static int
zcp_clones_iter(lua_State *state)
{
int err;
char clonename[ZFS_MAX_DATASET_NAME_LEN];
uint64_t dsobj = lua_tonumber(state, lua_upvalueindex(1));
uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2));
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
dsl_dataset_t *ds, *clone;
zap_attribute_t za;
zap_cursor_t zc;
err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (err == ENOENT) {
return (0);
} else if (err != 0) {
return (luaL_error(state,
"unexpected error %d from dsl_dataset_hold_obj(dsobj)",
err));
}
if (dsl_dataset_phys(ds)->ds_next_clones_obj == 0) {
dsl_dataset_rele(ds, FTAG);
return (0);
}
zap_cursor_init_serialized(&zc, dp->dp_meta_objset,
dsl_dataset_phys(ds)->ds_next_clones_obj, cursor);
dsl_dataset_rele(ds, FTAG);
err = zap_cursor_retrieve(&zc, &za);
if (err != 0) {
zap_cursor_fini(&zc);
if (err != ENOENT) {
return (luaL_error(state,
"unexpected error %d from zap_cursor_retrieve()",
err));
}
return (0);
}
zap_cursor_advance(&zc);
cursor = zap_cursor_serialize(&zc);
zap_cursor_fini(&zc);
err = dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &clone);
if (err != 0) {
return (luaL_error(state,
"unexpected error %d from "
"dsl_dataset_hold_obj(za_first_integer)", err));
}
dsl_dir_name(clone->ds_dir, clonename);
dsl_dataset_rele(clone, FTAG);
lua_pushnumber(state, cursor);
lua_replace(state, lua_upvalueindex(2));
(void) lua_pushstring(state, clonename);
return (1);
}
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 (!snapds->ds_is_snapshot)
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);
}
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);
}
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);
}
}
/*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);
}
static void
dsl_dataset_user_hold_sync(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_hold_arg_t *dduha = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvlist_t *tmpholds;
nvpair_t *pair;
uint64_t now = gethrestime_sec();
if (dduha->dduha_minor != 0)
tmpholds = fnvlist_alloc();
else
tmpholds = NULL;
for (pair = nvlist_next_nvpair(dduha->dduha_chkholds, NULL);
pair != NULL;
pair = nvlist_next_nvpair(dduha->dduha_chkholds, pair)) {
dsl_dataset_t *ds;
VERIFY0(dsl_dataset_hold(dp, nvpair_name(pair), FTAG, &ds));
dsl_dataset_user_hold_sync_one_impl(tmpholds, ds,
fnvpair_value_string(pair), dduha->dduha_minor, now, tx);
dsl_dataset_rele(ds, FTAG);
}
dsl_onexit_hold_cleanup(dp->dp_spa, tmpholds, dduha->dduha_minor);
}
static int
zcp_exists(lua_State *state)
{
zcp_run_info_t *ri = zcp_run_info(state);
dsl_pool_t *dp = ri->zri_pool;
zcp_lib_info_t *libinfo = &zcp_exists_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
const char *dsname = lua_tostring(state, 1);
dsl_dataset_t *ds;
int error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
if (error == 0) {
dsl_dataset_rele(ds, FTAG);
lua_pushboolean(state, B_TRUE);
} else if (error == ENOENT) {
lua_pushboolean(state, B_FALSE);
} else if (error == EXDEV) {
return (luaL_error(state, "dataset '%s' is not in the "
"target pool", dsname));
} else if (error == EIO) {
return (luaL_error(state, "I/O error opening dataset '%s'",
dsname));
} else if (error != 0) {
return (luaL_error(state, "unexpected error %d", error));
}
return (1);
}
int
dsl_destroy_snapshot_check(void *arg, dmu_tx_t *tx)
{
dsl_destroy_snapshot_arg_t *ddsa = arg;
const char *dsname = ddsa->ddsa_name;
boolean_t defer = ddsa->ddsa_defer;
dsl_pool_t *dp = dmu_tx_pool(tx);
int error = 0;
dsl_dataset_t *ds;
error = dsl_dataset_hold(dp, dsname, FTAG, &ds);
/*
* If the snapshot does not exist, silently ignore it, and
* dsl_destroy_snapshot_sync() will be a no-op
* (it's "already destroyed").
