int
spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
{
int err = 0;
dmu_tx_t *tx;
nvlist_t *nvarg;
if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
return (SET_ERROR(EINVAL));
tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err) {
dmu_tx_abort(tx);
return (err);
}
VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP));
if (spa_history_zone() != NULL) {
fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
spa_history_zone());
}
fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
/* Kick this off asynchronously; errors are ignored. */
dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
dmu_tx_commit(tx);
/* spa_history_log_sync will free nvl */
return (err);
}
/*
* The nvlist will be consumed by this call.
*/
static void
log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
dmu_tx_t *tx, const char *fmt, va_list adx)
{
char *msg;
/*
* If this is part of creating a pool, not everything is
* initialized yet, so don't bother logging the internal events.
* Likewise if the pool is not writeable.
*/
if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
fnvlist_free(nvl);
return;
}
msg = kmem_vasprintf(fmt, adx);
fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
strfree(msg);
fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);
if (dmu_tx_is_syncing(tx)) {
spa_history_log_sync(nvl, tx);
} else {
dsl_sync_task_nowait(spa_get_dsl(spa),
spa_history_log_sync, nvl, 0, ZFS_SPACE_CHECK_NONE, tx);
}
/* spa_history_log_sync() will free nvl */
}
void
dsl_sync_task_do_nowait(dsl_pool_t *dp,
dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx)
{
dsl_sync_task_group_t *dstg;
if (!spa_writeable(dp->dp_spa))
return;
dstg = dsl_sync_task_group_create(dp);
dsl_sync_task_create(dstg, checkfunc, syncfunc,
arg1, arg2, blocks_modified);
dsl_sync_task_group_nowait(dstg, tx);
}
static int
zvol_first_open(zvol_state_t *zv)
{
objset_t *os;
uint64_t volsize;
int error;
uint64_t ro;
/* lie and say we're read-only */
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
if (error)
return (SET_ERROR(-error));
zv->zv_objset = os;
error = dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL);
if (error)
goto out_owned;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_owned;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error)
goto out_owned;
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
zv->zv_zilog = zil_open(os, zvol_get_data);
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
set_disk_ro(zv->zv_disk, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
set_disk_ro(zv->zv_disk, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
out_owned:
if (error) {
dmu_objset_disown(os, zvol_tag);
zv->zv_objset = NULL;
}
return (SET_ERROR(-error));
}
void
mmp_thread_start(spa_t *spa)
{
mmp_thread_t *mmp = &spa->spa_mmp;
if (spa_writeable(spa)) {
mutex_enter(&mmp->mmp_thread_lock);
if (!mmp->mmp_thread) {
dprintf("mmp_thread_start pool %s\n",
spa->spa_name);
mmp->mmp_thread = thread_create(NULL, 0, mmp_thread,
spa, 0, &p0, TS_RUN, defclsyspri);
}
mutex_exit(&mmp->mmp_thread_lock);
}
}
int
dsl_sync_task_do(dsl_pool_t *dp,
dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
void *arg1, void *arg2, int blocks_modified)
{
dsl_sync_task_group_t *dstg;
int err;
ASSERT(spa_writeable(dp->dp_spa));
dstg = dsl_sync_task_group_create(dp);
dsl_sync_task_create(dstg, checkfunc, syncfunc,
arg1, arg2, blocks_modified);
err = dsl_sync_task_group_wait(dstg);
dsl_sync_task_group_destroy(dstg);
return (err);
}
static void
traverse_zil(traverse_data_t *td, zil_header_t *zh)
{
uint64_t claim_txg = zh->zh_claim_txg;
zilog_t *zilog;
/*
* We only want to visit blocks that have been claimed but not yet
* replayed; plus, in read-only mode, blocks that are already stable.
*/
if (claim_txg == 0 && spa_writeable(td->td_spa))
return;
zilog = zil_alloc(spa_get_dsl(td->td_spa)->dp_meta_objset, zh);
(void) zil_parse(zilog, traverse_zil_block, traverse_zil_record, td,
claim_txg);
zil_free(zilog);
}
static void
dsl_scan_zil(dsl_pool_t *dp, zil_header_t *zh)
{
uint64_t claim_txg = zh->zh_claim_txg;
zil_scan_arg_t zsa = { dp, zh };
zilog_t *zilog;
/*
* We only want to visit blocks that have been claimed but not yet
* replayed (or, in read-only mode, blocks that *would* be claimed).
*/
if (claim_txg == 0 && spa_writeable(dp->dp_spa))
return;
zilog = zil_alloc(dp->dp_meta_objset, zh);
(void) zil_parse(zilog, dsl_scan_zil_block, dsl_scan_zil_record, &zsa,
claim_txg);
zil_free(zilog);
}
/*
* The nvlist will be consumed by this call.
*/
static void
log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
dmu_tx_t *tx, const char *fmt, va_list adx)
{
char *msg;
va_list adx1;
int size;
/*
* If this is part of creating a pool, not everything is
* initialized yet, so don't bother logging the internal events.
* Likewise if the pool is not writeable.
*/
if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
fnvlist_free(nvl);
return;
}
va_copy(adx1, adx);
size = vsnprintf(NULL, 0, fmt, adx1) + 1;
msg = kmem_alloc(size, KM_PUSHPAGE);
va_end(adx1);
va_copy(adx1, adx);
(void) vsprintf(msg, fmt, adx1);
va_end(adx1);
fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
kmem_free(msg, size);
fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);
if (dmu_tx_is_syncing(tx)) {
spa_history_log_sync(nvl, tx);
} else {
dsl_sync_task_nowait(spa_get_dsl(spa),
spa_history_log_sync, nvl, 0, tx);
}
/* spa_history_log_sync() will free nvl */
}
int
zfs_sb_setup(zfs_sb_t *zsb, boolean_t mounting)
{
int error;
error = zfs_register_callbacks(zsb);
if (error)
return (error);
/*
* Set the objset user_ptr to track its zsb.
*/
mutex_enter(&zsb->z_os->os_user_ptr_lock);
dmu_objset_set_user(zsb->z_os, zsb);
mutex_exit(&zsb->z_os->os_user_ptr_lock);
zsb->z_log = zil_open(zsb->z_os, zfs_get_data);
/*
* If we are not mounting (ie: online recv), then we don't
* have to worry about replaying the log as we blocked all
* operations out since we closed the ZIL.
