void rwlock(pthread_rwlock_t *l, bool wr)
{
if (wr)
VERIFY0(pthread_rwlock_wrlock(l));
else
VERIFY0(pthread_rwlock_rdlock(l));
}
static int
zcp_props_list(lua_State *state)
{
const char *dsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
objset_t *os;
nvlist_t **props = lua_newuserdata(state, sizeof (nvlist_t *));
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
VERIFY0(dmu_objset_from_ds(ds, &os));
VERIFY0(dsl_prop_get_all(os, props));
dsl_dataset_rele(ds, FTAG);
/*
* Set the metatable for the properties list to free it on completion.
*/
luaL_getmetatable(state, zcp_props_list_info.name);
(void) lua_setmetatable(state, -2);
lua_pushlightuserdata(state, NULL);
lua_pushcclosure(state, &zcp_props_iter, 2);
return (1);
}
void
zfs_sa_symlink(znode_t *zp, char *link, int len, dmu_tx_t *tx)
{
dmu_buf_t *db = sa_get_db(zp->z_sa_hdl);
if (ZFS_OLD_ZNODE_PHYS_SIZE + len <= dmu_bonus_max()) {
VERIFY0(dmu_set_bonus(db, len + ZFS_OLD_ZNODE_PHYS_SIZE, tx));
if (len) {
bcopy(link, (caddr_t)db->db_data +
ZFS_OLD_ZNODE_PHYS_SIZE, len);
}
} else {
dmu_buf_t *dbp;
zfs_grow_blocksize(zp, len, tx);
VERIFY0(dmu_buf_hold(ZTOZSB(zp)->z_os, zp->z_id, 0, FTAG, &dbp,
DMU_READ_NO_PREFETCH));
dmu_buf_will_dirty(dbp, tx);
ASSERT3U(len, <=, dbp->db_size);
bcopy(link, dbp->db_data, len);
dmu_buf_rele(dbp, FTAG);
}
}
static void
dmu_objset_clone_sync(void *arg, dmu_tx_t *tx)
{
dmu_objset_clone_arg_t *doca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *pdd;
const char *tail;
dsl_dataset_t *origin, *ds;
uint64_t obj;
char namebuf[MAXNAMELEN];
VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail));
VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin));
obj = dsl_dataset_create_sync(pdd, tail, origin, 0,
doca->doca_cred, tx);
VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds));
dsl_dataset_name(origin, namebuf);
spa_history_log_internal_ds(ds, "clone", tx,
"origin=%s (%llu)", namebuf, origin->ds_object);
dsl_dataset_rele(ds, FTAG);
dsl_dataset_rele(origin, FTAG);
dsl_dir_rele(pdd, FTAG);
}
void
dsl_dataset_user_hold_sync_one(dsl_dataset_t *ds, const char *htag,
minor_t minor, uint64_t now, dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
uint64_t zapobj;
mutex_enter(&ds->ds_lock);
if (ds->ds_phys->ds_userrefs_obj == 0) {
/*
* This is the first user hold for this dataset. Create
* the userrefs zap object.
*/
dmu_buf_will_dirty(ds->ds_dbuf, tx);
zapobj = ds->ds_phys->ds_userrefs_obj =
zap_create(mos, DMU_OT_USERREFS, DMU_OT_NONE, 0, tx);
} else {
zapobj = ds->ds_phys->ds_userrefs_obj;
}
ds->ds_userrefs++;
mutex_exit(&ds->ds_lock);
VERIFY0(zap_add(mos, zapobj, htag, 8, 1, &now, tx));
if (minor != 0) {
VERIFY0(dsl_pool_user_hold(dp, ds->ds_object,
htag, now, tx));
dsl_register_onexit_hold_cleanup(ds, htag, minor);
}
spa_history_log_internal_ds(ds, "hold", tx,
"tag=%s temp=%d refs=%llu",
htag, minor != 0, ds->ds_userrefs);
}
static void
dmu_objset_create_sync(void *arg, dmu_tx_t *tx)
{
dmu_objset_create_arg_t *doca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
dsl_dir_t *pdd;
const char *tail;
dsl_dataset_t *ds;
uint64_t obj;
blkptr_t *bp;
objset_t *os;
VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail));
obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags,
doca->doca_cred, tx);
VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds));
bp = dsl_dataset_get_blkptr(ds);
os = dmu_objset_create_impl(pdd->dd_pool->dp_spa,
ds, bp, doca->doca_type, tx);
if (doca->doca_userfunc != NULL) {