*/
if (error == ENOENT)
return (0);
if (error == 0) {
error = dsl_destroy_snapshot_check_impl(ds, defer);
dsl_dataset_rele(ds, FTAG);
}
return (error);
}
void
dsl_dir_stats(dsl_dir_t *dd, nvlist_t *nv)
{
mutex_enter(&dd->dd_lock);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USED, dd->dd_used_bytes);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_QUOTA, dd->dd_phys->dd_quota);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_RESERVATION,
dd->dd_phys->dd_reserved);
dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_COMPRESSRATIO,
dd->dd_phys->dd_compressed_bytes == 0 ? 100 :
(dd->dd_phys->dd_uncompressed_bytes * 100 /
dd->dd_phys->dd_compressed_bytes));
mutex_exit(&dd->dd_lock);
rw_enter(&dd->dd_pool->dp_config_rwlock, RW_READER);
if (dsl_dir_is_clone(dd)) {
dsl_dataset_t *ds;
char buf[MAXNAMELEN];
VERIFY(0 == dsl_dataset_hold_obj(dd->dd_pool,
dd->dd_phys->dd_origin_obj, FTAG, &ds));
dsl_dataset_name(ds, buf);
dsl_dataset_rele(ds, FTAG);
dsl_prop_nvlist_add_string(nv, ZFS_PROP_ORIGIN, buf);
}
rw_exit(&dd->dd_pool->dp_config_rwlock);
}
int
dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
{
int err;
dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
dsl_dir_t *dd;
dsl_dataset_t *ds;
uint64_t obj;
rw_enter(&dp->dp_config_rwlock, RW_WRITER);
err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp,
&dp->dp_meta_objset);
if (err)
goto out;
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
&dp->dp_root_dir_obj);
if (err)
goto out;
err = dsl_dir_open_obj(dp, dp->dp_root_dir_obj,
NULL, dp, &dp->dp_root_dir);
if (err)
goto out;
err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
if (err)
goto out;
if (spa_version(spa) >= SPA_VERSION_ORIGIN) {
err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
if (err)
goto out;
err = dsl_dataset_hold_obj(dp, dd->dd_phys->dd_head_dataset_obj,
FTAG, &ds);
if (err == 0) {
err = dsl_dataset_hold_obj(dp,
ds->ds_phys->ds_prev_snap_obj, dp,
&dp->dp_origin_snap);
dsl_dataset_rele(ds, FTAG);
}
dsl_dir_close(dd, dp);
if (err)
goto out;
}
if (spa_version(spa) >= SPA_VERSION_DEADLISTS) {
err = dsl_pool_open_special_dir(dp, FREE_DIR_NAME,
&dp->dp_free_dir);
if (err)
goto out;
err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_FREE_BPOBJ, sizeof (uint64_t), 1, &obj);
if (err)
goto out;
VERIFY3U(0, ==, bpobj_open(&dp->dp_free_bpobj,
dp->dp_meta_objset, obj));
}
static void
dmu_objset_create_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
struct oscarg *oa = arg2;
dsl_dataset_t *ds;
blkptr_t *bp;
uint64_t dsobj;
ASSERT(dmu_tx_is_syncing(tx));
dsobj = dsl_dataset_create_sync(dd, oa->lastname,
oa->clone_parent, oa->flags, cr, tx);
VERIFY(0 == dsl_dataset_hold_obj(dd->dd_pool, dsobj, FTAG, &ds));
bp = dsl_dataset_get_blkptr(ds);
if (BP_IS_HOLE(bp)) {
objset_impl_t *osi;
/* This is an empty dmu_objset; not a clone. */
osi = dmu_objset_create_impl(dsl_dataset_get_spa(ds),
ds, bp, oa->type, tx);
if (oa->userfunc)
oa->userfunc(&osi->os, oa->userarg, cr, tx);
}
spa_history_internal_log(LOG_DS_CREATE, dd->dd_pool->dp_spa,
tx, cr, "dataset = %llu", dsobj);
dsl_dataset_rele(ds, FTAG);
}
/*
* Get a list of all visble properties and their values for a given dataset.
* Returned on the stack as a Lua table.