*/
if (mounting) {
boolean_t readonly;
/*
* During replay we remove the read only flag to
* allow replays to succeed.
*/
readonly = zfs_is_readonly(zsb);
if (readonly != 0)
readonly_changed_cb(zsb, B_FALSE);
else
zfs_unlinked_drain(zsb);
/*
* Parse and replay the intent log.
*
* Because of ziltest, this must be done after
* zfs_unlinked_drain(). (Further note: ziltest
* doesn't use readonly mounts, where
* zfs_unlinked_drain() isn't called.) This is because
* ziltest causes spa_sync() to think it's committed,
* but actually it is not, so the intent log contains
* many txg's worth of changes.
*
* In particular, if object N is in the unlinked set in
* the last txg to actually sync, then it could be
* actually freed in a later txg and then reallocated
* in a yet later txg. This would write a "create
* object N" record to the intent log. Normally, this
* would be fine because the spa_sync() would have
* written out the fact that object N is free, before
* we could write the "create object N" intent log
* record.
*
* But when we are in ziltest mode, we advance the "open
* txg" without actually spa_sync()-ing the changes to
* disk. So we would see that object N is still
* allocated and in the unlinked set, and there is an
* intent log record saying to allocate it.
*/
if (spa_writeable(dmu_objset_spa(zsb->z_os))) {
if (zil_replay_disable) {
zil_destroy(zsb->z_log, B_FALSE);
} else {
zsb->z_replay = B_TRUE;
zil_replay(zsb->z_os, zsb,
zfs_replay_vector);
zsb->z_replay = B_FALSE;
}
}
/* restore readonly bit */
if (readonly != 0)
readonly_changed_cb(zsb, B_TRUE);
}
return (0);
}
int
zfs_register_callbacks(zfs_sb_t *zsb)
{
struct dsl_dataset *ds = NULL;
objset_t *os = zsb->z_os;
zfs_mntopts_t *zmo = zsb->z_mntopts;
int error = 0;
ASSERT(zsb);
ASSERT(zmo);
/*
* The act of registering our callbacks will destroy any mount
* options we may have. In order to enable temporary overrides
* of mount options, we stash away the current values and
* restore them after we register the callbacks.
*/
if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os))) {
zmo->z_do_readonly = B_TRUE;
zmo->z_readonly = B_TRUE;
}
/*
* Register property callbacks.
*
* It would probably be fine to just check for i/o error from
* the first prop_register(), but I guess I like to go
* overboard...
*/
ds = dmu_objset_ds(os);
dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
error = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_RELATIME), relatime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_ACLTYPE), acltype_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
if (error)
goto unregister;
/*
* Invoke our callbacks to restore temporary mount options.
*/
if (zmo->z_do_readonly)
readonly_changed_cb(zsb, zmo->z_readonly);
if (zmo->z_do_setuid)
setuid_changed_cb(zsb, zmo->z_setuid);
if (zmo->z_do_exec)
exec_changed_cb(zsb, zmo->z_exec);
if (zmo->z_do_devices)
devices_changed_cb(zsb, zmo->z_devices);
if (zmo->z_do_xattr)
xattr_changed_cb(zsb, zmo->z_xattr);
if (zmo->z_do_atime)
atime_changed_cb(zsb, zmo->z_atime);
if (zmo->z_do_relatime)
relatime_changed_cb(zsb, zmo->z_relatime);
if (zmo->z_do_nbmand)
nbmand_changed_cb(zsb, zmo->z_nbmand);
return (0);
unregister:
/*
* We may attempt to unregister some callbacks that are not
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
atime_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RELATIME),
relatime_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
xattr_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
blksz_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
readonly_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
devices_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID),
setuid_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC),
exec_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR),
snapdir_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLTYPE),
acltype_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT),
acl_inherit_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN),
vscan_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND),
nbmand_changed_cb, zsb);
return (error);
}
static int
vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *ashift)
{
spa_t *spa = vd->vdev_spa;
vdev_disk_t *dvd;
struct dk_minfo_ext dkmext;
int error;
dev_t dev;
int otyp;
boolean_t validate_devid = B_FALSE;
ddi_devid_t devid;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (vd->vdev_tsd != NULL) {
ASSERT(vd->vdev_reopening);
dvd = vd->vdev_tsd;
goto skip_open;
}
dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP);
/*
* When opening a disk device, we want to preserve the user's original
* intent. We always want to open the device by the path the user gave
* us, even if it is one of multiple paths to the save device. But we
* also want to be able to survive disks being removed/recabled.
* Therefore the sequence of opening devices is:
*
* 1. Try opening the device by path. For legacy pools without the
* 'whole_disk' property, attempt to fix the path by appending 's0'.
*
* 2. If the devid of the device matches the stored value, return
* success.
*
* 3. Otherwise, the device may have moved. Try opening the device
* by the devid instead.
*/
if (vd->vdev_devid != NULL) {
if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid,
&dvd->vd_minor) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
}
error = EINVAL; /* presume failure */
if (vd->vdev_path != NULL) {
if (vd->vdev_wholedisk == -1ULL) {
size_t len = strlen(vd->vdev_path) + 3;
char *buf = kmem_alloc(len, KM_SLEEP);
ldi_handle_t lh;
(void) snprintf(buf, len, "%ss0", vd->vdev_path);
if (ldi_open_by_name(buf, spa_mode(spa), kcred,
&lh, zfs_li) == 0) {
spa_strfree(vd->vdev_path);
vd->vdev_path = buf;
vd->vdev_wholedisk = 1ULL;
(void) ldi_close(lh, spa_mode(spa), kcred);
} else {
kmem_free(buf, len);
}
}
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa), kcred,
&dvd->vd_lh, zfs_li);
/*
* Compare the devid to the stored value.
*/
if (error == 0 && vd->vdev_devid != NULL &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
error = SET_ERROR(EINVAL);
(void) ldi_close(dvd->vd_lh, spa_mode(spa),
kcred);
dvd->vd_lh = NULL;
}
ddi_devid_free(devid);
}
/*
* If we succeeded in opening the device, but 'vdev_wholedisk'
* is not yet set, then this must be a slice.
