *refcount_cache =
&spa->spa_feat_refcount_cache[feature->fi_feature];
VERIFY3U(*refcount_cache, ==,
atomic_swap_64(refcount_cache, refcount));
}
static void
dsl_scan_sync_state(dsl_scan_t *scn, dmu_tx_t *tx)
{
VERIFY0(zap_update(scn->scn_dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCAN, sizeof (uint64_t), SCAN_PHYS_NUMINTS,
&scn->scn_phys, tx));
}
static sa_lot_t *
sa_add_layout_entry(objset_t *os, sa_attr_type_t *attrs, int attr_count,
uint64_t lot_num, uint64_t hash, boolean_t zapadd, dmu_tx_t *tx)
{
sa_os_t *sa = os->os_sa;
sa_lot_t *tb, *findtb;
int i;
avl_index_t loc;
ASSERT(MUTEX_HELD(&sa->sa_lock));
tb = kmem_zalloc(sizeof (sa_lot_t), KM_SLEEP);
tb->lot_attr_count = attr_count;
tb->lot_attrs = kmem_alloc(sizeof (sa_attr_type_t) * attr_count,
KM_SLEEP);
bcopy(attrs, tb->lot_attrs, sizeof (sa_attr_type_t) * attr_count);
tb->lot_num = lot_num;
tb->lot_hash = hash;
tb->lot_instance = 0;
if (zapadd) {
char attr_name[8];
if (sa->sa_layout_attr_obj == 0) {
sa->sa_layout_attr_obj = zap_create(os,
DMU_OT_SA_ATTR_LAYOUTS, DMU_OT_NONE, 0, tx);
VERIFY(zap_add(os, sa->sa_master_obj, SA_LAYOUTS, 8, 1,
&sa->sa_layout_attr_obj, tx) == 0);
}
(void) snprintf(attr_name, sizeof (attr_name),
"%d", (int)lot_num);
VERIFY(0 == zap_update(os, os->os_sa->sa_layout_attr_obj,
attr_name, 2, attr_count, attrs, tx));
}
list_create(&tb->lot_idx_tab, sizeof (sa_idx_tab_t),
offsetof(sa_idx_tab_t, sa_next));
for (i = 0; i != attr_count; i++) {
if (sa->sa_attr_table[tb->lot_attrs[i]].sa_length == 0)
tb->lot_var_sizes++;
}
avl_add(&sa->sa_layout_num_tree, tb);
/* verify we don't have a hash collision */
if ((findtb = avl_find(&sa->sa_layout_hash_tree, tb, &loc)) != NULL) {
for (; findtb && findtb->lot_hash == hash;
findtb = AVL_NEXT(&sa->sa_layout_hash_tree, findtb)) {
if (findtb->lot_instance != tb->lot_instance)
break;
tb->lot_instance++;
}
}
avl_add(&sa->sa_layout_hash_tree, tb);
return (tb);
}
static void
vdev_initialize_zap_update_sync(void *arg, dmu_tx_t *tx)
{
/*
* We pass in the guid instead of the vdev_t since the vdev may
* have been freed prior to the sync task being processed. This
* happens when a vdev is detached as we call spa_config_vdev_exit(),
* stop the initializing thread, schedule the sync task, and free
* the vdev. Later when the scheduled sync task is invoked, it would
* find that the vdev has been freed.