*/
static int
zcp_system_props_list(lua_State *state)
{
int error;
char errbuf[128];
const char *dataset_name;
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
zcp_list_info_t *libinfo = &zcp_system_props_list_info;
zcp_parse_args(state, libinfo->name, libinfo->pargs, libinfo->kwargs);
dataset_name = lua_tostring(state, 1);
nvlist_t *nv = fnvlist_alloc();
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, dataset_name, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
/* Get the names of all valid properties for this dataset */
zcp_dataset_props(ds, nv);
dsl_dataset_rele(ds, FTAG);
/* push list as lua table */
error = zcp_nvlist_to_lua(state, nv, errbuf, sizeof (errbuf));
nvlist_free(nv);
if (error != 0) {
return (luaL_error(state,
"Error returning nvlist: %s", errbuf));
}
return (1);
}
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);
}
static int
zcp_children_list(lua_State *state)
{
const char *fsname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
boolean_t issnap;
uint64_t dsobj;
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, fsname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
issnap = ds->ds_is_snapshot;
dsobj = ds->ds_object;
dsl_dataset_rele(ds, FTAG);
if (issnap) {
return (zcp_argerror(state, 1,
"argument %s cannot be a snapshot", fsname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, 0);
lua_pushcclosure(state, &zcp_children_iter, 2);
return (1);
}
static int
zcp_clones_list(lua_State *state)
{
const char *snapname = lua_tostring(state, 1);
dsl_pool_t *dp = zcp_run_info(state)->zri_pool;
boolean_t issnap;
uint64_t dsobj, cursor;
/*
* zcp_dataset_hold will either successfully return the requested
* dataset or throw a lua error and longjmp out of the zfs.list.clones
* call without returning.
*/
dsl_dataset_t *ds = zcp_dataset_hold(state, dp, snapname, FTAG);
if (ds == NULL)
return (1); /* not reached; zcp_dataset_hold() longjmp'd */
cursor = 0;
issnap = ds->ds_is_snapshot;
dsobj = ds->ds_object;
dsl_dataset_rele(ds, FTAG);
if (!issnap) {
return (zcp_argerror(state, 1, "%s is not a snapshot",
snapname));
}
lua_pushnumber(state, dsobj);
lua_pushnumber(state, cursor);
lua_pushcclosure(state, &zcp_clones_iter, 2);
return (1);
}
/* ARGSUSED */
int
dmu_objset_prefetch(char *name, void *arg)
{
dsl_dataset_t *ds;
if (dsl_dataset_hold(name, FTAG, &ds))
return (0);
if (!BP_IS_HOLE(&ds->ds_phys->ds_bp)) {
mutex_enter(&ds->ds_opening_lock);
if (!dsl_dataset_get_user_ptr(ds)) {
uint32_t aflags = ARC_NOWAIT | ARC_PREFETCH;
zbookmark_t zb;
zb.zb_objset = ds->ds_object;
zb.zb_object = 0;
zb.zb_level = -1;
zb.zb_blkid = 0;
(void) arc_read_nolock(NULL, dsl_dataset_get_spa(ds),
&ds->ds_phys->ds_bp, NULL, NULL,
ZIO_PRIORITY_ASYNC_READ,
ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE,
&aflags, &zb);
}
mutex_exit(&ds->ds_opening_lock);
}
dsl_dataset_rele(ds, FTAG);
return (0);
}
/* called from zpl */
int
dmu_objset_open(const char *name, dmu_objset_type_t type, int mode,
objset_t **osp)
{
objset_t *os;
dsl_dataset_t *ds;
int err;
ASSERT(DS_MODE_TYPE(mode) == DS_MODE_USER ||
DS_MODE_TYPE(mode) == DS_MODE_OWNER);
os = kmem_alloc(sizeof (objset_t), KM_SLEEP);
if (DS_MODE_TYPE(mode) == DS_MODE_USER)
err = dsl_dataset_hold(name, os, &ds);
else
err = dsl_dataset_own(name, mode, os, &ds);
if (err) {
kmem_free(os, sizeof (objset_t));
return (err);
}
err = dmu_objset_open_ds_os(ds, os, type);
if (err) {
if (DS_MODE_TYPE(mode) == DS_MODE_USER)
dsl_dataset_rele(ds, os);
else
dsl_dataset_disown(ds, os);
kmem_free(os, sizeof (objset_t));
} else {
os->os_mode = mode;
*osp = os;
}
return (err);
}
int
dsl_dataset_get_holds(const char *dsname, nvlist_t *nvl)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
int err;
err = dsl_pool_hold(dsname, FTAG, &dp);
if (err != 0)
return (err);
err = dsl_dataset_hold(dp, dsname, FTAG, &ds);
if (err != 0) {
dsl_pool_rele(dp, FTAG);
return (err);
}
if (ds->ds_phys->ds_userrefs_obj != 0) {
zap_attribute_t *za;
zap_cursor_t zc;
za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP);
for (zap_cursor_init(&zc, ds->ds_dir->dd_pool->dp_meta_objset,
ds->ds_phys->ds_userrefs_obj);
zap_cursor_retrieve(&zc, za) == 0;
zap_cursor_advance(&zc)) {
fnvlist_add_uint64(nvl, za->za_name,
za->za_first_integer);
}
zap_cursor_fini(&zc);
kmem_free(za, sizeof (zap_attribute_t));
}
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (0);
}
static void
dsl_dataset_remove_clones_key(dsl_dataset_t *ds, uint64_t mintxg, dmu_tx_t *tx)
{
objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset;
zap_cursor_t zc;
zap_attribute_t za;
/*
* If it is the old version, dd_clones doesn't exist so we can't
* find the clones, but dsl_deadlist_remove_key() is a no-op so it
* doesn't matter.