doca->doca_userfunc(os, doca->doca_userarg,
doca->doca_cred, tx);
}
spa_history_log_internal_ds(ds, "create", tx, "");
dsl_dataset_rele(ds, FTAG);
dsl_dir_rele(pdd, FTAG);
}
static void
process_old_deadlist(dsl_dataset_t *ds, dsl_dataset_t *ds_prev,
dsl_dataset_t *ds_next, boolean_t after_branch_point, dmu_tx_t *tx)
{
struct process_old_arg poa = { 0 };
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
uint64_t deadlist_obj;
ASSERT(ds->ds_deadlist.dl_oldfmt);
ASSERT(ds_next->ds_deadlist.dl_oldfmt);
poa.ds = ds;
poa.ds_prev = ds_prev;
poa.after_branch_point = after_branch_point;
poa.pio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
VERIFY0(bpobj_iterate(&ds_next->ds_deadlist.dl_bpobj,
process_old_cb, &poa, tx));
VERIFY0(zio_wait(poa.pio));
ASSERT3U(poa.used, ==, ds->ds_phys->ds_unique_bytes);
/* change snapused */
dsl_dir_diduse_space(ds->ds_dir, DD_USED_SNAP,
-poa.used, -poa.comp, -poa.uncomp, tx);
/* swap next's deadlist to our deadlist */
dsl_deadlist_close(&ds->ds_deadlist);
dsl_deadlist_close(&ds_next->ds_deadlist);
deadlist_obj = ds->ds_phys->ds_deadlist_obj;
ds->ds_phys->ds_deadlist_obj = ds_next->ds_phys->ds_deadlist_obj;
ds_next->ds_phys->ds_deadlist_obj = deadlist_obj;
dsl_deadlist_open(&ds->ds_deadlist, mos, ds->ds_phys->ds_deadlist_obj);
dsl_deadlist_open(&ds_next->ds_deadlist, mos,
ds_next->ds_phys->ds_deadlist_obj);
}
/* 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);
}
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;
}
}
static void
dsl_bookmark_create_sync(void *arg, dmu_tx_t *tx)
{
dsl_bookmark_create_arg_t *dbca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
objset_t *mos = dp->dp_meta_objset;
nvpair_t *pair;
ASSERT(spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_BOOKMARKS));
for (pair = nvlist_next_nvpair(dbca->dbca_bmarks, NULL);
pair != NULL; pair = nvlist_next_nvpair(dbca->dbca_bmarks, pair)) {
dsl_dataset_t *snapds, *bmark_fs;
zfs_bookmark_phys_t bmark_phys;
char *shortname;
VERIFY0(dsl_dataset_hold(dp, fnvpair_value_string(pair),
FTAG, &snapds));
VERIFY0(dsl_bookmark_hold_ds(dp, nvpair_name(pair),
&bmark_fs, FTAG, &shortname));
if (bmark_fs->ds_bookmarks == 0) {
bmark_fs->ds_bookmarks =
zap_create_norm(mos, U8_TEXTPREP_TOUPPER,
DMU_OTN_ZAP_METADATA, DMU_OT_NONE, 0, tx);
spa_feature_incr(dp->dp_spa, SPA_FEATURE_BOOKMARKS, tx);
dsl_dataset_zapify(bmark_fs, tx);
VERIFY0(zap_add(mos, bmark_fs->ds_object,
DS_FIELD_BOOKMARK_NAMES,
sizeof (bmark_fs->ds_bookmarks), 1,
&bmark_fs->ds_bookmarks, tx));
}
bmark_phys.zbm_guid = dsl_dataset_phys(snapds)->ds_guid;
bmark_phys.zbm_creation_txg =
dsl_dataset_phys(snapds)->ds_creation_txg;
bmark_phys.zbm_creation_time =
dsl_dataset_phys(snapds)->ds_creation_time;
VERIFY0(zap_add(mos, bmark_fs->ds_bookmarks,
shortname, sizeof (uint64_t),
sizeof (zfs_bookmark_phys_t) / sizeof (uint64_t),
&bmark_phys, tx));
spa_history_log_internal_ds(bmark_fs, "bookmark", tx,
"name=%s creation_txg=%llu target_snap=%llu",
shortname,
(longlong_t)bmark_phys.zbm_creation_txg,
(longlong_t)snapds->ds_object);
dsl_dataset_rele(bmark_fs, FTAG);
dsl_dataset_rele(snapds, FTAG);
}
}
static void
dsl_dataset_user_hold_sync_one_impl(nvlist_t *tmpholds, dsl_dataset_t *ds,
const char *htag, minor_t minor, uint64_t now, dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
uint64_t zapobj;
ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
if (ds->ds_phys->ds_userrefs_obj == 0) {
/*
* This is the first user hold for this dataset. Create
* the userrefs zap object.