*/
uint64_t guid = *(uint64_t *)arg;
uint64_t txg = dmu_tx_get_txg(tx);
kmem_free(arg, sizeof (uint64_t));
vdev_t *vd = spa_lookup_by_guid(tx->tx_pool->dp_spa, guid, B_FALSE);
if (vd == NULL || vd->vdev_top->vdev_removing || !vdev_is_concrete(vd))
return;
uint64_t last_offset = vd->vdev_initialize_offset[txg & TXG_MASK];
vd->vdev_initialize_offset[txg & TXG_MASK] = 0;
VERIFY(vd->vdev_leaf_zap != 0);
objset_t *mos = vd->vdev_spa->spa_meta_objset;
if (last_offset > 0) {
vd->vdev_initialize_last_offset = last_offset;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_LAST_OFFSET,
sizeof (last_offset), 1, &last_offset, tx));
}
if (vd->vdev_initialize_action_time > 0) {
uint64_t val = (uint64_t)vd->vdev_initialize_action_time;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_ACTION_TIME, sizeof (val),
1, &val, tx));
}
uint64_t initialize_state = vd->vdev_initialize_state;
VERIFY0(zap_update(mos, vd->vdev_leaf_zap,
VDEV_LEAF_ZAP_INITIALIZE_STATE, sizeof (initialize_state), 1,
&initialize_state, tx));
}
static void
dsl_deleg_set_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
nvlist_t *nvp = arg2;
objset_t *mos = dd->dd_pool->dp_meta_objset;
nvpair_t *whopair = NULL;
uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
if (zapobj == 0) {
dmu_buf_will_dirty(dd->dd_dbuf, tx);
zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
}
while (whopair = nvlist_next_nvpair(nvp, whopair)) {
const char *whokey = nvpair_name(whopair);
nvlist_t *perms;
nvpair_t *permpair = NULL;
uint64_t jumpobj;
VERIFY(nvpair_value_nvlist(whopair, &perms) == 0);
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0) {
jumpobj = zap_create(mos, DMU_OT_DSL_PERMS,
DMU_OT_NONE, 0, tx);
VERIFY(zap_update(mos, zapobj,
whokey, 8, 1, &jumpobj, tx) == 0);
}
while (permpair = nvlist_next_nvpair(perms, permpair)) {
const char *perm = nvpair_name(permpair);
uint64_t n = 0;
VERIFY(zap_update(mos, jumpobj,
perm, 8, 1, &n, tx) == 0);
spa_history_internal_log(LOG_DS_PERM_UPDATE,
dd->dd_pool->dp_spa, tx, cr,
"%s %s dataset = %llu", whokey, perm,
dd->dd_phys->dd_head_dataset_obj);
}
}
}
int
zvol_set_volsize(zfs_cmd_t *zc)
{
zvol_state_t *zv;
dev_t dev = zc->zc_dev;
dmu_tx_t *tx;
int error;
dmu_object_info_t doi;
mutex_enter(&zvol_state_lock);
if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) {
mutex_exit(&zvol_state_lock);
return (ENXIO);
}
if ((error = dmu_object_info(zv->zv_objset, ZVOL_OBJ, &doi)) != 0 ||
(error = zvol_check_volsize(zc, doi.doi_data_block_size)) != 0) {
mutex_exit(&zvol_state_lock);
return (error);
}
if (zv->zv_readonly || (zv->zv_mode & DS_MODE_READONLY)) {
mutex_exit(&zvol_state_lock);
return (EROFS);
}
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
dmu_tx_hold_free(tx, ZVOL_OBJ, zc->zc_volsize, DMU_OBJECT_END);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
mutex_exit(&zvol_state_lock);
return (error);
}
error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1,
&zc->zc_volsize, tx);
if (error == 0) {
error = dmu_free_range(zv->zv_objset, ZVOL_OBJ, zc->zc_volsize,
DMU_OBJECT_END, tx);
}
dmu_tx_commit(tx);
if (error == 0) {
zv->zv_volsize = zc->zc_volsize;
zvol_size_changed(zv, dev);
}
mutex_exit(&zvol_state_lock);
return (error);
}
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);
}
int
zfs_set_version(const char *name, uint64_t newvers)
{
int error;
objset_t *os;
dmu_tx_t *tx;
uint64_t curvers;
/*
* XXX for now, require that the filesystem be unmounted. Would
* be nice to find the zfsvfs_t and just update that if
* possible.
*/
if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION)
return (EINVAL);
error = dmu_objset_open(name, DMU_OST_ZFS, DS_MODE_PRIMARY, &os);
if (error)
return (error);
error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_STR,
8, 1, &curvers);
if (error)
goto out;
if (newvers < curvers) {
error = EINVAL;
goto out;
}
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 0, ZPL_VERSION_STR);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
goto out;
}
error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 8, 1,
&newvers, tx);
spa_history_internal_log(LOG_DS_UPGRADE,
dmu_objset_spa(os), tx, CRED(),
"oldver=%llu newver=%llu dataset = %llu", curvers, newvers,
dmu_objset_id(os));
dmu_tx_commit(tx);
out:
dmu_objset_close(os);
return (error);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(zvol_state_t *zv, uint64_t volsize, objset_t *os)
{
struct block_device *bdev;
dmu_tx_t *tx;
int error;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(os);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
if (error)
return (error);
error = dmu_free_long_range(os,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
if (error)
return (error);
bdev = bdget_disk(zv->zv_disk, 0);
if (!bdev)
return (EIO);
/*
* 2.6.28 API change
* Added check_disk_size_change() helper function.