*/
if (dsl_dir_phys(ds->ds_dir)->dd_clones == 0)
return;
for (zap_cursor_init(&zc, mos, dsl_dir_phys(ds->ds_dir)->dd_clones);
zap_cursor_retrieve(&zc, &za) == 0;
zap_cursor_advance(&zc)) {
dsl_dataset_t *clone;
VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool,
za.za_first_integer, FTAG, &clone));
if (clone->ds_dir->dd_origin_txg > mintxg) {
dsl_deadlist_remove_key(&clone->ds_deadlist,
mintxg, tx);
dsl_dataset_remove_clones_key(clone, mintxg, tx);
}
dsl_dataset_rele(clone, FTAG);
}
zap_cursor_fini(&zc);
}
/*
* Called at spa_load time to release a stale temporary user hold.
* Also called by the onexit code.
*/
void
dsl_dataset_user_release_tmp(dsl_pool_t *dp, uint64_t dsobj, const char *htag)
{
dsl_dataset_user_release_tmp_arg_t ddurta;
#ifdef _KERNEL
dsl_dataset_t *ds;
int error;
/* Make sure it is not mounted. */
dsl_pool_config_enter(dp, FTAG);
error = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
if (error == 0) {
char name[MAXNAMELEN];
dsl_dataset_name(ds, name);
dsl_dataset_rele(ds, FTAG);
dsl_pool_config_exit(dp, FTAG);
zfs_unmount_snap(name);
} else {
dsl_pool_config_exit(dp, FTAG);
}
#endif
ddurta.ddurta_dsobj = dsobj;
ddurta.ddurta_holds = fnvlist_alloc();
fnvlist_add_boolean(ddurta.ddurta_holds, htag);
(void) dsl_sync_task(spa_name(dp->dp_spa),
dsl_dataset_user_release_tmp_check,
dsl_dataset_user_release_tmp_sync, &ddurta, 1);
fnvlist_free(ddurta.ddurta_holds);
}
static void
dmu_objset_create_sync(void *arg1, void *arg2, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
spa_t *spa = dd->dd_pool->dp_spa;
struct oscarg *oa = arg2;
uint64_t obj;
ASSERT(dmu_tx_is_syncing(tx));
obj = dsl_dataset_create_sync(dd, oa->lastname,
oa->clone_origin, oa->crypto_ctx, oa->flags, oa->cr, tx);
if (oa->clone_origin == NULL) {
dsl_pool_t *dp = dd->dd_pool;
dsl_dataset_t *ds;
blkptr_t *bp;
objset_t *os;
VERIFY3U(0, ==, dsl_dataset_hold_obj(dp, obj, FTAG, &ds));
bp = dsl_dataset_get_blkptr(ds);
ASSERT(BP_IS_HOLE(bp));
os = dmu_objset_create_impl(spa, ds, bp, oa->type,
oa->crypto_ctx, tx);
if (oa->userfunc)
oa->userfunc(os, oa->userarg, oa->cr, tx);
dsl_dataset_rele(ds, FTAG);
}
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);
}
}
void
dmu_objset_rele(objset_t *os, void *tag)
{
dsl_pool_t *dp = dmu_objset_pool(os);
dsl_dataset_rele(os->os_dsl_dataset, tag);
dsl_pool_rele(dp, tag);
}
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);
}
void
dsl_pool_close(dsl_pool_t *dp)
{
/*
* Drop our references from dsl_pool_open().
