/*
* dsl_pool must not be held when this is called.
* Upon successful return, there will be a longhold on the dataset,
* and the dsl_pool will not be held.
*/
int
dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, void *tag, objset_t **osp)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int err;
err = dsl_pool_hold(name, FTAG, &dp);
if (err != 0)
return (err);
err = dsl_dataset_own(dp, name, tag, &ds);
if (err != 0) {
dsl_pool_rele(dp, FTAG);
return (err);
}
err = dmu_objset_from_ds(ds, osp);
dsl_pool_rele(dp, FTAG);
if (err != 0) {
dsl_dataset_disown(ds, tag);
} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
dsl_dataset_disown(ds, tag);
return (SET_ERROR(EINVAL));
} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
dsl_dataset_disown(ds, tag);
return (SET_ERROR(EROFS));
}
return (err);
}
static int
dsl_bookmark_create_check_impl(dsl_dataset_t *snapds, const char *bookmark_name,
dmu_tx_t *tx)
{
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dataset_t *bmark_fs;
char *shortname;
int error;
zfs_bookmark_phys_t bmark_phys;
if (!dsl_dataset_is_snapshot(snapds))
return (SET_ERROR(EINVAL));
error = dsl_bookmark_hold_ds(dp, bookmark_name,
&bmark_fs, FTAG, &shortname);
if (error != 0)
return (error);
if (!dsl_dataset_is_before(bmark_fs, snapds, 0)) {
dsl_dataset_rele(bmark_fs, FTAG);
return (SET_ERROR(EINVAL));
}
error = dsl_dataset_bmark_lookup(bmark_fs, shortname,
&bmark_phys);
dsl_dataset_rele(bmark_fs, FTAG);
if (error == 0)
return (SET_ERROR(EEXIST));
if (error == ESRCH)
return (0);
return (error);
}
/*ARGSUSED*/
static int
dmu_objset_create_check(void *arg1, void *arg2, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
struct oscarg *oa = arg2;
objset_t *mos = dd->dd_pool->dp_meta_objset;
int err;
uint64_t ddobj;
err = zap_lookup(mos, dd->dd_phys->dd_child_dir_zapobj,
oa->lastname, sizeof (uint64_t), 1, &ddobj);
if (err != ENOENT)
return (err ? err : EEXIST);
if (oa->clone_origin != NULL) {
/* You can't clone across pools. */
if (oa->clone_origin->ds_dir->dd_pool != dd->dd_pool)
return (EXDEV);
/* You can only clone snapshots, not the head datasets. */
if (!dsl_dataset_is_snapshot(oa->clone_origin))
return (EINVAL);
}
return (0);
}
int
dsl_destroy_snapshot_check_impl(dsl_dataset_t *ds, boolean_t defer)
{
if (!dsl_dataset_is_snapshot(ds))
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 (ds->ds_phys->ds_num_children > 1)
return (SET_ERROR(EEXIST));
return (0);
}
int
dsl_prop_get_ds(dsl_dataset_t *ds, const char *propname,
int intsz, int numints, void *buf, char *setpoint)
{
zfs_prop_t prop = zfs_name_to_prop(propname);
boolean_t inheritable;
boolean_t snapshot;
uint64_t zapobj;
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
inheritable = (prop == ZPROP_INVAL || zfs_prop_inheritable(prop));
snapshot = (ds->ds_phys != NULL && dsl_dataset_is_snapshot(ds));
zapobj = (ds->ds_phys == NULL ? 0 : ds->ds_phys->ds_props_obj);
if (zapobj != 0) {
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
int err;
ASSERT(snapshot);
/* Check for a local value. */
err = zap_lookup(mos, zapobj, propname, intsz, numints, buf);
if (err != ENOENT) {
if (setpoint != NULL && err == 0)
dsl_dataset_name(ds, setpoint);
return (err);
}
/*
* Skip the check for a received value if there is an explicit
* inheritance entry.
