static void
dsl_deleg_unset_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) {
dsl_dir_rele(dd, FTAG);
return;
}
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;
if (nvpair_value_nvlist(whopair, &perms) != 0) {
if (zap_lookup(mos, zapobj, whokey, 8,
1, &jumpobj) == 0) {
(void) zap_remove(mos, zapobj, whokey, tx);
VERIFY(0 == zap_destroy(mos, jumpobj, tx));
}
spa_history_log_internal_dd(dd, "permission who remove",
tx, "%s", whokey);
continue;
}
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0)
continue;
while ((permpair = nvlist_next_nvpair(perms, permpair))) {
const char *perm = nvpair_name(permpair);
uint64_t n = 0;
(void) zap_remove(mos, jumpobj, perm, tx);
if (zap_count(mos, jumpobj, &n) == 0 && n == 0) {
(void) zap_remove(mos, zapobj,
whokey, tx);
VERIFY(0 == zap_destroy(mos,
jumpobj, tx));
}
spa_history_log_internal_dd(dd, "permission remove", tx,
"%s %s", whokey, perm);
}
}
dsl_dir_rele(dd, FTAG);
}
static void
dsl_deleg_unset_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)
return;
while (whopair = nvlist_next_nvpair(nvp, whopair)) {
const char *whokey = nvpair_name(whopair);
nvlist_t *perms;
nvpair_t *permpair = NULL;
uint64_t jumpobj;
if (nvpair_value_nvlist(whopair, &perms) != 0) {
if (zap_lookup(mos, zapobj, whokey, 8,
1, &jumpobj) == 0) {
(void) zap_remove(mos, zapobj, whokey, tx);
VERIFY(0 == zap_destroy(mos, jumpobj, tx));
}
spa_history_internal_log(LOG_DS_PERM_WHO_REMOVE,
dd->dd_pool->dp_spa, tx, cr,
"%s dataset = %llu", whokey,
dd->dd_phys->dd_head_dataset_obj);
continue;
}
if (zap_lookup(mos, zapobj, whokey, 8, 1, &jumpobj) != 0)
continue;
while (permpair = nvlist_next_nvpair(perms, permpair)) {
const char *perm = nvpair_name(permpair);
uint64_t n = 0;
(void) zap_remove(mos, jumpobj, perm, tx);
if (zap_count(mos, jumpobj, &n) == 0 && n == 0) {
(void) zap_remove(mos, zapobj,
whokey, tx);
VERIFY(0 == zap_destroy(mos,
jumpobj, tx));
}
spa_history_internal_log(LOG_DS_PERM_REMOVE,
dd->dd_pool->dp_spa, tx, cr,
"%s %s dataset = %llu", whokey, perm,
dd->dd_phys->dd_head_dataset_obj);
}
}
}
/*
* The match type in the code for this function should conform to:
*
* ------------------------------------------------------------------------
* fs type | z_norm | lookup type | match type
* ---------|-------------|-------------|----------------------------------
* CS !norm | 0 | 0 | 0 (exact)
* CS norm | formX | 0 | MT_NORMALIZE
* CI !norm | upper | !ZCIEXACT | MT_NORMALIZE
* CI !norm | upper | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
* CI norm | upper|formX | !ZCIEXACT | MT_NORMALIZE
* CI norm | upper|formX | ZCIEXACT | MT_NORMALIZE | MT_MATCH_CASE
* CM !norm | upper | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
* CM !norm | upper | ZCILOOK | MT_NORMALIZE
* CM norm | upper|formX | !ZCILOOK | MT_NORMALIZE | MT_MATCH_CASE
* CM norm | upper|formX | ZCILOOK | MT_NORMALIZE
*
* Abbreviations:
* CS = Case Sensitive, CI = Case Insensitive, CM = Case Mixed
* upper = case folding set by fs type on creation (U8_TEXTPREP_TOUPPER)
* formX = unicode normalization form set on fs creation
*/
static int
zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx,
int flag)
{
int error;
if (ZTOZSB(zp)->z_norm) {
matchtype_t mt = MT_NORMALIZE;
if ((ZTOZSB(zp)->z_case == ZFS_CASE_INSENSITIVE &&
(flag & ZCIEXACT)) ||
(ZTOZSB(zp)->z_case == ZFS_CASE_MIXED &&
!(flag & ZCILOOK))) {
mt |= MT_MATCH_CASE;
}
error = zap_remove_norm(ZTOZSB(zp)->z_os, dzp->z_id,
dl->dl_name, mt, tx);
} else {
error = zap_remove(ZTOZSB(zp)->z_os, dzp->z_id, dl->dl_name,
tx);
}
return (error);
}
void
dsl_dir_destroy_sync(void *arg1, void *tag, cred_t *cr, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t val, obj;
ASSERT(RW_WRITE_HELD(&dd->dd_pool->dp_config_rwlock));
ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
/* Remove our reservation. */
val = 0;
dsl_dir_set_reservation_sync(dd, &val, cr, tx);
ASSERT3U(dd->dd_used_bytes, ==, 0);
ASSERT3U(dd->dd_phys->dd_reserved, ==, 0);
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_child_dir_zapobj, tx));
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_props_zapobj, tx));
VERIFY(0 == dsl_deleg_destroy(mos, dd->dd_phys->dd_deleg_zapobj, tx));
VERIFY(0 == zap_remove(mos,
dd->dd_parent->dd_phys->dd_child_dir_zapobj, dd->dd_myname, tx));
obj = dd->dd_object;
dsl_dir_close(dd, tag);
VERIFY(0 == dmu_object_free(mos, obj, tx));
}
void
dsl_dir_destroy_sync(void *arg1, void *tag, dmu_tx_t *tx)
{
dsl_dir_t *dd = arg1;
objset_t *mos = dd->dd_pool->dp_meta_objset;
uint64_t obj;
dd_used_t t;
ASSERT(RW_WRITE_HELD(&dd->dd_pool->dp_config_rwlock));
ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
/*
* Remove our reservation. The impl() routine avoids setting the
* actual property, which would require the (already destroyed) ds.