*/
if (error == 0 && vd->vdev_wholedisk == -1ULL)
vd->vdev_wholedisk = 0;
}
/*
* If we were unable to open by path, or the devid check fails, open by
* devid instead.
*/
if (error != 0 && vd->vdev_devid != NULL) {
error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor,
spa_mode(spa), kcred, &dvd->vd_lh, zfs_li);
}
/*
* If all else fails, then try opening by physical path (if available)
* or the logical path (if we failed due to the devid check). While not
* as reliable as the devid, this will give us something, and the higher
* level vdev validation will prevent us from opening the wrong device.
*/
if (error) {
if (vd->vdev_devid != NULL)
validate_devid = B_TRUE;
if (vd->vdev_physpath != NULL &&
(dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != NODEV)
error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
/*
* Note that we don't support the legacy auto-wholedisk support
* as above. This hasn't been used in a very long time and we
* don't need to propagate its oddities to this edge condition.
*/
if (error && vd->vdev_path != NULL)
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
}
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (error);
}
/*
* Now that the device has been successfully opened, update the devid
* if necessary.
*/
if (validate_devid && spa_writeable(spa) &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
char *vd_devid;
vd_devid = ddi_devid_str_encode(devid, dvd->vd_minor);
zfs_dbgmsg("vdev %s: update devid from %s, "
"to %s", vd->vdev_path, vd->vdev_devid, vd_devid);
spa_strfree(vd->vdev_devid);
vd->vdev_devid = spa_strdup(vd_devid);
ddi_devid_str_free(vd_devid);
}
ddi_devid_free(devid);
}
/*
* Once a device is opened, verify that the physical device path (if
* available) is up to date.
*/
if (ldi_get_dev(dvd->vd_lh, &dev) == 0 &&
ldi_get_otyp(dvd->vd_lh, &otyp) == 0) {
char *physpath, *minorname;
physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP);
minorname = NULL;
if (ddi_dev_pathname(dev, otyp, physpath) == 0 &&
ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 &&
(vd->vdev_physpath == NULL ||
strcmp(vd->vdev_physpath, physpath) != 0)) {
if (vd->vdev_physpath)
spa_strfree(vd->vdev_physpath);
(void) strlcat(physpath, ":", MAXPATHLEN);
(void) strlcat(physpath, minorname, MAXPATHLEN);
vd->vdev_physpath = spa_strdup(physpath);
}
if (minorname)
kmem_free(minorname, strlen(minorname) + 1);
kmem_free(physpath, MAXPATHLEN);
}
skip_open:
/*
* Determine the actual size of the device.
*/
if (ldi_get_size(dvd->vd_lh, psize) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
return (SET_ERROR(EINVAL));
}
/*
* Determine the device's minimum transfer size.
* If the ioctl isn't supported, assume DEV_BSIZE.
*/
if (ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFOEXT, (intptr_t)&dkmext,
FKIOCTL, kcred, NULL) != 0)
dkmext.dki_pbsize = DEV_BSIZE;
*ashift = highbit(MAX(dkmext.dki_pbsize, SPA_MINBLOCKSIZE)) - 1;
if (vd->vdev_wholedisk == 1) {
uint64_t capacity = dkmext.dki_capacity - 1;
uint64_t blksz = dkmext.dki_lbsize;
int wce = 1;
/*
* If we own the whole disk, try to enable disk write caching.
* We ignore errors because it's OK if we can't do it.
*/
(void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce,
FKIOCTL, kcred, NULL);
*max_psize = *psize + vdev_disk_get_space(vd, capacity, blksz);
zfs_dbgmsg("capacity change: vdev %s, psize %llu, "
"max_psize %llu", vd->vdev_path, *psize, *max_psize);
} else {
*max_psize = *psize;
}
/*
* Clear the nowritecache bit, so that on a vdev_reopen() we will
* try again.
*/
vd->vdev_nowritecache = B_FALSE;
return (0);
}
static int
__zvol_create_minor(const char *name, boolean_t ignore_snapdev)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
unsigned minor = 0;
int error = 0;
ASSERT(MUTEX_HELD(&zvol_state_lock));
zv = zvol_find_by_name(name);
if (zv) {
error = EEXIST;
goto out;
}
if (ignore_snapdev == B_FALSE) {
error = __zvol_snapdev_hidden(name);
if (error)
goto out;
}
doi = kmem_alloc(sizeof(dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = EAGAIN;
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_queue, UINT_MAX);
blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
#ifdef HAVE_BLK_QUEUE_DISCARD
blk_queue_max_discard_sectors(zv->zv_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#endif
#ifdef HAVE_BLK_QUEUE_NONROT
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
if (spa_writeable(dmu_objset_spa(os))) {
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
}
zv->zv_objset = NULL;
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
out_doi:
kmem_free(doi, sizeof(dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
add_disk(zv->zv_disk);
}
return (error);
}
static int
vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *ashift)
{
spa_t *spa = vd->vdev_spa;
vdev_disk_t *dvd = vd->vdev_tsd;
ldi_ev_cookie_t ecookie;
vdev_disk_ldi_cb_t *lcb;
union {
struct dk_minfo_ext ude;
struct dk_minfo ud;
} dks;
struct dk_minfo_ext *dkmext = &dks.ude;
struct dk_minfo *dkm = &dks.ud;
int error;
/* XXX Apple - must leave devid unchanged */
#ifdef illumos
dev_t dev;
int otyp;
boolean_t validate_devid = B_FALSE;
ddi_devid_t devid;
#endif
uint64_t capacity = 0, blksz = 0, pbsize;
#ifdef __APPLE__
int isssd;
#endif
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (dvd != NULL) {
if (dvd->vd_ldi_offline && dvd->vd_lh == NULL) {
/*
* If we are opening a device in its offline notify
* context, the LDI handle was just closed. Clean
* up the LDI event callbacks and free vd->vdev_tsd.
*/
vdev_disk_free(vd);
} else {
ASSERT(vd->vdev_reopening);
goto skip_open;
}
}
/*
* Create vd->vdev_tsd.