*/
#ifdef HAVE_CHECK_DISK_SIZE_CHANGE
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
check_disk_size_change(zv->zv_disk, bdev);
#else
zv->zv_volsize = volsize;
zv->zv_changed = 1;
(void) check_disk_change(bdev);
#endif /* HAVE_CHECK_DISK_SIZE_CHANGE */
bdput(bdev);
return (0);
}
static void
dsl_keychain_set_key(dsl_dir_t *dd, dmu_tx_t *tx,
caddr_t wkeybuf, size_t wkeylen, uint64_t rekey_date)
{
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t keychain_zapobj = dsl_dir_phys(dd)->dd_keychain_obj;
uint64_t props_zapobj = dsl_dir_phys(dd)->dd_props_zapobj;
ASSERT(keychain_zapobj != 0);
VERIFY(zap_update_uint64(mos, keychain_zapobj,
(uint64_t *)&tx->tx_txg, 1, 1, wkeylen, wkeybuf, tx) == 0);
VERIFY(zap_update(mos, props_zapobj,
zfs_prop_to_name(ZFS_PROP_REKEYDATE),
8, 1, (void *)&rekey_date, tx) == 0);
}
void
zvol_create_cb(objset_t *os, void *arg, dmu_tx_t *tx)
{
zfs_cmd_t *zc = arg;
int error;
error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, zc->zc_volblocksize,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &zc->zc_volsize, tx);
ASSERT(error == 0);
}
static void
dsl_deleg_set_sync(void *arg, dmu_tx_t *tx)
{
dsl_deleg_arg_t *dda = arg;
dsl_dir_t *dd;
dsl_pool_t *dp = dmu_tx_pool(tx);
objset_t *mos = dp->dp_meta_objset;
nvpair_t *whopair = NULL;
uint64_t zapobj;
VERIFY0(dsl_dir_hold(dp, dda->dda_name, FTAG, &dd, NULL));
zapobj = dd->dd_phys->dd_deleg_zapobj;
if (zapobj == 0) {
dmu_buf_will_dirty(dd->dd_dbuf, tx);
zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
}
while ((whopair = nvlist_next_nvpair(dda->dda_nvlist, whopair))) {
const char *whokey = nvpair_name(whopair);
nvlist_t *perms;
nvpair_t *permpair = NULL;
uint64_t jumpobj;
perms = fnvpair_value_nvlist(whopair);
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0) {
jumpobj = zap_create_link(mos, DMU_OT_DSL_PERMS,
zapobj, whokey, tx);
}
while ((permpair = nvlist_next_nvpair(perms, permpair))) {
const char *perm = nvpair_name(permpair);
uint64_t n = 0;
VERIFY(zap_update(mos, jumpobj,
perm, 8, 1, &n, tx) == 0);
spa_history_log_internal_dd(dd, "permission update", tx,
"%s %s", whokey, perm);
}
}
dsl_dir_rele(dd, FTAG);
}
/*
* Ensure the zap is flushed then inform the VFS of the capacity change.