*
* Since we held the origin_snap from "syncing" context (which
* includes pool-opening context), it actually only got a "ref"
* and not a hold, so just drop that here.
*/
if (dp->dp_origin_snap != NULL)
dsl_dataset_rele(dp->dp_origin_snap, dp);
if (dp->dp_mos_dir != NULL)
dsl_dir_rele(dp->dp_mos_dir, dp);
if (dp->dp_free_dir != NULL)
dsl_dir_rele(dp->dp_free_dir, dp);
if (dp->dp_leak_dir != NULL)
dsl_dir_rele(dp->dp_leak_dir, dp);
if (dp->dp_root_dir != NULL)
dsl_dir_rele(dp->dp_root_dir, dp);
bpobj_close(&dp->dp_free_bpobj);
bpobj_close(&dp->dp_obsolete_bpobj);
/* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
if (dp->dp_meta_objset != NULL)
dmu_objset_evict(dp->dp_meta_objset);
txg_list_destroy(&dp->dp_dirty_datasets);
txg_list_destroy(&dp->dp_dirty_zilogs);
txg_list_destroy(&dp->dp_sync_tasks);
txg_list_destroy(&dp->dp_early_sync_tasks);
txg_list_destroy(&dp->dp_dirty_dirs);
taskq_destroy(dp->dp_zil_clean_taskq);
taskq_destroy(dp->dp_sync_taskq);
/*
* We can't set retry to TRUE since we're explicitly specifying
* a spa to flush. This is good enough; any missed buffers for
* this spa won't cause trouble, and they'll eventually fall
* out of the ARC just like any other unused buffer.
*/
arc_flush(dp->dp_spa, FALSE);
mmp_fini(dp->dp_spa);
txg_fini(dp);
dsl_scan_fini(dp);
dmu_buf_user_evict_wait();
rrw_destroy(&dp->dp_config_rwlock);
mutex_destroy(&dp->dp_lock);
cv_destroy(&dp->dp_spaceavail_cv);
taskq_destroy(dp->dp_iput_taskq);
if (dp->dp_blkstats != NULL) {
mutex_destroy(&dp->dp_blkstats->zab_lock);
vmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
}
kmem_free(dp, sizeof (dsl_pool_t));
}
static int
dsl_bookmark_create_check(void *arg, dmu_tx_t *tx)
{
dsl_bookmark_create_arg_t *dbca = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
int rv = 0;
if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_BOOKMARKS))
return (SET_ERROR(ENOTSUP));
for (nvpair_t *pair = nvlist_next_nvpair(dbca->dbca_bmarks, NULL);
pair != NULL; pair = nvlist_next_nvpair(dbca->dbca_bmarks, pair)) {
dsl_dataset_t *snapds;
int error;
/* note: validity of nvlist checked by ioctl layer */
error = dsl_dataset_hold(dp, fnvpair_value_string(pair),
FTAG, &snapds);
if (error == 0) {
error = dsl_bookmark_create_check_impl(snapds,
nvpair_name(pair), tx);
dsl_dataset_rele(snapds, FTAG);
}
if (error != 0) {
fnvlist_add_int32(dbca->dbca_errors,
nvpair_name(pair), error);
rv = error;
}
}
return (rv);
}
/* ARGSUSED */
static int
dsl_prop_notify_all_cb(dsl_pool_t *dp, dsl_dataset_t *ds, void *arg)
{
dsl_dir_t *dd = ds->ds_dir;
dsl_prop_record_t *pr;
dsl_prop_cb_record_t *cbr;
mutex_enter(&dd->dd_lock);
for (pr = list_head(&dd->dd_props);
pr; pr = list_next(&dd->dd_props, pr)) {
for (cbr = list_head(&pr->pr_cbs); cbr;
cbr = list_next(&pr->pr_cbs, cbr)) {
uint64_t value;
/*
* Callback entries do not have holds on their
* datasets so that datasets with registered
* callbacks are still eligible for eviction.
* Unlike operations to update properties on a
* single dataset, we are performing a recursive
* descent of related head datasets. The caller
* of this function only has a dataset hold on
* the passed in head dataset, not the snapshots
* associated with this dataset. Without a hold,
* the dataset pointer within callback records
* for snapshots can be invalidated by eviction
* at any time.