*/
vdev_disk_alloc(vd);
dvd = vd->vdev_tsd;
/*
* When opening a disk device, we want to preserve the user's original
* intent. We always want to open the device by the path the user gave
* us, even if it is one of multiple paths to the same device. But we
* also want to be able to survive disks being removed/recabled.
* Therefore the sequence of opening devices is:
*
* 1. Try opening the device by path. For legacy pools without the
* 'whole_disk' property, attempt to fix the path by appending 's0'.
*
* 2. If the devid of the device matches the stored value, return
* success.
*
* 3. Otherwise, the device may have moved. Try opening the device
* by the devid instead.
*/
/*
* XXX We must not set or modify the devid as this check would prevent
* import on Solaris/illumos.
*/
#ifdef illumos
if (vd->vdev_devid != NULL) {
if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid,
&dvd->vd_minor) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
vdev_dbgmsg(vd, "vdev_disk_open: invalid "
"vdev_devid '%s'", vd->vdev_devid);
return (SET_ERROR(EINVAL));
}
}
#endif
error = EINVAL; /* presume failure */
if (vd->vdev_path != NULL) {
/*
* XXX This assumes that if vdev_path refers to a device path /dev/dsk/cNtNdN,
* then the whole disk can be found by slice 0 at path /dev/dsk/cNtNdNs0.
*/
#ifdef illumos
if (vd->vdev_wholedisk == -1ULL) {
size_t len = strlen(vd->vdev_path) + 3;
char *buf = kmem_alloc(len, KM_SLEEP);
(void) snprintf(buf, len, "%ss0", vd->vdev_path);
error = ldi_open_by_name(buf, spa_mode(spa), kcred,
&dvd->vd_lh, zfs_li);
if (error == 0) {
spa_strfree(vd->vdev_path);
vd->vdev_path = buf;
vd->vdev_wholedisk = 1ULL;
} else {
kmem_free(buf, len);
}
}
#endif
/*
* If we have not yet opened the device, try to open it by the
* specified path.
*/
if (error != 0) {
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
}
/* XXX Apple - must leave devid unchanged */
#ifdef illumos
/*
* Compare the devid to the stored value.
*/
if (error == 0 && vd->vdev_devid != NULL &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
error = SET_ERROR(EINVAL);
(void) ldi_close(dvd->vd_lh, spa_mode(spa),
kcred);
dvd->vd_lh = NULL;
}
ddi_devid_free(devid);
}
#endif
/*
* If we succeeded in opening the device, but 'vdev_wholedisk'
* is not yet set, then this must be a slice.
*/
if (error == 0 && vd->vdev_wholedisk == -1ULL)
vd->vdev_wholedisk = 0;
}
/* XXX Apple - must leave devid unchanged */
#ifdef illumos
/*
* If we were unable to open by path, or the devid check fails, open by
* devid instead.
*/
if (error != 0 && vd->vdev_devid != NULL) {
error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor,
spa_mode(spa), kcred, &dvd->vd_lh, zfs_li);
}
#endif
/*
* If all else fails, then try opening by physical path (if available)
* or the logical path (if we failed due to the devid check). While not
* as reliable as the devid, this will give us something, and the higher
* level vdev validation will prevent us from opening the wrong device.
*/
if (error) {
/* XXX Apple - must leave devid unchanged */
#ifdef illumos
if (vd->vdev_devid != NULL)
validate_devid = B_TRUE;
#endif
/* XXX Apple to do - make ddi_ interface for this, using IORegistry path */
#ifdef illumos
if (vd->vdev_physpath != NULL &&
(dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != NODEV)
error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
#endif
/*
* Note that we don't support the legacy auto-wholedisk support
* as above. This hasn't been used in a very long time and we
* don't need to propagate its oddities to this edge condition.
*/
if (error && vd->vdev_path != NULL)
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
}
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
vdev_dbgmsg(vd, "vdev_disk_open: failed to open [error=%d]",
error);
return (error);
}
/*
* XXX Apple - We must not set or modify the devid. Import on Solaris/illumos
* expects a valid devid and fails if it cannot be decoded.
*/
#ifdef illumos
/*
* Now that the device has been successfully opened, update the devid
* if necessary.
*/
if (validate_devid && spa_writeable(spa) &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
char *vd_devid;
vd_devid = ddi_devid_str_encode(devid, dvd->vd_minor);
vdev_dbgmsg(vd, "vdev_disk_open: update devid from "
"'%s' to '%s'", vd->vdev_devid, vd_devid);
spa_strfree(vd->vdev_devid);
vd->vdev_devid = spa_strdup(vd_devid);
ddi_devid_str_free(vd_devid);
}
ddi_devid_free(devid);
}
#endif
/* XXX Apple to do, needs IORegistry physpath interface */
#ifdef illumos
/*
* Once a device is opened, verify that the physical device path (if
* available) is up to date.
*/
if (ldi_get_dev(dvd->vd_lh, &dev) == 0 &&
ldi_get_otyp(dvd->vd_lh, &otyp) == 0) {
char *physpath, *minorname;
physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP);
minorname = NULL;
if (ddi_dev_pathname(dev, otyp, physpath) == 0 &&
ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 &&
(vd->vdev_physpath == NULL ||
strcmp(vd->vdev_physpath, physpath) != 0)) {
if (vd->vdev_physpath)
spa_strfree(vd->vdev_physpath);
(void) strlcat(physpath, ":", MAXPATHLEN);
(void) strlcat(physpath, minorname, MAXPATHLEN);
vd->vdev_physpath = spa_strdup(physpath);
}
if (minorname)
kmem_free(minorname, strlen(minorname) + 1);
kmem_free(physpath, MAXPATHLEN);
}
#endif
/*
* Register callbacks for the LDI offline event.
*/
if (ldi_ev_get_cookie(dvd->vd_lh, LDI_EV_OFFLINE, &ecookie) ==
LDI_EV_SUCCESS) {
lcb = kmem_zalloc(sizeof (vdev_disk_ldi_cb_t), KM_SLEEP);
list_insert_tail(&dvd->vd_ldi_cbs, lcb);
(void) ldi_ev_register_callbacks(dvd->vd_lh, ecookie,
&vdev_disk_off_callb, (void *) vd, &lcb->lcb_id);
}
/* XXX Apple to do - we could support the degrade event, or just no-op */
#ifdef illumos
/*
* Register callbacks for the LDI degrade event.