*/
static int
zvol_update_volsize(zvol_state_t *zv, uint64_t volsize)
{
struct block_device *bdev;
dmu_tx_t *tx;
int error;
ASSERT(MUTEX_HELD(&zvol_state_lock));
tx = dmu_tx_create(zv->zv_objset);
dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
dmu_tx_abort(tx);
return (error);
}
error = zap_update(zv->zv_objset, ZVOL_ZAP_OBJ, "size", 8, 1,
&volsize, tx);
dmu_tx_commit(tx);
if (error)
return (error);
error = dmu_free_long_range(zv->zv_objset,
ZVOL_OBJ, volsize, DMU_OBJECT_END);
if (error)
return (error);
zv->zv_volsize = volsize;
zv->zv_changed = 1;
bdev = bdget_disk(zv->zv_disk, 0);
if (!bdev)
return EIO;
error = check_disk_change(bdev);
ASSERT3U(error, !=, 0);
bdput(bdev);
return (0);
}
static void
copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj,
boolean_t dosets, uint64_t uid, dmu_tx_t *tx)
{
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t jumpobj, pjumpobj;
uint64_t zapobj = dd->dd_phys->dd_deleg_zapobj;
zap_cursor_t zc;
zap_attribute_t za;
char whokey[ZFS_MAX_DELEG_NAME];
zfs_deleg_whokey(whokey,
dosets ? ZFS_DELEG_CREATE_SETS : ZFS_DELEG_CREATE,
ZFS_DELEG_LOCAL, NULL);
if (zap_lookup(mos, pzapobj, whokey, 8, 1, &pjumpobj) != 0)
return;
if (zapobj == 0) {
dmu_buf_will_dirty(dd->dd_dbuf, tx);
zapobj = dd->dd_phys->dd_deleg_zapobj = zap_create(mos,
DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
}
zfs_deleg_whokey(whokey,
dosets ? ZFS_DELEG_USER_SETS : ZFS_DELEG_USER,
ZFS_DELEG_LOCAL, &uid);
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) == ENOENT) {
jumpobj = zap_create(mos, DMU_OT_DSL_PERMS, DMU_OT_NONE, 0, tx);
VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0);
}
for (zap_cursor_init(&zc, mos, pjumpobj);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
uint64_t zero = 0;
ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1);
VERIFY(zap_update(mos, jumpobj, za.za_name,
8, 1, &zero, tx) == 0);
}
zap_cursor_fini(&zc);
}
/*
* dsl_crypto_key_clone
*
* Our caller (dmu_objset_create_sync) must have a lock on the dataset.
*/
int
dsl_crypto_key_clone(dsl_dir_t *dd, dsl_dataset_phys_t *dsphys,
uint64_t dsobj, dsl_dataset_t *clone_origin,
dsl_crypto_ctx_t *ctx, dmu_tx_t *tx)
{
zcrypt_key_t *txgkey;
zcrypt_key_t *wrappingkey = NULL;
dsl_dataset_t *origin;
uint64_t crypt;
spa_t *spa = dd->dd_pool->dp_spa;
int error = 0;
ASSERT(ctx != NULL);
if (BP_IS_HOLE(&dsphys->ds_bp)) {
origin = ctx->dcc_origin_ds;
} else {
origin = clone_origin;
}
ASSERT(origin != NULL);
/*
* Need to look at the value of crypt on the origin snapshot
* since it is sticky for encryption.
*/
VERIFY(dsl_prop_get_ds(origin, zfs_prop_to_name(ZFS_PROP_ENCRYPTION),
8, 1, &crypt,/* DSL_PROP_GET_EFFECTIVE,*/ NULL) == 0);
if (crypt == ZIO_CRYPT_OFF) {
return (0);
}
wrappingkey = ctx->dcc_wrap_key;
dsl_keychain_create_obj(dd, tx);
dsl_keychain_clone_phys(origin, dd, tx, wrappingkey);
if (ctx->dcc_salt != 0) {
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t props_zapobj = dsl_dir_phys(dd)->dd_props_zapobj;
error = zap_update(mos, props_zapobj,
zfs_prop_to_name(ZFS_PROP_SALT), 8, 1,
(void *)&ctx->dcc_salt, tx);
}
if (ctx->dcc_clone_newkey) {
caddr_t wkeybuf;
size_t wkeylen;
/*
* Generate a new key and add it to the keychain
* to be valid from this txg onwards.