*/
if (inheritable) {
char *inheritstr = kmem_asprintf("%s%s", propname,
ZPROP_INHERIT_SUFFIX);
err = zap_contains(mos, zapobj, inheritstr);
strfree(inheritstr);
if (err != 0 && err != ENOENT)
return (err);
}
if (err == ENOENT) {
/* Check for a received value. */
char *recvdstr = kmem_asprintf("%s%s", propname,
ZPROP_RECVD_SUFFIX);
err = zap_lookup(mos, zapobj, recvdstr,
intsz, numints, buf);
strfree(recvdstr);
if (err != ENOENT) {
if (setpoint != NULL && err == 0)
(void) strcpy(setpoint,
ZPROP_SOURCE_VAL_RECVD);
return (err);
}
}
}
return (dsl_prop_get_dd(ds->ds_dir, propname,
intsz, numints, buf, setpoint, snapshot));
}
static uint64_t
dsl_scan_ds_maxtxg(dsl_dataset_t *ds)
{
uint64_t smt = ds->ds_dir->dd_pool->dp_scan->scn_phys.scn_max_txg;
if (dsl_dataset_is_snapshot(ds))
return (MIN(smt, ds->ds_phys->ds_creation_txg));
return (smt);
}
int
dmu_objset_is_snapshot(objset_t *os)
{
if (os->os->os_dsl_dataset != NULL)
return (dsl_dataset_is_snapshot(os->os->os_dsl_dataset));
else
return (B_FALSE);
}
/*ARGSUSED*/
static int
dmu_objset_clone_check(void *arg, dmu_tx_t *tx)
{
dmu_objset_clone_arg_t *doca = arg;
dsl_dir_t *pdd;
const char *tail;
int error;
dsl_dataset_t *origin;
dsl_pool_t *dp = dmu_tx_pool(tx);
if (strchr(doca->doca_clone, '@') != NULL)
return (SET_ERROR(EINVAL));
error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail);
if (error != 0)
return (error);
if (tail == NULL) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EEXIST));
}
/* You can't clone across pools. */
if (pdd->dd_pool != dp) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EXDEV));
}
error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL,
doca->doca_cred);
if (error != 0) {
dsl_dir_rele(pdd, FTAG);
return (SET_ERROR(EDQUOT));
}
dsl_dir_rele(pdd, FTAG);
error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin);
if (error != 0)
return (error);
/* You can't clone across pools. */
if (origin->ds_dir->dd_pool != dp) {
dsl_dataset_rele(origin, FTAG);
return (SET_ERROR(EXDEV));
}
/* You can only clone snapshots, not the head datasets. */
if (!dsl_dataset_is_snapshot(origin)) {
dsl_dataset_rele(origin, FTAG);
return (SET_ERROR(EINVAL));
}
dsl_dataset_rele(origin, FTAG);
return (0);
}
void
dmu_objset_evict(dsl_dataset_t *ds, void *arg)
{
objset_impl_t *osi = arg;
objset_t os;
int i;
for (i = 0; i < TXG_SIZE; i++) {
ASSERT(list_head(&osi->os_dirty_dnodes[i]) == NULL);
ASSERT(list_head(&osi->os_free_dnodes[i]) == NULL);
}
if (ds) {
if (!dsl_dataset_is_snapshot(ds)) {
VERIFY(0 == dsl_prop_unregister(ds, "checksum",
checksum_changed_cb, osi));
VERIFY(0 == dsl_prop_unregister(ds, "compression",
compression_changed_cb, osi));
VERIFY(0 == dsl_prop_unregister(ds, "copies",
copies_changed_cb, osi));
}
VERIFY(0 == dsl_prop_unregister(ds, "primarycache",
primary_cache_changed_cb, osi));
VERIFY(0 == dsl_prop_unregister(ds, "secondarycache",
secondary_cache_changed_cb, osi));
}
/*
* We should need only a single pass over the dnode list, since
* nothing can be added to the list at this point.