*/
if (ldi_ev_get_cookie(dvd->vd_lh, LDI_EV_DEGRADE, &ecookie) ==
LDI_EV_SUCCESS) {
lcb = kmem_zalloc(sizeof (vdev_disk_ldi_cb_t), KM_SLEEP);
list_insert_tail(&dvd->vd_ldi_cbs, lcb);
(void) ldi_ev_register_callbacks(dvd->vd_lh, ecookie,
&vdev_disk_dgrd_callb, (void *) vd, &lcb->lcb_id);
}
#endif
#if 0
int len = MAXPATHLEN;
if (vn_getpath(devvp, dvd->vd_readlinkname, &len) == 0) {
dprintf("ZFS: '%s' resolved name is '%s'\n",
vd->vdev_path, dvd->vd_readlinkname);
} else {
dvd->vd_readlinkname[0] = 0;
}
#endif
skip_open:
/*
* Determine the actual size of the device.
*/
if (ldi_get_size(dvd->vd_lh, psize) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
vdev_dbgmsg(vd, "vdev_disk_open: failed to get size");
return (SET_ERROR(EINVAL));
}
*max_psize = *psize;
/*
* Determine the device's minimum transfer size.
* If the ioctl isn't supported, assume DEV_BSIZE.
*/
if ((error = ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFOEXT,
(intptr_t)dkmext, FKIOCTL, kcred, NULL)) == 0) {
capacity = dkmext->dki_capacity - 1;
blksz = dkmext->dki_lbsize;
pbsize = dkmext->dki_pbsize;
} else if ((error = ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFO,
(intptr_t)dkm, FKIOCTL, kcred, NULL)) == 0) {
VDEV_DEBUG(
"vdev_disk_open(\"%s\"): fallback to DKIOCGMEDIAINFO\n",
vd->vdev_path);
capacity = dkm->dki_capacity - 1;
blksz = dkm->dki_lbsize;
pbsize = blksz;
} else {
VDEV_DEBUG("vdev_disk_open(\"%s\"): "
"both DKIOCGMEDIAINFO{,EXT} calls failed, %d\n",
vd->vdev_path, error);
pbsize = DEV_BSIZE;
}
*ashift = highbit64(MAX(pbsize, SPA_MINBLOCKSIZE)) - 1;
/* XXX Now that we opened the device, determine if it is a whole disk. */
#ifdef __APPLE__
/*
* XXX Apple to do - provide an ldi_ mechanism
* to report whether this is a whole disk or a
* partition.
* Return 0 (no), 1 (yes), or -1 (error).
*/
// vd->vdev_wholedisk = ldi_is_wholedisk(vd->vd_lh);
#endif
if (vd->vdev_wholedisk == 1) {
int wce = 1;
/* Gets information about the disk if it has GPT partitions */
#ifdef illumos
if (error == 0) {
/*
* If we have the capability to expand, we'd have
* found out via success from DKIOCGMEDIAINFO{,EXT}.
* Adjust max_psize upward accordingly since we know
* we own the whole disk now.
*/
*max_psize = capacity * blksz;
}
#endif
/*
* Since we own the whole disk, try to enable disk write
* caching. We ignore errors because it's OK if we can't do it.
*/
(void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce,
FKIOCTL, kcred, NULL);
}
/*
* Clear the nowritecache bit, so that on a vdev_reopen() we will
* try again.
*/
vd->vdev_nowritecache = B_FALSE;
#ifdef __APPLE__
/* Inform the ZIO pipeline that we are non-rotational */
vd->vdev_nonrot = B_FALSE;
if (ldi_ioctl(dvd->vd_lh, DKIOCISSOLIDSTATE, (intptr_t)&isssd,
FKIOCTL, kcred, NULL) == 0) {
vd->vdev_nonrot = (isssd ? B_TRUE : B_FALSE);
}
#endif //__APPLE__
return (0);
}
static int
vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
uint64_t *ashift)
{
spa_t *spa = vd->vdev_spa;
vdev_disk_t *dvd;
ldi_ev_cookie_t ecookie;
vdev_disk_ldi_cb_t *lcb;
union {
struct dk_minfo_ext ude;
struct dk_minfo ud;
} dks;
struct dk_minfo_ext *dkmext = &dks.ude;
struct dk_minfo *dkm = &dks.ud;
int error;
dev_t dev;
int otyp;
boolean_t validate_devid = B_FALSE;
ddi_devid_t devid;
uint64_t capacity = 0, blksz = 0, pbsize;
/*
* We must have a pathname, and it must be absolute.
*/
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return (SET_ERROR(EINVAL));
}
rw_enter(&vd->vdev_tsd_lock, RW_WRITER);
dvd = vd->vdev_tsd;
/*
* Reopen the device if it's not currently open. Otherwise,
* just update the physical size of the device.
*/
if (dvd != NULL) {
ASSERT(vd->vdev_reopening);
/*
* Here vd_lh is protected by vdev_tsd_lock
*/
ASSERT(dvd->vd_lh != NULL);
/* This should not happen, but let's be safe */
if (dvd->vd_lh == NULL) {
/* What are we going to do here??? */
rw_exit(&vd->vdev_tsd_lock);
return (SET_ERROR(ENXIO));
}
goto skip_open;
}
/*
* Create dvd to be used as vd->vdev_tsd.
*/
vd->vdev_tsd = dvd = vdev_disk_alloc();
/*
* When opening a disk device, we want to preserve the user's original
* intent. We always want to open the device by the path the user gave
* us, even if it is one of multiple paths to the same device. But we
* also want to be able to survive disks being removed/recabled.
* Therefore the sequence of opening devices is:
*
* 1. Try opening the device by path. For legacy pools without the
* 'whole_disk' property, attempt to fix the path by appending 's0'.
*
* 2. If the devid of the device matches the stored value, return
* success.
*
* 3. Otherwise, the device may have moved. Try opening the device
* by the devid instead.