*/
dsl_dir_set_reservation_sync_impl(dd, 0, tx);
ASSERT0(dd->dd_phys->dd_used_bytes);
ASSERT0(dd->dd_phys->dd_reserved);
for (t = 0; t < DD_USED_NUM; t++)
ASSERT0(dd->dd_phys->dd_used_breakdown[t]);
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_child_dir_zapobj, tx));
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_props_zapobj, tx));
VERIFY(0 == dsl_deleg_destroy(mos, dd->dd_phys->dd_deleg_zapobj, tx));
VERIFY(0 == zap_remove(mos,
dd->dd_parent->dd_phys->dd_child_dir_zapobj, dd->dd_myname, tx));
obj = dd->dd_object;
dsl_dir_close(dd, tag);
VERIFY(0 == dmu_object_free(mos, obj, tx));
}
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);
}
static int osd_dir_delete(const struct lu_env *env, struct dt_object *dt,
const struct dt_key *key, struct thandle *th,
struct lustre_capa *capa)
{
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
struct osd_thandle *oh;
dmu_buf_t *zap_db = obj->oo_db;
int rc;
ENTRY;
LASSERT(obj->oo_db);
LASSERT(udmu_object_is_zap(obj->oo_db));
LASSERT(th != NULL);
oh = container_of0(th, struct osd_thandle, ot_super);
/* Remove key from the ZAP */
rc = -zap_remove(osd->od_objset.os, zap_db->db_object,
(char *) key, oh->ot_tx);
if (rc && rc != -ENOENT)
CERROR("%s: zap_remove failed: rc = %d\n", osd->od_svname, rc);
RETURN(rc);
}
int __osd_xattr_del(const struct lu_env *env, struct osd_object *obj,
const char *name, struct osd_thandle *oh)
{
struct osd_device *osd = osd_obj2dev(obj);
udmu_objset_t *uos = &osd->od_objset;
uint64_t xa_data_obj;
int rc;
/* try remove xattr from SA at first */
rc = __osd_sa_xattr_del(env, obj, name, oh);
if (rc != -ENOENT)
return rc;
if (obj->oo_xattr == ZFS_NO_OBJECT)
return 0;
rc = -zap_lookup(uos->os, obj->oo_xattr, name, sizeof(uint64_t), 1,
&xa_data_obj);
if (rc == -ENOENT) {
rc = 0;
} else if (rc == 0) {
/*
* Entry exists.
* We'll delete the existing object and ZAP entry.
*/
rc = __osd_object_free(uos, xa_data_obj, oh->ot_tx);
if (rc)
return rc;
rc = -zap_remove(uos->os, obj->oo_xattr, name, oh->ot_tx);
}
return rc;
}
static void
dsl_dataset_user_release_sync_one(dsl_dataset_t *ds, nvlist_t *holds,
dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
nvpair_t *pair;
for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
int error;
const char *holdname = nvpair_name(pair);
/* Remove temporary hold if one exists. */
error = dsl_pool_user_release(dp, ds->ds_object, holdname, tx);
VERIFY(error == 0 || error == ENOENT);
VERIFY0(zap_remove(mos, dsl_dataset_phys(ds)->ds_userrefs_obj,
holdname, tx));
ds->ds_userrefs--;
spa_history_log_internal_ds(ds, "release", tx,
"tag=%s refs=%lld", holdname, (longlong_t)ds->ds_userrefs);
}
}
/*
* This sync task completes (finishes) a condense, deleting the old
* mapping and replacing it with the new one.
*/
static void
spa_condense_indirect_complete_sync(void *arg, dmu_tx_t *tx)
{
spa_condensing_indirect_t *sci = arg;
spa_t *spa = dmu_tx_pool(tx)->dp_spa;
spa_condensing_indirect_phys_t *scip =
&spa->spa_condensing_indirect_phys;
vdev_t *vd = vdev_lookup_top(spa, scip->scip_vdev);
vdev_indirect_config_t *vic = &vd->vdev_indirect_config;
objset_t *mos = spa->spa_meta_objset;
vdev_indirect_mapping_t *old_mapping = vd->vdev_indirect_mapping;
uint64_t old_count = vdev_indirect_mapping_num_entries(old_mapping);
uint64_t new_count =
vdev_indirect_mapping_num_entries(sci->sci_new_mapping);
ASSERT(dmu_tx_is_syncing(tx));
ASSERT3P(vd->vdev_ops, ==, &vdev_indirect_ops);
ASSERT3P(sci, ==, spa->spa_condensing_indirect);
for (int i = 0; i < TXG_SIZE; i++) {
ASSERT(list_is_empty(&sci->sci_new_mapping_entries[i]));
}
ASSERT(vic->vic_mapping_object != 0);
ASSERT3U(vd->vdev_id, ==, scip->scip_vdev);
ASSERT(scip->scip_next_mapping_object != 0);
ASSERT(scip->scip_prev_obsolete_sm_object != 0);
/*
* Reset vdev_indirect_mapping to refer to the new object.
*/
rw_enter(&vd->vdev_indirect_rwlock, RW_WRITER);
vdev_indirect_mapping_close(vd->vdev_indirect_mapping);
vd->vdev_indirect_mapping = sci->sci_new_mapping;
rw_exit(&vd->vdev_indirect_rwlock);
sci->sci_new_mapping = NULL;
vdev_indirect_mapping_free(mos, vic->vic_mapping_object, tx);
vic->vic_mapping_object = scip->scip_next_mapping_object;
scip->scip_next_mapping_object = 0;
space_map_free_obj(mos, scip->scip_prev_obsolete_sm_object, tx);
spa_feature_decr(spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
scip->scip_prev_obsolete_sm_object = 0;
scip->scip_vdev = 0;
VERIFY0(zap_remove(mos, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_CONDENSING_INDIRECT, tx));
spa_condensing_indirect_destroy(spa->spa_condensing_indirect);
spa->spa_condensing_indirect = NULL;
zfs_dbgmsg("finished condense of vdev %llu in txg %llu: "
"new mapping object %llu has %llu entries "
"(was %llu entries)",
vd->vdev_id, dmu_tx_get_txg(tx), vic->vic_mapping_object,
new_count, old_count);
vdev_config_dirty(spa->spa_root_vdev);
}
/*
* Unlink zp from dl, and mark zp for reaping if this was the last link.
* Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
* If 'reaped_ptr' is NULL, we put reaped znodes on the delete queue.
* If it's non-NULL, we use it to indicate whether the znode needs reaping,
* and it's the caller's job to do it.