*/
os.os = osi;
(void) dmu_objset_evict_dbufs(&os);
dnode_special_close(osi->os_meta_dnode);
if (osi->os_userused_dnode) {
dnode_special_close(osi->os_userused_dnode);
dnode_special_close(osi->os_groupused_dnode);
}
zil_free(osi->os_zil);
ASSERT3P(list_head(&osi->os_dnodes), ==, NULL);
VERIFY(arc_buf_remove_ref(osi->os_phys_buf, &osi->os_phys_buf) == 1);
mutex_destroy(&osi->os_lock);
mutex_destroy(&osi->os_obj_lock);
mutex_destroy(&osi->os_user_ptr_lock);
kmem_free(osi, sizeof (objset_impl_t));
}
static int
dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type,
boolean_t readonly, void *tag, objset_t **osp)
{
int err;
err = dmu_objset_from_ds(ds, osp);
if (err != 0) {
dsl_dataset_disown(ds, tag);
} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
dsl_dataset_disown(ds, tag);
return (SET_ERROR(EINVAL));
} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
dsl_dataset_disown(ds, tag);
return (SET_ERROR(EROFS));
}
return (err);
}
zcrypt_keystore_node_t *
zcrypt_keystore_find_node(spa_t *spa, uint64_t dsobj,
boolean_t config_rwlock_held)
{
zcrypt_keystore_node_t search;
zcrypt_keystore_node_t *found = NULL;
rw_enter(&spa->spa_keystore->sk_lock, RW_READER);
if (avl_is_empty(&spa->spa_keystore->sk_dslkeys))
goto out;
search.skn_os = dsobj;
found = avl_find(&spa->spa_keystore->sk_dslkeys, &search, NULL);
if (found == NULL) {
int error;
dsl_pool_t *dp = spa_get_dsl(spa);
dsl_dataset_t *ds;
boolean_t need_lock;
rw_exit(&spa->spa_keystore->sk_lock);
need_lock = !dsl_pool_sync_context(dp) && !config_rwlock_held;
if (need_lock)
rw_enter(&dp->dp_config_rwlock, RW_READER);
error = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (need_lock)
rw_exit(&dp->dp_config_rwlock);
rw_enter(&spa->spa_keystore->sk_lock, RW_READER);
if (!error) {
if (dsl_dataset_is_snapshot(ds)) {
search.skn_os =
ds->ds_dir->dd_phys->dd_head_dataset_obj;
found = avl_find(&spa->spa_keystore->sk_dslkeys,
&search, NULL);
}
dsl_dataset_rele(ds, FTAG);
}
}
out:
rw_exit(&spa->spa_keystore->sk_lock);
return (found);
}
static int
dsl_dataset_user_release_check_one(dsl_dataset_t *ds,
nvlist_t *holds, boolean_t *todelete)
{
uint64_t zapobj;
nvpair_t *pair;
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
int error;
int numholds = 0;
*todelete = B_FALSE;
if (!dsl_dataset_is_snapshot(ds))
return (EINVAL);
zapobj = ds->ds_phys->ds_userrefs_obj;
if (zapobj == 0)
return (ESRCH);
for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
/* Make sure the hold exists */
uint64_t tmp;
error = zap_lookup(mos, zapobj, nvpair_name(pair), 8, 1, &tmp);
if (error == ENOENT)
error = ESRCH;
if (error != 0)
return (error);
numholds++;
}
if (DS_IS_DEFER_DESTROY(ds) && ds->ds_phys->ds_num_children == 1 &&
ds->ds_userrefs == numholds) {
/* we need to destroy the snapshot as well */
if (dsl_dataset_long_held(ds))
return (EBUSY);
*todelete = B_TRUE;
}
return (0);
}
/*ARGSUSED*/
static int
dmu_objset_create_check(void *arg1, void *arg2, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
struct oscarg *oa = arg2;
objset_t *mos = dd->dd_pool->dp_meta_objset;
int err;
uint64_t ddobj;
static boolean_t dp_config_rwlock_held = B_TRUE;
err = zap_lookup(mos, dd->dd_phys->dd_child_dir_zapobj,
oa->lastname, sizeof (uint64_t), 1, &ddobj);
if (err != ENOENT)
return (err ? err : EEXIST);
if (oa->clone_origin != NULL) {
/* You can't clone across pools. */
if (oa->clone_origin->ds_dir->dd_pool != dd->dd_pool)
return (EXDEV);
/* You can only clone snapshots, not the head datasets. */
if (!dsl_dataset_is_snapshot(oa->clone_origin))
return (EINVAL);
if (dsl_dataset_keystatus(oa->clone_origin,
dp_config_rwlock_held) == ZFS_CRYPT_KEY_UNAVAILABLE)
return (ENOKEY);
}
/*
* Check we have the required crypto algorithms available
* via kcf since this is our last chance to fail the dataset creation.