*/
if (vd->vdev_devid != NULL) {
if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid,
&dvd->vd_minor) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
vdev_disk_free_locked(vd);
rw_exit(&vd->vdev_tsd_lock);
return (SET_ERROR(EINVAL));
}
}
error = EINVAL; /* presume failure */
if (vd->vdev_path != NULL) {
if (vd->vdev_wholedisk == -1ULL) {
size_t len = strlen(vd->vdev_path) + 3;
char *buf = kmem_alloc(len, KM_SLEEP);
(void) snprintf(buf, len, "%ss0", vd->vdev_path);
error = ldi_open_by_name(buf, spa_mode(spa), kcred,
&dvd->vd_lh, zfs_li);
if (error == 0) {
spa_strfree(vd->vdev_path);
vd->vdev_path = buf;
vd->vdev_wholedisk = 1ULL;
} else {
kmem_free(buf, len);
}
}
/*
* If we have not yet opened the device, try to open it by the
* specified path.
*/
if (error != 0) {
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
}
/*
* Compare the devid to the stored value.
*/
if (error == 0 && vd->vdev_devid != NULL &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
error = SET_ERROR(EINVAL);
(void) ldi_close(dvd->vd_lh, spa_mode(spa),
kcred);
dvd->vd_lh = NULL;
}
ddi_devid_free(devid);
}
/*
* If we succeeded in opening the device, but 'vdev_wholedisk'
* is not yet set, then this must be a slice.
*/
if (error == 0 && vd->vdev_wholedisk == -1ULL)
vd->vdev_wholedisk = 0;
}
/*
* If we were unable to open by path, or the devid check fails, open by
* devid instead.
*/
if (error != 0 && vd->vdev_devid != NULL) {
error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor,
spa_mode(spa), kcred, &dvd->vd_lh, zfs_li);
}
/*
* If all else fails, then try opening by physical path (if available)
* or the logical path (if we failed due to the devid check). While not
* as reliable as the devid, this will give us something, and the higher
* level vdev validation will prevent us from opening the wrong device.
*/
if (error) {
if (vd->vdev_devid != NULL)
validate_devid = B_TRUE;
if (vd->vdev_physpath != NULL &&
(dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != NODEV)
error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
/*
* Note that we don't support the legacy auto-wholedisk support
* as above. This hasn't been used in a very long time and we
* don't need to propagate its oddities to this edge condition.
*/
if (error && vd->vdev_path != NULL)
error = ldi_open_by_name(vd->vdev_path, spa_mode(spa),
kcred, &dvd->vd_lh, zfs_li);
}
if (error) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
vdev_disk_free_locked(vd);
rw_exit(&vd->vdev_tsd_lock);
return (error);
}
/*
* Now that the device has been successfully opened, update the devid
* if necessary.
*/
if (validate_devid && spa_writeable(spa) &&
ldi_get_devid(dvd->vd_lh, &devid) == 0) {
if (ddi_devid_compare(devid, dvd->vd_devid) != 0) {
char *vd_devid;
vd_devid = ddi_devid_str_encode(devid, dvd->vd_minor);
zfs_dbgmsg("vdev %s: update devid from %s, "
"to %s", vd->vdev_path, vd->vdev_devid, vd_devid);
spa_strfree(vd->vdev_devid);
vd->vdev_devid = spa_strdup(vd_devid);
ddi_devid_str_free(vd_devid);
}
ddi_devid_free(devid);
}
/*
* Once a device is opened, verify that the physical device path (if
* available) is up to date.
*/
if (ldi_get_dev(dvd->vd_lh, &dev) == 0 &&
ldi_get_otyp(dvd->vd_lh, &otyp) == 0) {
char *physpath, *minorname;
physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP);
minorname = NULL;
if (ddi_dev_pathname(dev, otyp, physpath) == 0 &&
ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 &&
(vd->vdev_physpath == NULL ||
strcmp(vd->vdev_physpath, physpath) != 0)) {
if (vd->vdev_physpath)
spa_strfree(vd->vdev_physpath);
(void) strlcat(physpath, ":", MAXPATHLEN);
(void) strlcat(physpath, minorname, MAXPATHLEN);
vd->vdev_physpath = spa_strdup(physpath);
}
if (minorname)
kmem_free(minorname, strlen(minorname) + 1);
kmem_free(physpath, MAXPATHLEN);
}
/*
* Register callbacks for the LDI offline event.
*/
if (ldi_ev_get_cookie(dvd->vd_lh, LDI_EV_OFFLINE, &ecookie) ==
LDI_EV_SUCCESS) {
lcb = kmem_zalloc(sizeof (vdev_disk_ldi_cb_t), KM_SLEEP);
list_insert_tail(&dvd->vd_ldi_cbs, lcb);
(void) ldi_ev_register_callbacks(dvd->vd_lh, ecookie,
&vdev_disk_off_callb, (void *) vd, &lcb->lcb_id);
}
/*
* Register callbacks for the LDI degrade event.
*/
if (ldi_ev_get_cookie(dvd->vd_lh, LDI_EV_DEGRADE, &ecookie) ==
LDI_EV_SUCCESS) {
lcb = kmem_zalloc(sizeof (vdev_disk_ldi_cb_t), KM_SLEEP);
list_insert_tail(&dvd->vd_ldi_cbs, lcb);
(void) ldi_ev_register_callbacks(dvd->vd_lh, ecookie,
&vdev_disk_dgrd_callb, (void *) vd, &lcb->lcb_id);
}
/* Reset TRIM flag, as underlying device support may have changed */
vd->vdev_notrim = B_FALSE;
skip_open:
ASSERT(dvd != NULL);
/*
* Determine the actual size of the device.
*/
if (ldi_get_size(dvd->vd_lh, psize) != 0) {
vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
vdev_disk_free_locked(vd);
rw_exit(&vd->vdev_tsd_lock);
return (SET_ERROR(EINVAL));
}
*max_psize = *psize;
/*
* Determine the device's minimum transfer size.
* If the ioctl isn't supported, assume DEV_BSIZE.