*/
int
zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
int *reaped_ptr)
{
znode_t *dzp = dl->dl_dzp;
vnode_t *vp = ZTOV(zp);
int zp_is_dir = (vp->v_type == VDIR);
int reaped = 0;
int error;
dnlc_remove(ZTOV(dzp), dl->dl_name);
if (!(flag & ZRENAMING)) {
dmu_buf_will_dirty(zp->z_dbuf, tx);
if (vn_vfswlock(vp)) /* prevent new mounts on zp */
return (EBUSY);
if (vn_ismntpt(vp)) { /* don't remove mount point */
vn_vfsunlock(vp);
return (EBUSY);
}
mutex_enter(&zp->z_lock);
if (zp_is_dir && !zfs_dirempty(zp)) { /* dir not empty */
mutex_exit(&zp->z_lock);
vn_vfsunlock(vp);
return (EEXIST);
}
ASSERT(zp->z_phys->zp_links > zp_is_dir);
if (--zp->z_phys->zp_links == zp_is_dir) {
zp->z_reap = 1;
zp->z_phys->zp_links = 0;
reaped = 1;
} else {
zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
}
mutex_exit(&zp->z_lock);
vn_vfsunlock(vp);
}
dmu_buf_will_dirty(dzp->z_dbuf, tx);
mutex_enter(&dzp->z_lock);
dzp->z_phys->zp_size--; /* one dirent removed */
dzp->z_phys->zp_links -= zp_is_dir; /* ".." link from zp */
zfs_time_stamper_locked(dzp, CONTENT_MODIFIED, tx);
mutex_exit(&dzp->z_lock);
error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name, tx);
ASSERT(error == 0);
if (reaped_ptr != NULL)
*reaped_ptr = reaped;
else if (reaped)
zfs_dq_add(zp, tx);
return (0);
}
void
dsl_pool_destroy_obsolete_bpobj(dsl_pool_t *dp, dmu_tx_t *tx)
{
spa_feature_decr(dp->dp_spa, SPA_FEATURE_OBSOLETE_COUNTS, tx);
VERIFY0(zap_remove(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_OBSOLETE_BPOBJ, tx));
bpobj_free(dp->dp_meta_objset,
dp->dp_obsolete_bpobj.bpo_object, tx);
bpobj_close(&dp->dp_obsolete_bpobj);
}
/*
* 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);
}
static int osd_dir_delete(const struct lu_env *env, struct dt_object *dt,
const struct dt_key *key, struct thandle *th,
struct lustre_capa *capa)
{
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
struct osd_thandle *oh;
dmu_buf_t *zap_db = obj->oo_db;
char *name = (char *)key;
int rc;
ENTRY;
LASSERT(obj->oo_db);
LASSERT(udmu_object_is_zap(obj->oo_db));
LASSERT(th != NULL);
oh = container_of0(th, struct osd_thandle, ot_super);
/*
* In Orion . and .. were stored in the directory (not generated upon
* request as now). we preserve them for backward compatibility
*/
if (name[0] == '.') {
if (name[1] == 0) {
RETURN(0);
} else if (name[1] == '.' && name[2] == 0) {
RETURN(0);
}
}
/* Remove key from the ZAP */
rc = -zap_remove(osd->od_objset.os, zap_db->db_object,
(char *) key, oh->ot_tx);
#if LUSTRE_VERSION_CODE <= OBD_OCD_VERSION(2, 4, 53, 0)
if (unlikely(rc == -ENOENT && name[0] == '.' &&
(name[1] == 0 || (name[1] == '.' && name[2] == 0))))
rc = 0;
#endif
if (unlikely(rc && rc != -ENOENT))
CERROR("%s: zap_remove failed: rc = %d\n", osd->od_svname, rc);
RETURN(rc);
}
static void
dsl_bookmark_destroy_sync(void *arg, dmu_tx_t *tx)
{
dsl_bookmark_destroy_arg_t *dbda = arg;
dsl_pool_t *dp = dmu_tx_pool(tx);
objset_t *mos = dp->dp_meta_objset;
nvpair_t *pair;
for (pair = nvlist_next_nvpair(dbda->dbda_success, NULL);
pair != NULL; pair = nvlist_next_nvpair(dbda->dbda_success, pair)) {
dsl_dataset_t *ds;
char *shortname;
uint64_t zap_cnt;
VERIFY0(dsl_bookmark_hold_ds(dp, nvpair_name(pair),
&ds, FTAG, &shortname));
VERIFY0(dsl_dataset_bookmark_remove(ds, shortname, tx));
/*
* If all of this dataset's bookmarks have been destroyed,
* free the zap object and decrement the feature's use count.
*/
VERIFY0(zap_count(mos, ds->ds_bookmarks,
&zap_cnt));
if (zap_cnt == 0) {
dmu_buf_will_dirty(ds->ds_dbuf, tx);
VERIFY0(zap_destroy(mos, ds->ds_bookmarks, tx));
ds->ds_bookmarks = 0;
spa_feature_decr(dp->dp_spa, SPA_FEATURE_BOOKMARKS, tx);
VERIFY0(zap_remove(mos, ds->ds_object,
DS_FIELD_BOOKMARK_NAMES, tx));
}
spa_history_log_internal_ds(ds, "remove bookmark", tx,
"name=%s", shortname);
dsl_dataset_rele(ds, FTAG);
}
}
static int
zfs_dropname(zfs_dirlock_t *dl, znode_t *zp, znode_t *dzp, dmu_tx_t *tx,
int flag)
{
int error;
if (zp->z_zfsvfs->z_norm) {
if (((zp->z_zfsvfs->z_case == ZFS_CASE_INSENSITIVE) &&
(flag & ZCIEXACT)) ||
((zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) &&
!(flag & ZCILOOK)))
error = zap_remove_norm(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, MT_EXACT, tx);
else
error = zap_remove_norm(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, MT_FIRST, tx);
} else {
error = zap_remove(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, tx);
}
return (error);
}
void
dsl_dir_destroy_sync(void *arg1, void *tag, dmu_tx_t *tx)
{
dsl_dataset_t *ds = arg1;