*/
zcrypt_key_hold(wrappingkey, FTAG);
txgkey = zcrypt_key_gen(crypt);
VERIFY(zcrypt_wrap_key(wrappingkey, txgkey,
&wkeybuf, &wkeylen, zio_crypt_select_wrap(crypt)) == 0);
zcrypt_key_release(wrappingkey, FTAG);
dsl_keychain_set_key(dd, tx, wkeybuf, wkeylen,
dsphys->ds_creation_time);
kmem_free(wkeybuf, wkeylen);
zcrypt_key_free(txgkey);
spa_history_log_internal(spa, "key create", tx,
"rekey succeeded dataset = %llu from dataset = %llu",
dsobj, clone_origin->ds_object);
} else {
spa_history_log_internal(spa, "key create", tx,
"succeeded dataset = %llu from dataset = %llu",
dsobj, clone_origin->ds_object);
}
if (wrappingkey != NULL) {
error = dsl_keychain_load_dd(dd, dsobj, crypt, wrappingkey);
}
return (error);
}
/*
* dsl_crypto_key_gen - Generate dataset key
*
* Generate a new key for this dataset based on its encryption property type.
* Store the key as a usable zcrypt_key_t in the in memory keystore and
* put the wrapped version of it in the on disk keychain.
*
* returns 0 on success
*/
int
dsl_crypto_key_create(dsl_dir_t *dd, dsl_dataset_phys_t *dsphys,
uint64_t dsobj, dsl_crypto_ctx_t *ctx, dmu_tx_t *tx)
{
int error = -1;
#ifdef DEBUG
zcrypt_key_t *debugkey;
#endif /* DEBUG */
zcrypt_key_t *wrappingkey, *dslkey;
caddr_t wkeybuf = NULL;
size_t wkeylen = 0;
uint64_t crypt;
spa_t *spa = dd->dd_pool->dp_spa;
zcrypt_keystore_node_t *skn;
if (ctx == NULL) {
return (0);
}
crypt = ctx->dcc_crypt;
if (crypt == ZIO_CRYPT_OFF || crypt == ZIO_CRYPT_INHERIT) {
return (0);
}
crypt = zio_crypt_select(crypt, ZIO_CRYPT_ON_VALUE);
wrappingkey = ctx->dcc_wrap_key;
skn = zcrypt_keystore_insert(spa, dsobj, wrappingkey);
if (skn == NULL) {
error = ENOKEY;
goto out;
}
dslkey = zcrypt_key_gen(crypt);
zcrypt_key_hold(wrappingkey, FTAG);
error = zcrypt_wrap_key(wrappingkey, dslkey,
&wkeybuf, &wkeylen, zio_crypt_select_wrap(crypt));
if (error != 0) {
zcrypt_key_free(dslkey);
spa_history_log_internal(spa, "key create", tx,
"failed dataset = %llu unable to wrap key", dsobj, error);
goto out;
}
dsl_keychain_create_obj(dd, tx);
dsl_keychain_set_key(dd, tx, wkeybuf, wkeylen,
dsphys->ds_creation_time);
zcrypt_keychain_insert(&skn->skn_keychain,
dsphys->ds_creation_txg, dslkey);
if (ctx->dcc_salt != 0) {
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t props_zapobj = dsl_dir_phys(dd)->dd_props_zapobj;
error = zap_update(mos, props_zapobj,
zfs_prop_to_name(ZFS_PROP_SALT), 8, 1,
(void *)&ctx->dcc_salt, tx);
}
if (error == 0)
spa_history_log_internal(spa, "key create", tx,
"succeeded dataset = %llu", dsobj);
#ifdef DEBUG
ASSERT3U(zcrypt_unwrap_key(wrappingkey, crypt,
wkeybuf, wkeylen, &debugkey), ==, 0);
zcrypt_key_compare(dslkey, debugkey);
zcrypt_key_free(debugkey);
#endif /* DEBUG */
out:
zcrypt_key_release(wrappingkey, FTAG);
if (wkeylen > 0) {
kmem_free(wkeybuf, wkeylen);
}
return (error);
}
void
zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx)
{
struct super_block *sb;
zfs_sb_t *zsb;
uint64_t moid, obj, sa_obj, version;
uint64_t sense = ZFS_CASE_SENSITIVE;
uint64_t norm = 0;
nvpair_t *elem;
int error;
int i;
znode_t *rootzp = NULL;
vattr_t vattr;
znode_t *zp;
zfs_acl_ids_t acl_ids;
/*
* First attempt to create master node.