*/
if ((error = ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFOEXT,
(intptr_t)dkmext, FKIOCTL, kcred, NULL)) == 0) {
capacity = dkmext->dki_capacity - 1;
blksz = dkmext->dki_lbsize;
pbsize = dkmext->dki_pbsize;
} else if ((error = ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFO,
(intptr_t)dkm, FKIOCTL, kcred, NULL)) == 0) {
VDEV_DEBUG(
"vdev_disk_open(\"%s\"): fallback to DKIOCGMEDIAINFO\n",
vd->vdev_path);
capacity = dkm->dki_capacity - 1;
blksz = dkm->dki_lbsize;
pbsize = blksz;
} else {
VDEV_DEBUG("vdev_disk_open(\"%s\"): "
"both DKIOCGMEDIAINFO{,EXT} calls failed, %d\n",
vd->vdev_path, error);
pbsize = DEV_BSIZE;
}
*ashift = highbit64(MAX(pbsize, SPA_MINBLOCKSIZE)) - 1;
if (vd->vdev_wholedisk == 1) {
int wce = 1;
if (error == 0) {
/*
* If we have the capability to expand, we'd have
* found out via success from DKIOCGMEDIAINFO{,EXT}.
* Adjust max_psize upward accordingly since we know
* we own the whole disk now.
*/
*max_psize += vdev_disk_get_space(vd, capacity, blksz);
zfs_dbgmsg("capacity change: vdev %s, psize %llu, "
"max_psize %llu", vd->vdev_path, *psize,
*max_psize);
}
/*
* Since we own the whole disk, try to enable disk write
* caching. We ignore errors because it's OK if we can't do it.
*/
(void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce,
FKIOCTL, kcred, NULL);
}
/*
* We are done with vd_lh and vdev_tsd, release the vdev_tsd_lock
*/
rw_exit(&vd->vdev_tsd_lock);
/*
* Clear the nowritecache bit, so that on a vdev_reopen() we will
* try again.
*/
vd->vdev_nowritecache = B_FALSE;
return (0);
}
/*
* Synchronize pool configuration to disk. This must be called with the
* namespace lock held. Synchronizing the pool cache is typically done after
* the configuration has been synced to the MOS. This exposes a window where
* the MOS config will have been updated but the cache file has not. If
* the system were to crash at that instant then the cached config may not
* contain the correct information to open the pool and an explicity import
* would be required.
*/
void
spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
{
spa_config_dirent_t *dp, *tdp;
nvlist_t *nvl;
char *pool_name;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (rootdir == NULL || !(spa_mode_global & FWRITE))
return;
/*
* Iterate over all cachefiles for the pool, past or present. When the
* cachefile is changed, the new one is pushed onto this list, allowing
* us to update previous cachefiles that no longer contain this pool.
*/
for (dp = list_head(&target->spa_config_list); dp != NULL;
dp = list_next(&target->spa_config_list, dp)) {
spa_t *spa = NULL;
if (dp->scd_path == NULL)
continue;
/*
* Iterate over all pools, adding any matching pools to 'nvl'.
*/
nvl = NULL;
while ((spa = spa_next(spa)) != NULL) {
/*
* Skip over our own pool if we're about to remove
* ourselves from the spa namespace or any pool that
* is readonly. Since we cannot guarantee that a
* readonly pool would successfully import upon reboot,
* we don't allow them to be written to the cache file.
*/
if ((spa == target && removing) ||
!spa_writeable(spa))
continue;
mutex_enter(&spa->spa_props_lock);
tdp = list_head(&spa->spa_config_list);
if (spa->spa_config == NULL ||
tdp->scd_path == NULL ||
strcmp(tdp->scd_path, dp->scd_path) != 0) {
mutex_exit(&spa->spa_props_lock);
continue;
}
if (nvl == NULL)
VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
KM_SLEEP) == 0);
if (spa->spa_import_flags & ZFS_IMPORT_TEMP_NAME) {
VERIFY0(nvlist_lookup_string(spa->spa_config,
ZPOOL_CONFIG_POOL_NAME, &pool_name));
} else
pool_name = spa_name(spa);
VERIFY(nvlist_add_nvlist(nvl, pool_name,
spa->spa_config) == 0);
mutex_exit(&spa->spa_props_lock);
}
spa_config_write(dp, nvl);
nvlist_free(nvl);
}
/*
* Remove any config entries older than the current one.
*/
dp = list_head(&target->spa_config_list);
while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
list_remove(&target->spa_config_list, tdp);
if (tdp->scd_path != NULL)
spa_strfree(tdp->scd_path);
kmem_free(tdp, sizeof (spa_config_dirent_t));
}
spa_config_generation++;
if (postsysevent)
spa_event_notify(target, NULL, FM_EREPORT_ZFS_CONFIG_SYNC);
}
/*
* Synchronize pool configuration to disk. This must be called with the
* namespace lock held. Synchronizing the pool cache is typically done after
* the configuration has been synced to the MOS. This exposes a window where
* the MOS config will have been updated but the cache file has not. If
* the system were to crash at that instant then the cached config may not
* contain the correct information to open the pool and an explicity import
* would be required.
*/
void
spa_config_sync(spa_t *target, boolean_t removing, boolean_t postsysevent)
{
spa_config_dirent_t *dp, *tdp;
nvlist_t *nvl;
boolean_t ccw_failure;
int error;
ASSERT(MUTEX_HELD(&spa_namespace_lock));
if (rootdir == NULL || !(spa_mode_global & FWRITE))
return;
/*
* Iterate over all cachefiles for the pool, past or present. When the
* cachefile is changed, the new one is pushed onto this list, allowing
* us to update previous cachefiles that no longer contain this pool.
*/
ccw_failure = B_FALSE;
for (dp = list_head(&target->spa_config_list); dp != NULL;
dp = list_next(&target->spa_config_list, dp)) {
spa_t *spa = NULL;
if (dp->scd_path == NULL)
continue;
/*
* Iterate over all pools, adding any matching pools to 'nvl'.
*/
nvl = NULL;
while ((spa = spa_next(spa)) != NULL) {
nvlist_t *nvroot = NULL;
/*
* Skip over our own pool if we're about to remove
* ourselves from the spa namespace or any pool that
* is readonly. Since we cannot guarantee that a
* readonly pool would successfully import upon reboot,
* we don't allow them to be written to the cache file.