dsl_dir_t *dd = ds->ds_dir;
objset_t *mos = dd->dd_pool->dp_meta_objset;
dsl_prop_setarg_t psa;
uint64_t value = 0;
uint64_t obj;
dd_used_t t;
ASSERT(RW_WRITE_HELD(&dd->dd_pool->dp_config_rwlock));
ASSERT(dd->dd_phys->dd_head_dataset_obj == 0);
/* Remove our reservation. */
dsl_prop_setarg_init_uint64(&psa, "reservation",
(ZPROP_SRC_NONE | ZPROP_SRC_LOCAL | ZPROP_SRC_RECEIVED),
&value);
psa.psa_effective_value = 0; /* predict default value */
dsl_dir_set_reservation_sync(ds, &psa, tx);
ASSERT3U(dd->dd_phys->dd_used_bytes, ==, 0);
ASSERT3U(dd->dd_phys->dd_reserved, ==, 0);
for (t = 0; t < DD_USED_NUM; t++)
ASSERT3U(dd->dd_phys->dd_used_breakdown[t], ==, 0);
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_child_dir_zapobj, tx));
VERIFY(0 == zap_destroy(mos, dd->dd_phys->dd_props_zapobj, tx));
VERIFY(0 == dsl_deleg_destroy(mos, dd->dd_phys->dd_deleg_zapobj, tx));
VERIFY(0 == zap_remove(mos,
dd->dd_parent->dd_phys->dd_child_dir_zapobj, dd->dd_myname, tx));
obj = dd->dd_object;
dsl_dir_close(dd, tag);
VERIFY(0 == dmu_object_free(mos, obj, tx));
}
static void
dsl_dataset_user_release_sync_one(dsl_dataset_t *ds, nvlist_t *holds,
dmu_tx_t *tx)
{
dsl_pool_t *dp = ds->ds_dir->dd_pool;
objset_t *mos = dp->dp_meta_objset;
uint64_t zapobj;
int error;
nvpair_t *pair;
for (pair = nvlist_next_nvpair(holds, NULL); pair != NULL;
pair = nvlist_next_nvpair(holds, pair)) {
ds->ds_userrefs--;
error = dsl_pool_user_release(dp, ds->ds_object,
nvpair_name(pair), tx);
VERIFY(error == 0 || error == ENOENT);
zapobj = ds->ds_phys->ds_userrefs_obj;
VERIFY0(zap_remove(mos, zapobj, nvpair_name(pair), tx));
spa_history_log_internal_ds(ds, "release", tx,
"tag=%s refs=%lld", nvpair_name(pair),
(longlong_t)ds->ds_userrefs);
}
}
int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
const struct lu_buf *buf, const char *name, int fl,
struct osd_thandle *oh)
{
dmu_buf_impl_t *db;
uchar_t *nv_value;
size_t size;
int nv_size;
int rc;
int too_big = 0;
LASSERT(obj->oo_sa_hdl);
if (obj->oo_sa_xattr == NULL) {
rc = __osd_xattr_cache(env, obj);
if (rc)
return rc;
}
LASSERT(obj->oo_sa_xattr);
/* Limited to 32k to keep nvpair memory allocations small */
if (buf->lb_len > DXATTR_MAX_ENTRY_SIZE) {
too_big = 1;
} else {
/* Prevent the DXATTR SA from consuming the entire SA
* region */
rc = -nvlist_size(obj->oo_sa_xattr, &size, NV_ENCODE_XDR);
if (rc)
return rc;
if (size + buf->lb_len > DXATTR_MAX_SA_SIZE)
too_big = 1;
}
/* even in case of -EFBIG we must lookup xattr and check can we
* rewrite it then delete from SA */
rc = -nvlist_lookup_byte_array(obj->oo_sa_xattr, name, &nv_value,
&nv_size);
if (rc == 0) {
if (fl & LU_XATTR_CREATE) {
return -EEXIST;
} else if (too_big) {
rc = -nvlist_remove(obj->oo_sa_xattr, name,
DATA_TYPE_BYTE_ARRAY);
if (rc < 0)
return rc;
rc = __osd_sa_xattr_schedule_update(env, obj, oh);
return rc == 0 ? -EFBIG : rc;
}
} else if (rc == -ENOENT) {
if (fl & LU_XATTR_REPLACE)
return -ENODATA;
else if (too_big)
return -EFBIG;
} else {
return rc;
}
/* Ensure xattr doesn't exist in ZAP */
if (obj->oo_xattr != ZFS_NO_OBJECT) {
struct osd_device *osd = osd_obj2dev(obj);
uint64_t objid;
rc = -zap_lookup(osd->od_os, obj->oo_xattr,
name, 8, 1, &objid);
if (rc == 0) {
rc = -dmu_object_free(osd->od_os, objid, oh->ot_tx);
if (rc == 0)
zap_remove(osd->od_os, obj->oo_xattr,
name, oh->ot_tx);
}
}
rc = -nvlist_add_byte_array(obj->oo_sa_xattr, name,
(uchar_t *)buf->lb_buf, buf->lb_len);
if (rc)
return rc;
/* batch updates only for just created dnodes where we
* used to set number of EAs in a single transaction */
db = (dmu_buf_impl_t *)obj->oo_db;
if (DB_DNODE(db)->dn_allocated_txg == oh->ot_tx->tx_txg)
rc = __osd_sa_xattr_schedule_update(env, obj, oh);
else
rc = __osd_sa_xattr_update(env, obj, oh);
return rc;
}
/* ARGSUSED */
static void
dsl_scan_done(dsl_scan_t *scn, boolean_t complete, dmu_tx_t *tx)
{
static const char *old_names[] = {
"scrub_bookmark",
"scrub_ddt_bookmark",
"scrub_ddt_class_max",
"scrub_queue",
"scrub_min_txg",
"scrub_max_txg",
"scrub_func",
"scrub_errors",
NULL
};
dsl_pool_t *dp = scn->scn_dp;
spa_t *spa = dp->dp_spa;
int i;
/* Remove any remnants of an old-style scrub. */
for (i = 0; old_names[i]; i++) {
(void) zap_remove(dp->dp_meta_objset,
DMU_POOL_DIRECTORY_OBJECT, old_names[i], tx);
}
if (scn->scn_phys.scn_queue_obj != 0) {
VERIFY(0 == dmu_object_free(dp->dp_meta_objset,
scn->scn_phys.scn_queue_obj, tx));
scn->scn_phys.scn_queue_obj = 0;
}
/*
* If we were "restarted" from a stopped state, don't bother
* with anything else.