*/
/*
* In an empty objset, there are no blocks to read and thus
* there can be no i/o errors (which we assert below).
*/
moid = MASTER_NODE_OBJ;
error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE,
DMU_OT_NONE, 0, tx);
ASSERT(error == 0);
/*
* Set starting attributes.
*/
version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os)));
elem = NULL;
while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) {
/* For the moment we expect all zpl props to be uint64_ts */
uint64_t val;
char *name;
ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64);
VERIFY(nvpair_value_uint64(elem, &val) == 0);
name = nvpair_name(elem);
if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) {
if (val < version)
version = val;
} else {
error = zap_update(os, moid, name, 8, 1, &val, tx);
}
ASSERT(error == 0);
if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0)
norm = val;
else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0)
sense = val;
}
ASSERT(version != 0);
error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx);
/*
* Create zap object used for SA attribute registration
*/
if (version >= ZPL_VERSION_SA) {
sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE,
DMU_OT_NONE, 0, tx);
error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx);
ASSERT(error == 0);
} else {
sa_obj = 0;
}
/*
* Create a delete queue.
*/
obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx);
error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx);
ASSERT(error == 0);
/*
* Create root znode. Create minimal znode/inode/zsb/sb
* to allow zfs_mknode to work.
*/
vattr.va_mask = ATTR_MODE|ATTR_UID|ATTR_GID;
vattr.va_mode = S_IFDIR|0755;
vattr.va_uid = crgetuid(cr);
vattr.va_gid = crgetgid(cr);
rootzp = kmem_cache_alloc(znode_cache, KM_PUSHPAGE);
rootzp->z_moved = 0;
rootzp->z_unlinked = 0;
rootzp->z_atime_dirty = 0;
rootzp->z_is_sa = USE_SA(version, os);
zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_PUSHPAGE | KM_NODEBUG);
zsb->z_os = os;
zsb->z_parent = zsb;
zsb->z_version = version;
zsb->z_use_fuids = USE_FUIDS(version, os);
zsb->z_use_sa = USE_SA(version, os);
zsb->z_norm = norm;
sb = kmem_zalloc(sizeof (struct super_block), KM_PUSHPAGE);
sb->s_fs_info = zsb;
ZTOI(rootzp)->i_sb = sb;
error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
&zsb->z_attr_table);
ASSERT(error == 0);
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
*/
if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED)
zsb->z_norm |= U8_TEXTPREP_TOUPPER;
mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zsb->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr,
cr, NULL, &acl_ids));
zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids);
ASSERT3P(zp, ==, rootzp);
error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx);
ASSERT(error == 0);
zfs_acl_ids_free(&acl_ids);
atomic_set(&ZTOI(rootzp)->i_count, 0);
sa_handle_destroy(rootzp->z_sa_hdl);
kmem_cache_free(znode_cache, rootzp);
/*
* Create shares directory
*/
error = zfs_create_share_dir(zsb, tx);
ASSERT(error == 0);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_destroy(&zsb->z_hold_mtx[i]);
kmem_free(sb, sizeof (struct super_block));
kmem_free(zsb, sizeof (zfs_sb_t));
}
/**
* update/insert/delete the specified OI mapping (@fid @id) according to the ops
*
* \retval 1, changed nothing