*/
if (oa->crypto_ctx != NULL &&
!zcrypt_mech_available(oa->crypto_ctx->dcc_crypt)) {
return (ENOTSUP);
}
return (0);
}
/* called from zpl */
int
dmu_objset_own(const char *name, dmu_objset_type_t type,
boolean_t readonly, void *tag, objset_t **osp)
{
dsl_dataset_t *ds;
int err;
err = dsl_dataset_own(name, B_FALSE, tag, &ds);
if (err)
return (err);
err = dmu_objset_from_ds(ds, osp);
if (err) {
dsl_dataset_disown(ds, tag);
} else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) {
dmu_objset_disown(*osp, tag);
return (EINVAL);
} else if (!readonly && dsl_dataset_is_snapshot(ds)) {
dmu_objset_disown(*osp, tag);
return (EROFS);
}
return (err);
}
int
dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
objset_impl_t **osip)
{
objset_impl_t *osi;
int i, err;
ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
osi = kmem_zalloc(sizeof (objset_impl_t), KM_SLEEP);
osi->os.os = osi;
osi->os_dsl_dataset = ds;
osi->os_spa = spa;
osi->os_rootbp = bp;
if (!BP_IS_HOLE(osi->os_rootbp)) {
uint32_t aflags = ARC_WAIT;
zbookmark_t zb;
zb.zb_objset = ds ? ds->ds_object : 0;
zb.zb_object = 0;
zb.zb_level = -1;
zb.zb_blkid = 0;
if (DMU_OS_IS_L2CACHEABLE(osi))
aflags |= ARC_L2CACHE;
dprintf_bp(osi->os_rootbp, "reading %s", "");
/*
* NB: 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 = arc_read_nolock(NULL, spa, osi->os_rootbp,
arc_getbuf_func, &osi->os_phys_buf,
ZIO_PRIORITY_SYNC_READ, ZIO_FLAG_CANFAIL, &aflags, &zb);
if (err) {
kmem_free(osi, sizeof (objset_impl_t));
/* convert checksum errors into IO errors */
if (err == ECKSUM)
err = EIO;
return (err);
}
osi->os_phys = osi->os_phys_buf->b_data;
} else {
osi->os_phys_buf = arc_buf_alloc(spa, sizeof (objset_phys_t),
&osi->os_phys_buf, ARC_BUFC_METADATA);
osi->os_phys = osi->os_phys_buf->b_data;
bzero(osi->os_phys, sizeof (objset_phys_t));
}
/*
* 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) {
err = dsl_prop_register(ds, "primarycache",
primary_cache_changed_cb, osi);
if (err == 0)
err = dsl_prop_register(ds, "secondarycache",
secondary_cache_changed_cb, osi);
if (!dsl_dataset_is_snapshot(ds)) {
if (err == 0)
err = dsl_prop_register(ds, "checksum",
checksum_changed_cb, osi);
if (err == 0)
err = dsl_prop_register(ds, "compression",
compression_changed_cb, osi);
if (err == 0)
err = dsl_prop_register(ds, "copies",
copies_changed_cb, osi);
}
if (err) {
VERIFY(arc_buf_remove_ref(osi->os_phys_buf,
&osi->os_phys_buf) == 1);
kmem_free(osi, sizeof (objset_impl_t));
return (err);
}
} else if (ds == NULL) {
/* It's the meta-objset. */
osi->os_checksum = ZIO_CHECKSUM_FLETCHER_4;
osi->os_compress = ZIO_COMPRESS_LZJB;
osi->os_copies = spa_max_replication(spa);
osi->os_primary_cache = ZFS_CACHE_ALL;
osi->os_secondary_cache = ZFS_CACHE_ALL;
}
osi->os_zil_header = osi->os_phys->os_zil_header;
osi->os_zil = zil_alloc(&osi->os, &osi->os_zil_header);
for (i = 0; i < TXG_SIZE; i++) {
list_create(&osi->os_dirty_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
list_create(&osi->os_free_dnodes[i], sizeof (dnode_t),
offsetof(dnode_t, dn_dirty_link[i]));
}
list_create(&osi->os_dnodes, sizeof (dnode_t),
offsetof(dnode_t, dn_link));
list_create(&osi->os_downgraded_dbufs, sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_link));