*
* Use dsl_dataset_try_add_ref() to verify
* that the dataset for a snapshot has not
* begun eviction processing and to prevent
* eviction from occurring for the duration of
* the callback. If the hold attempt fails,
* this object is already being evicted and the
* callback can be safely ignored.
*/
if (ds != cbr->cbr_ds &&
!dsl_dataset_try_add_ref(dp, cbr->cbr_ds, FTAG))
continue;
if (dsl_prop_get_ds(cbr->cbr_ds,
cbr->cbr_pr->pr_propname, sizeof (value), 1,
&value, NULL) == 0)
cbr->cbr_func(cbr->cbr_arg, value);
if (ds != cbr->cbr_ds)
dsl_dataset_rele(cbr->cbr_ds, FTAG);
}
}
mutex_exit(&dd->dd_lock);
return (0);
}
void
dmu_objset_close(objset_t *os)
{
ASSERT(DS_MODE_TYPE(os->os_mode) == DS_MODE_USER ||
DS_MODE_TYPE(os->os_mode) == DS_MODE_OWNER ||
DS_MODE_TYPE(os->os_mode) == DS_MODE_NOHOLD);
if (DS_MODE_TYPE(os->os_mode) == DS_MODE_USER)
dsl_dataset_rele(os->os->os_dsl_dataset, os);
else if (DS_MODE_TYPE(os->os_mode) == DS_MODE_OWNER)
dsl_dataset_disown(os->os->os_dsl_dataset, os);
kmem_free(os, sizeof (objset_t));
}
static int
dsl_dataset_user_release_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_release_arg_t *ddura;
dsl_holdfunc_t *holdfunc;
dsl_pool_t *dp;
nvpair_t *pair;
if (!dmu_tx_is_syncing(tx))
return (0);
dp = dmu_tx_pool(tx);
ASSERT(RRW_WRITE_HELD(&dp->dp_config_rwlock));
ddura = arg;
holdfunc = ddura->ddura_holdfunc;
for (pair = nvlist_next_nvpair(ddura->ddura_holds, NULL);
pair != NULL; pair = nvlist_next_nvpair(ddura->ddura_holds, pair)) {
int error;
dsl_dataset_t *ds;
nvlist_t *holds;
const char *snapname = nvpair_name(pair);
error = nvpair_value_nvlist(pair, &holds);
if (error != 0)
error = (SET_ERROR(EINVAL));
else
error = holdfunc(dp, snapname, FTAG, &ds);
if (error == 0) {
error = dsl_dataset_user_release_check_one(ddura, ds,
holds, snapname);
dsl_dataset_rele(ds, FTAG);
}
if (error != 0) {
if (ddura->ddura_errlist != NULL) {
fnvlist_add_int32(ddura->ddura_errlist,
snapname, error);
}
/*
* Non-existent snapshots are put on the errlist,
* but don't cause an overall failure.
*/
if (error != ENOENT)
return (error);
}
}
return (0);
}
static int
dsl_dataset_user_hold_check(void *arg, dmu_tx_t *tx)
{
dsl_dataset_user_hold_arg_t *dduha = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
nvpair_t *pair;
if (spa_version(dp->dp_spa) < SPA_VERSION_USERREFS)
return (SET_ERROR(ENOTSUP));
if (!dmu_tx_is_syncing(tx))
return (0);
for (pair = nvlist_next_nvpair(dduha->dduha_holds, NULL);
pair != NULL; pair = nvlist_next_nvpair(dduha->dduha_holds, pair)) {
dsl_dataset_t *ds;
int error = 0;
char *htag, *name;
/* must be a snapshot */
name = nvpair_name(pair);
if (strchr(name, '@') == NULL)
error = SET_ERROR(EINVAL);
if (error == 0)
error = nvpair_value_string(pair, &htag);
if (error == 0)
error = dsl_dataset_hold(dp, name, FTAG, &ds);
if (error == 0) {
error = dsl_dataset_user_hold_check_one(ds, htag,
dduha->dduha_minor != 0, tx);
dsl_dataset_rele(ds, FTAG);
}
if (error == 0) {
fnvlist_add_string(dduha->dduha_chkholds, name, htag);
} else {
/*
* We register ENOENT errors so they can be correctly
* reported if needed, such as when all holds fail.
*/
fnvlist_add_int32(dduha->dduha_errlist, name, error);
if (error != ENOENT)
return (error);
}
}
return (0);
}