* \retval 0, changed successfully
* \retval -ve, on error
*/
static int osd_scrub_refresh_mapping(const struct lu_env *env,
struct osd_device *dev,
const struct lu_fid *fid,
uint64_t oid, int ops,
bool force, const char *name)
{
struct osd_thread_info *info = osd_oti_get(env);
struct zpl_direntry *zde = &info->oti_zde.lzd_reg;
char *buf = info->oti_str;
dmu_tx_t *tx = NULL;
dnode_t *dn = NULL;
uint64_t zapid;
int rc;
ENTRY;
if (dev->od_scrub.os_file.sf_param & SP_DRYRUN && !force)
GOTO(log, rc = 0);
tx = dmu_tx_create(dev->od_os);
if (!tx)
GOTO(log, rc = -ENOMEM);
zapid = osd_get_name_n_idx(env, dev, fid, buf,
sizeof(info->oti_str), &dn);
osd_tx_hold_zap(tx, zapid, dn,
ops == DTO_INDEX_INSERT ? TRUE : FALSE, NULL);
rc = -dmu_tx_assign(tx, TXG_WAIT);
if (rc) {
dmu_tx_abort(tx);
GOTO(log, rc);
}
switch (ops) {
case DTO_INDEX_UPDATE:
zde->zde_pad = 0;
zde->zde_dnode = oid;
zde->zde_type = 0; /* The type in OI mapping is useless. */
rc = -zap_update(dev->od_os, zapid, buf, 8, sizeof(*zde) / 8,
zde, tx);
if (unlikely(rc == -ENOENT)) {
/* Some unlink thread may removed the OI mapping. */
rc = 1;
}
break;
case DTO_INDEX_INSERT:
zde->zde_pad = 0;
zde->zde_dnode = oid;
zde->zde_type = 0; /* The type in OI mapping is useless. */
rc = osd_zap_add(dev, zapid, dn, buf, 8, sizeof(*zde) / 8,
zde, tx);
if (unlikely(rc == -EEXIST))
rc = 1;
break;
case DTO_INDEX_DELETE:
rc = osd_zap_remove(dev, zapid, dn, buf, tx);
if (rc == -ENOENT) {
/* It is normal that the unlink thread has removed the
* OI mapping already. */
rc = 1;
}
break;
default:
LASSERTF(0, "Unexpected ops %d\n", ops);
rc = -EINVAL;
break;
}
dmu_tx_commit(tx);
GOTO(log, rc);
log:
CDEBUG(D_LFSCK, "%s: refresh OI map for scrub, op %d, force %s, "
DFID" => %llu (%s): rc = %d\n", osd_name(dev), ops,
force ? "yes" : "no", PFID(fid), oid, name ? name : "null", rc);
return rc;
}
/* ARGSUSED */
static void
dsl_pool_scrub_setup_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
{
dsl_pool_t *dp = arg1;
enum scrub_func *funcp = arg2;
dmu_object_type_t ot = 0;
boolean_t complete = B_FALSE;
dsl_pool_scrub_cancel_sync(dp, &complete, cr, tx);
ASSERT(dp->dp_scrub_func == SCRUB_FUNC_NONE);
ASSERT(*funcp > SCRUB_FUNC_NONE);
ASSERT(*funcp < SCRUB_FUNC_NUMFUNCS);
dp->dp_scrub_min_txg = 0;
dp->dp_scrub_max_txg = tx->tx_txg;
dp->dp_scrub_ddt_class_max = zfs_scrub_ddt_class_max;
if (*funcp == SCRUB_FUNC_CLEAN) {
vdev_t *rvd = dp->dp_spa->spa_root_vdev;
/* rewrite all disk labels */
vdev_config_dirty(rvd);
if (vdev_resilver_needed(rvd,
&dp->dp_scrub_min_txg, &dp->dp_scrub_max_txg)) {
spa_event_notify(dp->dp_spa, NULL,
ESC_ZFS_RESILVER_START);
dp->dp_scrub_max_txg = MIN(dp->dp_scrub_max_txg,
tx->tx_txg);
} else {
spa_event_notify(dp->dp_spa, NULL,
ESC_ZFS_SCRUB_START);
}
/* zero out the scrub stats in all vdev_stat_t's */
vdev_scrub_stat_update(rvd,
dp->dp_scrub_min_txg ? POOL_SCRUB_RESILVER :
POOL_SCRUB_EVERYTHING, B_FALSE);
/*
* 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 (dp->dp_scrub_min_txg > TXG_INITIAL)
dp->dp_scrub_ddt_class_max = DDT_CLASS_DITTO;
dp->dp_spa->spa_scrub_started = B_TRUE;
}
/* 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(dp->dp_spa) < SPA_VERSION_DSL_SCRUB)
ot = DMU_OT_ZAP_OTHER;
dp->dp_scrub_func = *funcp;
dp->dp_scrub_queue_obj = zap_create(dp->dp_meta_objset,