mutex_init(&osi->os_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&osi->os_obj_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&osi->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL);
osi->os_meta_dnode = dnode_special_open(osi,
&osi->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT);
/*
* We should be the only thread trying to do this because we
* have ds_opening_lock
*/
if (ds) {
VERIFY(NULL == dsl_dataset_set_user_ptr(ds, osi,
dmu_objset_evict));
}
*osip = osi;
return (0);
}
void
dmu_objset_evict(objset_t *os)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
int t;
for (t = 0; t < TXG_SIZE; t++)
ASSERT(!dmu_objset_is_dirty(os, t));
if (ds) {
if (!dsl_dataset_is_snapshot(ds)) {
VERIFY(0 == dsl_prop_unregister(ds, "checksum",
checksum_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "compression",
compression_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "copies",
copies_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "dedup",
dedup_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "logbias",
logbias_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "sync",
sync_changed_cb, os));
}
VERIFY(0 == dsl_prop_unregister(ds, "primarycache",
primary_cache_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "secondarycache",
secondary_cache_changed_cb, os));
VERIFY(0 == dsl_prop_unregister(ds, "encryption",
crypt_changed_cb, os));
}
if (os->os_sa)
sa_tear_down(os);
/*
* We should need only a single pass over the dnode list, since
* nothing can be added to the list at this point.
*/
(void) dmu_objset_evict_dbufs(os);
dnode_special_close(&os->os_meta_dnode);
if (DMU_USERUSED_DNODE(os)) {
dnode_special_close(&os->os_userused_dnode);
dnode_special_close(&os->os_groupused_dnode);
}
zil_free(os->os_zil);
ASSERT3P(list_head(&os->os_dnodes), ==, NULL);
VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf) == 1);
/*
* This is a barrier to prevent the objset from going away in
* dnode_move() until we can safely ensure that the objset is still in
* use. We consider the objset valid before the barrier and invalid
* after the barrier.
*/
rw_enter(&os_lock, RW_READER);
rw_exit(&os_lock);
mutex_destroy(&os->os_lock);
mutex_destroy(&os->os_obj_lock);
mutex_destroy(&os->os_user_ptr_lock);
kmem_free(os, sizeof (objset_t));
}
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);
}
int
dsl_crypto_key_new(const char *dsname)
{
dsl_dataset_t *ds;
objset_t *os;
zcrypt_keystore_node_t *skn;
spa_t *spa;
struct knarg arg;
int error;
dsl_pool_t *dp;
void *cookie;
error = dsl_pool_hold(dsname, FTAG, &dp);
if (error != 0)
return (error);
if ((error = dsl_dataset_hold(dp, dsname, FTAG, &ds)) != 0) {
dsl_pool_rele(dp, FTAG);
return (error);
}
if (dsl_dataset_is_snapshot(ds)) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (ENOTSUP);
}
if ((error = dmu_objset_from_ds(ds, &os)) != 0) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (error);
}
if (os->os_crypt == ZIO_CRYPT_OFF) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (ENOTSUP);
}
ASSERT(os->os_crypt != ZIO_CRYPT_INHERIT);
/*
* Need the keychain and wrapping key to already be available.