*/
if (scn->scn_phys.scn_state != DSS_SCANNING)
return;
if (complete)
scn->scn_phys.scn_state = DSS_FINISHED;
else
scn->scn_phys.scn_state = DSS_CANCELED;
spa_history_log_internal(spa, "scan done", tx,
"complete=%u", complete);
if (DSL_SCAN_IS_SCRUB_RESILVER(scn)) {
mutex_enter(&spa->spa_scrub_lock);
while (spa->spa_scrub_inflight > 0) {
cv_wait(&spa->spa_scrub_io_cv,
&spa->spa_scrub_lock);
}
mutex_exit(&spa->spa_scrub_lock);
spa->spa_scrub_started = B_FALSE;
spa->spa_scrub_active = B_FALSE;
/*
* If the scrub/resilver completed, update all DTLs to
* reflect this. Whether it succeeded or not, vacate
* all temporary scrub DTLs.
*/
vdev_dtl_reassess(spa->spa_root_vdev, tx->tx_txg,
complete ? scn->scn_phys.scn_max_txg : 0, B_TRUE);
if (complete) {
spa_event_notify(spa, NULL, scn->scn_phys.scn_min_txg ?
ESC_ZFS_RESILVER_FINISH : ESC_ZFS_SCRUB_FINISH);
}
spa_errlog_rotate(spa);
/*
* We may have finished replacing a device.
* Let the async thread assess this and handle the detach.
*/
spa_async_request(spa, SPA_ASYNC_RESILVER_DONE);
}
scn->scn_phys.scn_end_time = gethrestime_sec();
}
int __osd_sa_xattr_set(const struct lu_env *env, struct osd_object *obj,
const struct lu_buf *buf, const char *name, int fl,
struct osd_thandle *oh)
{
uchar_t *nv_value;
size_t size;
int nv_size;
int rc;
int too_big = 0;
LASSERT(obj->oo_sa_hdl);
if (obj->oo_sa_xattr == NULL) {
rc = __osd_xattr_cache(env, obj);
if (rc)
return rc;
}
LASSERT(obj->oo_sa_xattr);
/* Limited to 32k to keep nvpair memory allocations small */
if (buf->lb_len > DXATTR_MAX_ENTRY_SIZE) {
too_big = 1;
} else {
/* Prevent the DXATTR SA from consuming the entire SA
* region */
rc = -nvlist_size(obj->oo_sa_xattr, &size, NV_ENCODE_XDR);
if (rc)
return rc;
if (size + buf->lb_len > DXATTR_MAX_SA_SIZE)
too_big = 1;
}
/* even in case of -EFBIG we must lookup xattr and check can we
* rewrite it then delete from SA */
rc = -nvlist_lookup_byte_array(obj->oo_sa_xattr, name, &nv_value,
&nv_size);
if (rc == 0) {
if (fl & LU_XATTR_CREATE) {
return -EEXIST;
} else if (too_big) {
rc = -nvlist_remove(obj->oo_sa_xattr, name,
DATA_TYPE_BYTE_ARRAY);
if (rc < 0)
return rc;
rc = __osd_sa_xattr_update(env, obj, oh);
return rc == 0 ? -EFBIG : rc;
}
} else if (rc == -ENOENT) {
if (fl & LU_XATTR_REPLACE)
return -ENODATA;
else if (too_big)
return -EFBIG;
} else {
return rc;
}
/* Ensure xattr doesn't exist in ZAP */
if (obj->oo_xattr != ZFS_NO_OBJECT) {
udmu_objset_t *uos = &osd_obj2dev(obj)->od_objset;
uint64_t xa_data_obj;
rc = -zap_lookup(uos->os, obj->oo_xattr,
name, 8, 1, &xa_data_obj);
if (rc == 0) {
rc = __osd_object_free(uos, xa_data_obj, oh->ot_tx);
if (rc == 0)
zap_remove(uos->os, obj->oo_xattr,
name, oh->ot_tx);
}
}
rc = -nvlist_add_byte_array(obj->oo_sa_xattr, name,
(uchar_t *)buf->lb_buf, buf->lb_len);
if (rc)
return rc;
rc = __osd_sa_xattr_update(env, obj, oh);
return rc;
}
void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
char obj_name[17];
dmu_tx_t *tx;
uint64_t acl_obj;
int error;
ASSERT(ZTOV(zp)->v_count == 0);
ASSERT(zp->z_phys->zp_links == 0);
/*
* If this is an attribute directory, purge its contents.
*/
if (ZTOV(zp)->v_type == VDIR && (zp->z_phys->zp_flags & ZFS_XATTR)) {
if (zfs_purgedir(zp) != 0) {
/*
* Not enough space to delete some xattrs.
* Leave it on the unlinked set.
*/
return;
}
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
if (zp->z_phys->zp_xattr) {
error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
ASSERT(error == 0);
}
acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
/*
* Set up the transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
if (xzp) {
dmu_tx_hold_bonus(tx, xzp->z_id);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
/*
* Not enough space to delete the file. Leave it in the
* unlinked set, leaking it until the fs is remounted (at
* which point we'll call zfs_unlinked_drain() to process it).