ot ? ot : DMU_OT_SCRUB_QUEUE, DMU_OT_NONE, 0, tx);
bzero(&dp->dp_scrub_bookmark, sizeof (zbookmark_t));
bzero(&dp->dp_scrub_ddt_bookmark, sizeof (ddt_bookmark_t));
dp->dp_scrub_restart = B_FALSE;
dp->dp_spa->spa_scrub_errors = 0;
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_FUNC, sizeof (uint32_t), 1,
&dp->dp_scrub_func, tx));
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_QUEUE, sizeof (uint64_t), 1,
&dp->dp_scrub_queue_obj, tx));
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_MIN_TXG, sizeof (uint64_t), 1,
&dp->dp_scrub_min_txg, tx));
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_MAX_TXG, sizeof (uint64_t), 1,
&dp->dp_scrub_max_txg, tx));
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_BOOKMARK, sizeof (uint64_t),
sizeof (dp->dp_scrub_bookmark) / sizeof (uint64_t),
&dp->dp_scrub_bookmark, tx));
VERIFY(0 == zap_update(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_DDT_BOOKMARK, sizeof (uint64_t),
sizeof (dp->dp_scrub_ddt_bookmark) / sizeof (uint64_t),
&dp->dp_scrub_ddt_bookmark, tx));
VERIFY(0 == zap_update(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_DDT_CLASS_MAX, sizeof (uint64_t), 1,
&dp->dp_scrub_ddt_class_max, tx));
VERIFY(0 == zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_ERRORS, sizeof (uint64_t), 1,
&dp->dp_spa->spa_scrub_errors, tx));
spa_history_internal_log(LOG_POOL_SCRUB, dp->dp_spa, tx, cr,
"func=%u mintxg=%llu maxtxg=%llu",
*funcp, dp->dp_scrub_min_txg, dp->dp_scrub_max_txg);
}
static int
feature_do_action(objset_t *os, uint64_t read_obj, uint64_t write_obj,
uint64_t desc_obj, zfeature_info_t *feature, feature_action_t action,
dmu_tx_t *tx)
{
int error;
uint64_t refcount;
uint64_t zapobj = feature->fi_can_readonly ? write_obj : read_obj;
ASSERT(0 != zapobj);
ASSERT(zfeature_is_valid_guid(feature->fi_guid));
error = zap_lookup(os, zapobj, feature->fi_guid,
sizeof (uint64_t), 1, &refcount);
/*
* If we can't ascertain the status of the specified feature, an I/O
* error occurred.
*/
if (error != 0 && error != ENOENT)
return (error);
switch (action) {
case FEATURE_ACTION_ENABLE:
/*
* If the feature is already enabled, ignore the request.
*/
if (error == 0)
return (0);
refcount = 0;
break;
case FEATURE_ACTION_INCR:
if (error == ENOENT)
return (ENOTSUP);
if (refcount == UINT64_MAX)
return (EOVERFLOW);
refcount++;
break;
case FEATURE_ACTION_DECR:
if (error == ENOENT)
return (ENOTSUP);
if (refcount == 0)
return (EOVERFLOW);
refcount--;
break;
default:
ASSERT(0);
break;
}
if (action == FEATURE_ACTION_ENABLE) {
int i;
for (i = 0; feature->fi_depends[i] != NULL; i++) {
zfeature_info_t *dep = feature->fi_depends[i];
error = feature_do_action(os, read_obj, write_obj,
desc_obj, dep, FEATURE_ACTION_ENABLE, tx);
if (error != 0)
return (error);
}
}
error = zap_update(os, zapobj, feature->fi_guid,
sizeof (uint64_t), 1, &refcount, tx);
if (error != 0)
return (error);
if (action == FEATURE_ACTION_ENABLE) {
error = zap_update(os, desc_obj,
feature->fi_guid, 1, strlen(feature->fi_desc) + 1,
feature->fi_desc, tx);
if (error != 0)
return (error);
}
if (action == FEATURE_ACTION_INCR && refcount == 1 && feature->fi_mos) {
spa_activate_mos_feature(dmu_objset_spa(os), feature->fi_guid);
}
if (action == FEATURE_ACTION_DECR && refcount == 0) {
spa_deactivate_mos_feature(dmu_objset_spa(os),
feature->fi_guid);
}
return (0);
}