*/
spa = dsl_dataset_get_spa(ds);
skn = zcrypt_keystore_find_node(spa, ds->ds_object, B_FALSE);
if (skn == NULL) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (ENOENT);
}
ASSERT(ds != NULL);
ASSERT(ds->ds_objset != NULL);
//zil_suspend_dmu_sync(dmu_objset_zil(os));
zil_suspend(dsname, &cookie);
arg.kn_skn = skn;
arg.kn_txgkey = zcrypt_key_gen(os->os_crypt);
arg.kn_ds = ds;
zcrypt_key_hold(skn->skn_wrapkey, FTAG);
VERIFY(zcrypt_wrap_key(skn->skn_wrapkey, arg.kn_txgkey,
&arg.kn_wkeybuf, &arg.kn_wkeylen,
zio_crypt_select_wrap(os->os_crypt)) == 0);
error = dsl_sync_task(spa->spa_name, dsl_crypto_key_new_check,
dsl_crypto_key_new_sync, &arg, 1,
ZFS_SPACE_CHECK_NONE);
kmem_free(arg.kn_wkeybuf, arg.kn_wkeylen);
zcrypt_key_release(skn->skn_wrapkey, FTAG);
//zil_resume_dmu_sync(dmu_objset_zil(os));
zil_resume(os);
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
if (error)
zcrypt_key_free(arg.kn_txgkey);
return (error);
}
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)) {
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", "");
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) {
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 (!dsl_dataset_is_snapshot(ds)) {
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) {
VERIFY(arc_buf_remove_ref(os->os_phys_buf,
&os->os_phys_buf));
kmem_free(os, sizeof (objset_t));
return (err);
}
} else if (ds == NULL) {
/* It's the meta-objset. */
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;
}
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);
}
/*
* Check if user has requested permission.
*/
int
dsl_deleg_access(const char *dsname, const char *perm, cred_t *cr)
{
dsl_dataset_t *ds;
dsl_dir_t *dd;
dsl_pool_t *dp;
void *cookie;
int error;
char checkflag;
objset_t *mos;
avl_tree_t permsets;
perm_set_t *setnode;
error = dsl_dataset_hold(dsname, FTAG, &ds);
if (error)
return (error);
dp = ds->ds_dir->dd_pool;
mos = dp->dp_meta_objset;
if (dsl_delegation_on(mos) == B_FALSE) {
dsl_dataset_rele(ds, FTAG);
return (ECANCELED);
}
if (spa_version(dmu_objset_spa(dp->dp_meta_objset)) <
SPA_VERSION_DELEGATED_PERMS) {
dsl_dataset_rele(ds, FTAG);
return (EPERM);
}
if (dsl_dataset_is_snapshot(ds)) {
/*
* Snapshots are treated as descendents only,
* local permissions do not apply.
*/
checkflag = ZFS_DELEG_DESCENDENT;
} else {
checkflag = ZFS_DELEG_LOCAL;
}
avl_create(&permsets, perm_set_compare, sizeof (perm_set_t),
offsetof(perm_set_t, p_node));
rw_enter(&dp->dp_config_rwlock, RW_READER);
for (dd = ds->ds_dir; dd != NULL; dd = dd->dd_parent,
checkflag = ZFS_DELEG_DESCENDENT) {
uint64_t zapobj;
boolean_t expanded;
/*
* If not in global zone then make sure
* the zoned property is set
*/
if (!INGLOBALZONE(curproc)) {
uint64_t zoned;
if (dsl_prop_get_dd(dd,
zfs_prop_to_name(ZFS_PROP_ZONED),
8, 1, &zoned, NULL) != 0)
break;
if (!zoned)
break;
}
zapobj = dd->dd_phys->dd_deleg_zapobj;
if (zapobj == 0)
continue;
dsl_load_user_sets(mos, zapobj, &permsets, checkflag, cr);
again:
expanded = B_FALSE;
for (setnode = avl_first(&permsets); setnode;
setnode = AVL_NEXT(&permsets, setnode)) {
if (setnode->p_matched == B_TRUE)
continue;
/* See if this set directly grants this permission */
error = dsl_check_access(mos, zapobj,
ZFS_DELEG_NAMED_SET, 0, setnode->p_setname, perm);
if (error == 0)
goto success;
if (error == EPERM)
setnode->p_matched = B_TRUE;
/* See if this set includes other sets */
error = dsl_load_sets(mos, zapobj,
ZFS_DELEG_NAMED_SET_SETS, 0,
setnode->p_setname, &permsets);
if (error == 0)
setnode->p_matched = expanded = B_TRUE;
}
/*
* If we expanded any sets, that will define more sets,
* which we need to check.