*/
dmu_tx_abort(tx);
return;
}
if (xzp) {
dmu_buf_will_dirty(xzp->z_dbuf, tx);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
xzp->z_phys->zp_links = 0; /* no more links to it */
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
}
/* Remove this znode from the unlinked set */
error = zap_remove(os, zfsvfs->z_unlinkedobj,
zfs_unlinked_hexname(obj_name, zp->z_id), tx);
ASSERT3U(error, ==, 0);
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
if (xzp)
VN_RELE(ZTOV(xzp));
}
static int osd_object_destroy(const struct lu_env *env,
struct dt_object *dt, struct thandle *th)
{
struct osd_thread_info *info = osd_oti_get(env);
char *buf = info->oti_str;
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
const struct lu_fid *fid = lu_object_fid(&dt->do_lu);
struct osd_thandle *oh;
int rc;
uint64_t oid, zapid;
ENTRY;
down_write(&obj->oo_guard);
if (unlikely(!dt_object_exists(dt) || obj->oo_destroyed))
GOTO(out, rc = -ENOENT);
LASSERT(obj->oo_db != NULL);
oh = container_of0(th, struct osd_thandle, ot_super);
LASSERT(oh != NULL);
LASSERT(oh->ot_tx != NULL);
/* remove obj ref from index dir (it depends) */
zapid = osd_get_name_n_idx(env, osd, fid, buf, sizeof(info->oti_str));
rc = -zap_remove(osd->od_os, zapid, buf, oh->ot_tx);
if (rc) {
CERROR("%s: zap_remove(%s) failed: rc = %d\n",
osd->od_svname, buf, rc);
GOTO(out, rc);
}
rc = osd_xattrs_destroy(env, obj, oh);
if (rc) {
CERROR("%s: cannot destroy xattrs for %s: rc = %d\n",
osd->od_svname, buf, rc);
GOTO(out, rc);
}
/* Remove object from inode accounting. It is not fatal for the destroy
* operation if something goes wrong while updating accounting, but we
* still log an error message to notify the administrator */
rc = -zap_increment_int(osd->od_os, osd->od_iusr_oid,
obj->oo_attr.la_uid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to remove "DFID" from accounting ZAP for usr"
" %d: rc = %d\n", osd->od_svname, PFID(fid),
obj->oo_attr.la_uid, rc);
rc = -zap_increment_int(osd->od_os, osd->od_igrp_oid,
obj->oo_attr.la_gid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to remove "DFID" from accounting ZAP for grp"
" %d: rc = %d\n", osd->od_svname, PFID(fid),
obj->oo_attr.la_gid, rc);
oid = obj->oo_db->db_object;
if (unlikely(obj->oo_destroy == OSD_DESTROY_NONE)) {
/* this may happen if the destroy wasn't declared
* e.g. when the object is created and then destroyed
* in the same transaction - we don't need additional
* space for destroy specifically */
LASSERT(obj->oo_attr.la_size <= osd_sync_destroy_max_size);
rc = -dmu_object_free(osd->od_os, oid, oh->ot_tx);
if (rc)
CERROR("%s: failed to free %s %llu: rc = %d\n",
osd->od_svname, buf, oid, rc);
} else if (obj->oo_destroy == OSD_DESTROY_SYNC) {
rc = -dmu_object_free(osd->od_os, oid, oh->ot_tx);
if (rc)
CERROR("%s: failed to free %s %llu: rc = %d\n",
osd->od_svname, buf, oid, rc);
} else { /* asynchronous destroy */
rc = osd_object_unlinked_add(obj, oh);
if (rc)
GOTO(out, rc);
rc = -zap_add_int(osd->od_os, osd->od_unlinkedid,
oid, oh->ot_tx);
if (rc)
CERROR("%s: zap_add_int() failed %s %llu: rc = %d\n",
osd->od_svname, buf, oid, rc);
}
out:
/* not needed in the cache anymore */
set_bit(LU_OBJECT_HEARD_BANSHEE, &dt->do_lu.lo_header->loh_flags);
if (rc == 0)
obj->oo_destroyed = 1;
up_write(&obj->oo_guard);
RETURN (0);
}
void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
char obj_name[17];
dmu_tx_t *tx;
uint64_t acl_obj;
int error;
#ifndef __APPLE__
ASSERT(ZTOV(zp)->v_count == 0);
#endif /*!__APPLE__*/
ASSERT(zp->z_phys->zp_links == 0);
#ifdef ZFS_DEBUG
znode_stalker(zp, N_zfs_rmnode);
#endif
/*
* If this is an attribute directory, purge its contents.
*/
#ifdef __APPLE__
if (S_ISDIR(zp->z_phys->zp_mode) && (zp->z_phys->zp_flags & ZFS_XATTR))
#else
if (ZTOV(zp)->v_type == VDIR && (zp->z_phys->zp_flags & ZFS_XATTR))
#endif
{
if (zfs_purgedir(zp) != 0) {
/*
* Not enough space to delete some xattrs.
* Leave it on the unlinked set.
*/
return;
}
}
/*
* If the file has extended attributes, we're going to unlink
* the xattr dir.
*/
if (zp->z_phys->zp_xattr) {
error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
ASSERT(error == 0);
}
acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
/*
* Set up the transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL);
if (xzp) {
dmu_tx_hold_bonus(tx, xzp->z_id);
dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, TRUE, NULL);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
/*
* Not enough space to delete the file. Leave it in the
* unlinked set, leaking it until the fs is remounted (at
* which point we'll call zfs_unlinked_drain() to process it).
*/
dmu_tx_abort(tx);
#ifdef __APPLE__
/*XXX NOEL: double check this path logic. see radar 5182217.
* This may be disturbing some of the evict logic
* and hence causing the NULL ptr drefs seen every great while
* in some of the test cases*/
zp->z_dbuf_held = 0;
ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id);
dmu_buf_rele(zp->z_dbuf, NULL);
#endif /* __APPLE__ */
return;
}
if (xzp) {
dmu_buf_will_dirty(xzp->z_dbuf, tx);
mutex_enter(&xzp->z_lock);
xzp->z_unlinked = B_TRUE; /* mark xzp for deletion */
xzp->z_phys->zp_links = 0; /* no more links to it */
mutex_exit(&xzp->z_lock);
zfs_unlinked_add(xzp, tx);
}
/* Remove this znode from the unlinked set */
error = zap_remove(os, zfsvfs->z_unlinkedobj,
zfs_unlinked_hexname(obj_name, zp->z_id), tx);
ASSERT3U(error, ==, 0);
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
if (xzp)
VN_RELE(ZTOV(xzp));
}
/*
* Unlink zp from dl, and mark zp for deletion if this was the last link.
* Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
* If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
* If it's non-NULL, we use it to indicate whether the znode needs deletion,
* and it's the caller's job to do it.