*/
dmu_buf_will_dirty(ds->ds_dbuf, tx);
zapobj = ds->ds_phys->ds_userrefs_obj =
zap_create(mos, DMU_OT_USERREFS, DMU_OT_NONE, 0, tx);
} else {
zapobj = ds->ds_phys->ds_userrefs_obj;
}
ds->ds_userrefs++;
VERIFY0(zap_add(mos, zapobj, htag, 8, 1, &now, tx));
if (minor != 0) {
char name[MAXNAMELEN];
nvlist_t *tags;
VERIFY0(dsl_pool_user_hold(dp, ds->ds_object,
htag, now, tx));
(void) snprintf(name, sizeof (name), "%llx",
(u_longlong_t)ds->ds_object);
if (nvlist_lookup_nvlist(tmpholds, name, &tags) != 0) {
VERIFY0(nvlist_alloc(&tags, NV_UNIQUE_NAME,
KM_PUSHPAGE));
fnvlist_add_boolean(tags, htag);
fnvlist_add_nvlist(tmpholds, name, tags);
fnvlist_free(tags);
} else {
fnvlist_add_boolean(tags, htag);
}
}
spa_history_log_internal_ds(ds, "hold", tx,
"tag=%s temp=%d refs=%llu",
htag, minor != 0, ds->ds_userrefs);
}
/*
* Sync task to begin the condensing process.
*/
void
spa_condense_indirect_start_sync(vdev_t *vd, dmu_tx_t *tx)
{
spa_t *spa = vd->vdev_spa;
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
ASSERT0(scip->scip_next_mapping_object);
ASSERT0(scip->scip_prev_obsolete_sm_object);
ASSERT0(scip->scip_vdev);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
ASSERT(spa_feature_is_active(spa, SPA_FEATURE_OBSOLETE_COUNTS));
ASSERT(vdev_indirect_mapping_num_entries(vd->vdev_indirect_mapping));
uint64_t obsolete_sm_obj = vdev_obsolete_sm_object(vd);
ASSERT(obsolete_sm_obj != 0);
scip->scip_vdev = vd->vdev_id;
scip->scip_next_mapping_object =
vdev_indirect_mapping_alloc(spa->spa_meta_objset, tx);
scip->scip_prev_obsolete_sm_object = obsolete_sm_obj;
/*
* We don't need to allocate a new space map object, since
* vdev_indirect_sync_obsolete will allocate one when needed.
*/
space_map_close(vd->vdev_obsolete_sm);
vd->vdev_obsolete_sm = NULL;
VERIFY0(zap_remove(spa->spa_meta_objset, vd->vdev_top_zap,
VDEV_TOP_ZAP_INDIRECT_OBSOLETE_SM, tx));
VERIFY0(zap_add(spa->spa_dsl_pool->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CONDENSING_INDIRECT, sizeof (uint64_t),
sizeof (*scip) / sizeof (uint64_t), scip, tx));
ASSERT3P(spa->spa_condensing_indirect, ==, NULL);
spa->spa_condensing_indirect = spa_condensing_indirect_create(spa);
zfs_dbgmsg("starting condense of vdev %llu in txg %llu: "
"posm=%llu nm=%llu",
vd->vdev_id, dmu_tx_get_txg(tx),
(u_longlong_t)scip->scip_prev_obsolete_sm_object,
(u_longlong_t)scip->scip_next_mapping_object);
zthr_wakeup(spa->spa_condense_zthr);
}
uint64_t
fnvpair_value_uint64(nvpair_t *nvp)
{
uint64_t rv;
VERIFY0(nvpair_value_uint64(nvp, &rv));
return (rv);
}
int32_t
fnvpair_value_int32(nvpair_t *nvp)
{
int32_t rv;
VERIFY0(nvpair_value_int32(nvp, &rv));
return (rv);
}
uchar_t
fnvpair_value_byte(nvpair_t *nvp)
{
uchar_t rv;
VERIFY0(nvpair_value_byte(nvp, &rv));
return (rv);
}
boolean_t
fnvpair_value_boolean_value(nvpair_t *nvp)
{
boolean_t rv;
VERIFY0(nvpair_value_boolean_value(nvp, &rv));
return (rv);
}
nvlist_t *
fnvlist_lookup_nvlist(nvlist_t *nvl, const char *name)