*/
if (expanded)
goto again;
error = dsl_check_user_access(mos, zapobj, perm, checkflag, cr);
if (error == 0)
goto success;
}
error = EPERM;
success:
rw_exit(&dp->dp_config_rwlock);
dsl_dataset_rele(ds, FTAG);
cookie = NULL;
while ((setnode = avl_destroy_nodes(&permsets, &cookie)) != NULL)
kmem_free(setnode, sizeof (perm_set_t));
return (error);
}
static int
dsl_dataset_user_release_check_one(dsl_dataset_user_release_arg_t *ddura,
dsl_dataset_t *ds, nvlist_t *holds, const char *snapname)
{
uint64_t zapobj;
nvlist_t *holds_found;
objset_t *mos;
int numholds;
if (!dsl_dataset_is_snapshot(ds))
return (SET_ERROR(EINVAL));
if (nvlist_empty(holds))
return (0);
numholds = 0;
mos = ds->ds_dir->dd_pool->dp_meta_objset;
zapobj = ds->ds_phys->ds_userrefs_obj;
holds_found = fnvlist_alloc();
for (nvpair_t *pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
uint64_t tmp;
int error;
const char *holdname = nvpair_name(pair);
if (zapobj != 0)
error = zap_lookup(mos, zapobj, holdname, 8, 1, &tmp);
else
error = SET_ERROR(ENOENT);
/*
* Non-existent holds are put on the errlist, but don't
* cause an overall failure.
*/
if (error == ENOENT) {
if (ddura->ddura_errlist != NULL) {
char *errtag = kmem_asprintf("%s#%s",
snapname, holdname);
fnvlist_add_int32(ddura->ddura_errlist, errtag,
ENOENT);
strfree(errtag);
}
continue;
}
if (error != 0) {
fnvlist_free(holds_found);
return (error);
}
fnvlist_add_boolean(holds_found, holdname);
numholds++;
}
if (DS_IS_DEFER_DESTROY(ds) && ds->ds_phys->ds_num_children == 1 &&
ds->ds_userrefs == numholds) {
/* we need to destroy the snapshot as well */
if (dsl_dataset_long_held(ds)) {
fnvlist_free(holds_found);
return (SET_ERROR(EBUSY));
}
fnvlist_add_boolean(ddura->ddura_todelete, snapname);
}
if (numholds != 0) {
fnvlist_add_nvlist(ddura->ddura_chkholds, snapname,
holds_found);
}
fnvlist_free(holds_found);
return (0);
}
void
dmu_objset_evict(objset_t *os)
{
dsl_dataset_t *ds = os->os_dsl_dataset;
for (int t = 0; t < TXG_SIZE; t++)
ASSERT(!dmu_objset_is_dirty(os, t));
if (ds) {
if (!dsl_dataset_is_snapshot(ds)) {
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_CHECKSUM),
checksum_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_COMPRESSION),
compression_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_COPIES),
copies_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_DEDUP),
dedup_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_LOGBIAS),
logbias_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_SYNC),
sync_changed_cb, os));
}
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
primary_cache_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE),
secondary_cache_changed_cb, os));
}
if (os->os_sa)
sa_tear_down(os);
dmu_objset_evict_dbufs(os);
dnode_special_close(&os->os_meta_dnode);
if (DMU_USERUSED_DNODE(os)) {
dnode_special_close(&os->os_userused_dnode);
dnode_special_close(&os->os_groupused_dnode);
}
zil_free(os->os_zil);
ASSERT3P(list_head(&os->os_dnodes), ==, NULL);
VERIFY(arc_buf_remove_ref(os->os_phys_buf, &os->os_phys_buf));
/*
* This is a barrier to prevent the objset from going away in
* dnode_move() until we can safely ensure that the objset is still in
* use. We consider the objset valid before the barrier and invalid
* after the barrier.
*/
rw_enter(&os_lock, RW_READER);
rw_exit(&os_lock);
mutex_destroy(&os->os_lock);
mutex_destroy(&os->os_obj_lock);
mutex_destroy(&os->os_user_ptr_lock);
kmem_free(os, sizeof (objset_t));
}