*/
int
zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
boolean_t *unlinkedp)
{
znode_t *dzp = dl->dl_dzp;
vnode_t *vp = ZTOV(zp);
#ifdef __APPLE__
int zp_is_dir = S_ISDIR(zp->z_phys->zp_mode);
#else
int zp_is_dir = (vp->v_type == VDIR);
#endif
boolean_t unlinked = B_FALSE;
int error;
#ifndef __APPLE__
dnlc_remove(ZTOV(dzp), dl->dl_name);
#endif
if (!(flag & ZRENAMING)) {
dmu_buf_will_dirty(zp->z_dbuf, tx);
#ifdef __APPLE__
if (vp) {
#endif /* __APPLE__ */
if (vn_vfswlock(vp)) /* prevent new mounts on zp */
return (EBUSY);
if (vn_ismntpt(vp)) { /* don't remove mount point */
vn_vfsunlock(vp);
return (EBUSY);
}
#ifdef __APPLE__
} /* if (vp) */
#endif /* __APPLE__ */
mutex_enter(&zp->z_lock);
if (zp_is_dir && !zfs_dirempty(zp)) { /* dir not empty */
mutex_exit(&zp->z_lock);
#ifdef __APPLE__
return (ENOTEMPTY);
#else
vn_vfsunlock(vp);
return (EEXIST);
#endif
}
if (zp->z_phys->zp_links <= zp_is_dir) {
#ifndef __APPLE__
zfs_panic_recover("zfs: link count on %s is %u, "
"should be at least %u",
zp->z_vnode->v_path ? zp->z_vnode->v_path :
"<unknown>", (int)zp->z_phys->zp_links,
zp_is_dir + 1);
#endif
zp->z_phys->zp_links = zp_is_dir + 1;
}
if (--zp->z_phys->zp_links == zp_is_dir) {
zp->z_unlinked = B_TRUE;
zp->z_phys->zp_links = 0;
unlinked = B_TRUE;
} else {
zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
}
mutex_exit(&zp->z_lock);
#ifndef __APPLE__
vn_vfsunlock(vp);
#endif
}
dmu_buf_will_dirty(dzp->z_dbuf, tx);
mutex_enter(&dzp->z_lock);
dzp->z_phys->zp_size--; /* one dirent removed */
dzp->z_phys->zp_links -= zp_is_dir; /* ".." link from zp */
zfs_time_stamper_locked(dzp, CONTENT_MODIFIED, tx);
mutex_exit(&dzp->z_lock);
error = zap_remove(zp->z_zfsvfs->z_os, dzp->z_id, dl->dl_name, tx);
ASSERT(error == 0);
if (unlinkedp != NULL)
*unlinkedp = unlinked;
else if (unlinked)
zfs_unlinked_add(zp, tx);
return (0);
}
/* ARGSUSED */
static void
dsl_pool_scrub_cancel_sync(void *arg1, void *arg2, cred_t *cr, dmu_tx_t *tx)
{
dsl_pool_t *dp = arg1;
boolean_t *completep = arg2;
if (dp->dp_scrub_func == SCRUB_FUNC_NONE)
return;
mutex_enter(&dp->dp_scrub_cancel_lock);
if (dp->dp_scrub_restart) {
dp->dp_scrub_restart = B_FALSE;
*completep = B_FALSE;
}
/* XXX this is scrub-clean specific */
mutex_enter(&dp->dp_spa->spa_scrub_lock);
while (dp->dp_spa->spa_scrub_inflight > 0) {
cv_wait(&dp->dp_spa->spa_scrub_io_cv,
&dp->dp_spa->spa_scrub_lock);
}
mutex_exit(&dp->dp_spa->spa_scrub_lock);
dp->dp_spa->spa_scrub_active = B_FALSE;
dp->dp_scrub_func = SCRUB_FUNC_NONE;
VERIFY(0 == dmu_object_free(dp->dp_meta_objset,
dp->dp_scrub_queue_obj, tx));
dp->dp_scrub_queue_obj = 0;
bzero(&dp->dp_scrub_bookmark, sizeof (zbookmark_t));
bzero(&dp->dp_scrub_ddt_bookmark, sizeof (ddt_bookmark_t));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_QUEUE, tx));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_MIN_TXG, tx));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_MAX_TXG, tx));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_BOOKMARK, tx));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_FUNC, tx));
VERIFY(0 == zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_ERRORS, tx));
(void) zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_DDT_BOOKMARK, tx);
(void) zap_remove(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
DMU_POOL_SCRUB_DDT_CLASS_MAX, tx);
spa_history_internal_log(LOG_POOL_SCRUB_DONE, dp->dp_spa, tx, cr,
"complete=%u", *completep);
/* below is scrub-clean specific */
vdev_scrub_stat_update(dp->dp_spa->spa_root_vdev, POOL_SCRUB_NONE,
*completep);
/*
* If the scrub/resilver completed, update all DTLs to reflect this.
* Whether it succeeded or not, vacate all temporary scrub DTLs.
*/
vdev_dtl_reassess(dp->dp_spa->spa_root_vdev, tx->tx_txg,
*completep ? dp->dp_scrub_max_txg : 0, B_TRUE);
dp->dp_spa->spa_scrub_started = B_FALSE;
if (*completep)
spa_event_notify(dp->dp_spa, NULL, dp->dp_scrub_min_txg ?
ESC_ZFS_RESILVER_FINISH : ESC_ZFS_SCRUB_FINISH);
spa_errlog_rotate(dp->dp_spa);
/*
* We may have finished replacing a device.
* Let the async thread assess this and handle the detach.
*/
spa_async_request(dp->dp_spa, SPA_ASYNC_RESILVER_DONE);
dp->dp_scrub_min_txg = dp->dp_scrub_max_txg = 0;
mutex_exit(&dp->dp_scrub_cancel_lock);
}
static int osd_object_destroy(const struct lu_env *env,
struct dt_object *dt,
struct thandle *th)
{
char *buf = osd_oti_get(env)->oti_str;
struct osd_object *obj = osd_dt_obj(dt);
struct osd_device *osd = osd_obj2dev(obj);
const struct lu_fid *fid = lu_object_fid(&dt->do_lu);
struct osd_thandle *oh;
int rc;
uint64_t zapid;
ENTRY;
LASSERT(obj->oo_db != NULL);
LASSERT(dt_object_exists(dt));
LASSERT(!lu_object_is_dying(dt->do_lu.lo_header));
oh = container_of0(th, struct osd_thandle, ot_super);
LASSERT(oh != NULL);
LASSERT(oh->ot_tx != NULL);
zapid = osd_get_name_n_idx(env, osd, fid, buf);
/* remove obj ref from index dir (it depends) */
rc = -zap_remove(osd->od_objset.os, zapid, buf, oh->ot_tx);
if (rc) {
CERROR("%s: zap_remove() failed: rc = %d\n",
osd->od_svname, rc);
GOTO(out, rc);
}
/* Remove object from inode accounting. It is not fatal for the destroy
* operation if something goes wrong while updating accounting, but we
* still log an error message to notify the administrator */
rc = -zap_increment_int(osd->od_objset.os, osd->od_iusr_oid,
obj->oo_attr.la_uid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to remove "DFID" from accounting ZAP for usr"
" %d: rc = %d\n", osd->od_svname, PFID(fid),
obj->oo_attr.la_uid, rc);
rc = -zap_increment_int(osd->od_objset.os, osd->od_igrp_oid,
obj->oo_attr.la_gid, -1, oh->ot_tx);
if (rc)
CERROR("%s: failed to remove "DFID" from accounting ZAP for grp"
" %d: rc = %d\n", osd->od_svname, PFID(fid),
obj->oo_attr.la_gid, rc);
/* kill object */
rc = __osd_object_destroy(env, obj, oh->ot_tx, osd_obj_tag);
if (rc) {
CERROR("%s: __osd_object_destroy() failed: rc = %d\n",
osd->od_svname, rc);
GOTO(out, rc);
}
out:
/* not needed in the cache anymore */
set_bit(LU_OBJECT_HEARD_BANSHEE, &dt->do_lu.lo_header->loh_flags);
RETURN (0);
}
/*
* Unlink zp from dl, and mark zp for deletion if this was the last link.