{
nvlist_t *rv;
VERIFY0(nvlist_lookup_nvlist(nvl, name, &rv));
return (rv);
}
char *
fnvlist_lookup_string(nvlist_t *nvl, const char *name)
{
char *rv;
VERIFY0(nvlist_lookup_string(nvl, name, &rv));
return (rv);
}
uint64_t
fnvlist_lookup_uint64(nvlist_t *nvl, const char *name)
{
uint64_t rv;
VERIFY0(nvlist_lookup_uint64(nvl, name, &rv));
return (rv);
}
int32_t
fnvlist_lookup_int32(nvlist_t *nvl, const char *name)
{
int32_t rv;
VERIFY0(nvlist_lookup_int32(nvl, name, &rv));
return (rv);
}
nvlist_t *
fnvlist_unpack(char *buf, size_t buflen)
{
nvlist_t *rv;
VERIFY0(nvlist_unpack(buf, buflen, &rv, KM_SLEEP));
return (rv);
}
nvlist_t *
fnvlist_alloc(void)
{
nvlist_t *nvl;
VERIFY0(nvlist_alloc(&nvl, NV_UNIQUE_NAME, KM_SLEEP));
return (nvl);
}
char *
fnvpair_value_string(nvpair_t *nvp)
{
char *rv;
VERIFY0(nvpair_value_string(nvp, &rv));
return (rv);
}
void
dsl_pool_create_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx)
{
uint64_t obj;
/*
* Currently, we only create the obsolete_bpobj where there are
* indirect vdevs with referenced mappings.
*/
ASSERT(spa_feature_is_active(dp->dp_spa, SPA_FEATURE_DEVICE_REMOVAL));
/* create and open the obsolete_bpobj */
obj = bpobj_alloc(dp->dp_meta_objset, SPA_OLD_MAXBLOCKSIZE, tx);
VERIFY0(bpobj_open(&dp->dp_obsolete_bpobj, dp->dp_meta_objset, obj));
VERIFY0(zap_add(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ, sizeof (uint64_t), 1, &obj, tx));
spa_feature_incr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
}
nvlist_t *
fnvpair_value_nvlist(nvpair_t *nvp)
{
nvlist_t *rv;
VERIFY0(nvpair_value_nvlist(nvp, &rv));
return (rv);
}
nvlist_t *
fnvlist_dup(nvlist_t *nvl)
{
nvlist_t *rv;
VERIFY0(nvlist_dup(nvl, &rv, KM_SLEEP));
return (rv);
}
size_t
fnvlist_size(nvlist_t *nvl)
{
size_t size;
VERIFY0(nvlist_size(nvl, &size, NV_ENCODE_NATIVE));
return (size);
}
boolean_t
fnvlist_lookup_boolean_value(nvlist_t *nvl, const char *name)
{
boolean_t rv;
VERIFY0(nvlist_lookup_boolean_value(nvl, name, &rv));
return (rv);
}
static void
dsl_dataset_user_release_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_release_arg_t *ddura = arg;
dsl_holdfunc_t *holdfunc = ddura->ddura_holdfunc;
dsl_pool_t *dp = dmu_tx_pool(tx);
ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
for (nvpair_t *pair = nvlist_next_nvpair(ddura->ddura_chkholds, NULL);
pair != NULL; pair = nvlist_next_nvpair(ddura->ddura_chkholds,
pair)) {
dsl_dataset_t *ds;
const char *name = nvpair_name(pair);
VERIFY0(holdfunc(dp, name, FTAG, &ds));
dsl_dataset_user_release_sync_one(ds,
fnvpair_value_nvlist(pair), tx);
if (nvlist_exists(ddura->ddura_todelete, name)) {
ASSERT(ds->ds_userrefs == 0 &&
ds->ds_phys->ds_num_children == 1 &&
DS_IS_DEFER_DESTROY(ds));
dsl_destroy_snapshot_sync_impl(ds, B_FALSE, tx);
}
dsl_dataset_rele(ds, FTAG);
}
}
uchar_t
fnvlist_lookup_byte(nvlist_t *nvl, const char *name)
{
uchar_t rv;
VERIFY0(nvlist_lookup_byte(nvl, name, &rv));
return (rv);
}