* Can fail if zp is a mount point (EBUSY) or a non-empty directory (EEXIST).
* If 'unlinkedp' is NULL, we put unlinked znodes on the unlinked list.
* If it's non-NULL, we use it to indicate whether the znode needs deletion,
* and it's the caller's job to do it.
*/
int
zfs_link_destroy(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag,
boolean_t *unlinkedp)
{
znode_t *dzp = dl->dl_dzp;
vnode_t *vp = ZTOV(zp);
int zp_is_dir = (vp->v_type == VDIR);
boolean_t unlinked = B_FALSE;
int error;
dnlc_remove(ZTOV(dzp), dl->dl_name);
if (!(flag & ZRENAMING)) {
dmu_buf_will_dirty(zp->z_dbuf, tx);
if (vn_vfswlock(vp)) /* prevent new mounts on zp */
return (EBUSY);
if (vn_ismntpt(vp)) { /* don't remove mount point */
vn_vfsunlock(vp);
return (EBUSY);
}
mutex_enter(&zp->z_lock);
if (zp_is_dir && !zfs_dirempty(zp)) { /* dir not empty */
mutex_exit(&zp->z_lock);
vn_vfsunlock(vp);
return (ENOTEMPTY);
}
if (zp->z_phys->zp_links <= zp_is_dir) {
zfs_panic_recover("zfs: link count on vnode %p is %u, "
"should be at least %u", zp->z_vnode,
(int)zp->z_phys->zp_links,
zp_is_dir + 1);
zp->z_phys->zp_links = zp_is_dir + 1;
}
if (--zp->z_phys->zp_links == zp_is_dir) {
zp->z_unlinked = B_TRUE;
zp->z_phys->zp_links = 0;
unlinked = B_TRUE;
} else {
zfs_time_stamper_locked(zp, STATE_CHANGED, tx);
}
mutex_exit(&zp->z_lock);
vn_vfsunlock(vp);
}
dmu_buf_will_dirty(dzp->z_dbuf, tx);
mutex_enter(&dzp->z_lock);
dzp->z_phys->zp_size--; /* one dirent removed */
dzp->z_phys->zp_links -= zp_is_dir; /* ".." link from zp */
zfs_time_stamper_locked(dzp, CONTENT_MODIFIED, tx);
mutex_exit(&dzp->z_lock);
if (zp->z_zfsvfs->z_norm) {
if (((zp->z_zfsvfs->z_case == ZFS_CASE_INSENSITIVE) &&
(flag & ZCIEXACT)) ||
((zp->z_zfsvfs->z_case == ZFS_CASE_MIXED) &&
!(flag & ZCILOOK)))
error = zap_remove_norm(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, MT_EXACT, tx);
else
error = zap_remove_norm(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, MT_FIRST, tx);
} else {
error = zap_remove(zp->z_zfsvfs->z_os,
dzp->z_id, dl->dl_name, tx);
}
ASSERT(error == 0);
if (unlinkedp != NULL)
*unlinkedp = unlinked;
else if (unlinked)
zfs_unlinked_add(zp, tx);
return (0);
}
void
zfs_rmnode(znode_t *zp)
{
zfsvfs_t *zfsvfs = zp->z_zfsvfs;
objset_t *os = zfsvfs->z_os;
znode_t *xzp = NULL;
char obj_name[17];
dmu_tx_t *tx;
uint64_t acl_obj;
int error;
ASSERT(ZTOV(zp)->v_count == 0);
ASSERT(zp->z_phys->zp_links == 0);
/*
* If this is an attribute directory, purge its contents.
*/
if (ZTOV(zp)->v_type == VDIR && (zp->z_phys->zp_flags & ZFS_XATTR))
if (zfs_purgedir(zp) != 0) {
zfs_delete_t *delq = &zfsvfs->z_delete_head;
/*
* Add this back to the delete list to be retried later.
*
* XXX - this could just busy loop on us...
*/
mutex_enter(&delq->z_mutex);
list_insert_tail(&delq->z_znodes, zp);
delq->z_znode_count++;
mutex_exit(&delq->z_mutex);
return;
}
/*
* If the file has extended attributes, unlink the xattr dir.
*/
if (zp->z_phys->zp_xattr) {
error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp);
ASSERT(error == 0);
}
acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj;
/*
* Set up the transaction.
*/
tx = dmu_tx_create(os);
dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END);
dmu_tx_hold_zap(tx, zfsvfs->z_dqueue, FALSE, NULL);
if (xzp) {
dmu_tx_hold_bonus(tx, xzp->z_id);
dmu_tx_hold_zap(tx, zfsvfs->z_dqueue, TRUE, NULL);
}
if (acl_obj)
dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END);
error = dmu_tx_assign(tx, TXG_WAIT);
if (error) {
zfs_delete_t *delq = &zfsvfs->z_delete_head;
dmu_tx_abort(tx);
/*
* Add this back to the delete list to be retried later.
*
* XXX - this could just busy loop on us...
*/
mutex_enter(&delq->z_mutex);
list_insert_tail(&delq->z_znodes, zp);
delq->z_znode_count++;
mutex_exit(&delq->z_mutex);
return;
}
if (xzp) {
dmu_buf_will_dirty(xzp->z_dbuf, tx);
mutex_enter(&xzp->z_lock);
xzp->z_reap = 1; /* mark xzp for deletion */
xzp->z_phys->zp_links = 0; /* no more links to it */
mutex_exit(&xzp->z_lock);
zfs_dq_add(xzp, tx); /* add xzp to delete queue */
}
/*
* Remove this znode from delete queue
*/
error = zap_remove(os, zfsvfs->z_dqueue,
zfs_dq_hexname(obj_name, zp->z_id), tx);
ASSERT3U(error, ==, 0);
zfs_znode_delete(zp, tx);
dmu_tx_commit(tx);
if (xzp)
VN_RELE(ZTOV(xzp));
}