*/
if ((spa == target && removing) ||
(spa_state(spa) == POOL_STATE_ACTIVE &&
!spa_writeable(spa)))
continue;
mutex_enter(&spa->spa_props_lock);
tdp = list_head(&spa->spa_config_list);
if (spa->spa_config == NULL ||
tdp->scd_path == NULL ||
strcmp(tdp->scd_path, dp->scd_path) != 0) {
mutex_exit(&spa->spa_props_lock);
continue;
}
if (nvl == NULL)
VERIFY(nvlist_alloc(&nvl, NV_UNIQUE_NAME,
KM_SLEEP) == 0);
VERIFY(nvlist_add_nvlist(nvl, spa->spa_name,
spa->spa_config) == 0);
mutex_exit(&spa->spa_props_lock);
if (nvlist_lookup_nvlist(nvl, spa->spa_name, &nvroot) == 0)
spa_config_clean(nvroot);
}
error = spa_config_write(dp, nvl);
if (error != 0)
ccw_failure = B_TRUE;
nvlist_free(nvl);
}
if (ccw_failure) {
/*
* Keep trying so that configuration data is
* written if/when any temporary filesystem
* resource issues are resolved.
*/
if (target->spa_ccw_fail_time == 0) {
zfs_ereport_post(FM_EREPORT_ZFS_CONFIG_CACHE_WRITE,
target, NULL, NULL, 0, 0);
}
target->spa_ccw_fail_time = gethrtime();
spa_async_request(target, SPA_ASYNC_CONFIG_UPDATE);
} else {
/*
* Do not rate limit future attempts to update
* the config cache.
*/
target->spa_ccw_fail_time = 0;
}
/*
* Remove any config entries older than the current one.
*/
dp = list_head(&target->spa_config_list);
while ((tdp = list_next(&target->spa_config_list, dp)) != NULL) {
list_remove(&target->spa_config_list, tdp);
if (tdp->scd_path != NULL)
spa_strfree(tdp->scd_path);
kmem_free(tdp, sizeof (spa_config_dirent_t));
}
spa_config_generation++;
if (postsysevent)
spa_event_notify(target, NULL, ESC_ZFS_CONFIG_SYNC);
}
static int
zvol_first_open(zvol_state_t *zv)
{
objset_t *os;
uint64_t volsize;
int locked = 0;
int error;
uint64_t ro;
/*
* In all other cases the spa_namespace_lock is taken before the
* bdev->bd_mutex lock. But in this case the Linux __blkdev_get()
* function calls fops->open() with the bdev->bd_mutex lock held.
*
* To avoid a potential lock inversion deadlock we preemptively
* try to take the spa_namespace_lock(). Normally it will not
* be contended and this is safe because spa_open_common() handles
* the case where the caller already holds the spa_namespace_lock.
*
* When it is contended we risk a lock inversion if we were to
* block waiting for the lock. Luckily, the __blkdev_get()
* function allows us to return -ERESTARTSYS which will result in
* bdev->bd_mutex being dropped, reacquired, and fops->open() being
* called again. This process can be repeated safely until both
* locks are acquired.
*/
if (!mutex_owned(&spa_namespace_lock)) {
locked = mutex_tryenter(&spa_namespace_lock);
if (!locked)
return (-ERESTARTSYS);
}
/* lie and say we're read-only */
error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, 1, zvol_tag, &os);
if (error)
goto out_mutex;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error) {
dmu_objset_disown(os, zvol_tag);
goto out_mutex;
}
zv->zv_objset = os;
error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf);
if (error) {
dmu_objset_disown(os, zvol_tag);
goto out_mutex;
}
set_capacity(zv->zv_disk, volsize >> 9);
zv->zv_volsize = volsize;
zv->zv_zilog = zil_open(os, zvol_get_data);
VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &ro, NULL) == 0);
if (ro || dmu_objset_is_snapshot(os) ||
!spa_writeable(dmu_objset_spa(os))) {
set_disk_ro(zv->zv_disk, 1);
zv->zv_flags |= ZVOL_RDONLY;
} else {
set_disk_ro(zv->zv_disk, 0);
zv->zv_flags &= ~ZVOL_RDONLY;
}
out_mutex:
if (locked)
mutex_exit(&spa_namespace_lock);
return (-error);
}
/*
* Create a block device minor node and setup the linkage between it
* and the specified volume. Once this function returns the block
* device is live and ready for use.
*/
static int
zvol_create_minor_impl(const char *name)
{
zvol_state_t *zv;
objset_t *os;
dmu_object_info_t *doi;
uint64_t volsize;
uint64_t len;
unsigned minor = 0;
int error = 0;
mutex_enter(&zvol_state_lock);
zv = zvol_find_by_name(name);
if (zv) {
error = SET_ERROR(EEXIST);
goto out;
}
doi = kmem_alloc(sizeof (dmu_object_info_t), KM_SLEEP);
error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os);
if (error)
goto out_doi;
error = dmu_object_info(os, ZVOL_OBJ, doi);
if (error)
goto out_dmu_objset_disown;
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error)
goto out_dmu_objset_disown;
error = zvol_find_minor(&minor);
if (error)
goto out_dmu_objset_disown;
zv = zvol_alloc(MKDEV(zvol_major, minor), name);
if (zv == NULL) {
error = SET_ERROR(EAGAIN);
goto out_dmu_objset_disown;
}
if (dmu_objset_is_snapshot(os))
zv->zv_flags |= ZVOL_RDONLY;
zv->zv_volblocksize = doi->doi_data_block_size;
zv->zv_volsize = volsize;
zv->zv_objset = os;
set_capacity(zv->zv_disk, zv->zv_volsize >> 9);
blk_queue_max_hw_sectors(zv->zv_queue, (DMU_MAX_ACCESS / 4) >> 9);
blk_queue_max_segments(zv->zv_queue, UINT16_MAX);
blk_queue_max_segment_size(zv->zv_queue, UINT_MAX);
blk_queue_physical_block_size(zv->zv_queue, zv->zv_volblocksize);
blk_queue_io_opt(zv->zv_queue, zv->zv_volblocksize);
blk_queue_max_discard_sectors(zv->zv_queue,
(zvol_max_discard_blocks * zv->zv_volblocksize) >> 9);
blk_queue_discard_granularity(zv->zv_queue, zv->zv_volblocksize);
queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, zv->zv_queue);
#ifdef QUEUE_FLAG_NONROT
queue_flag_set_unlocked(QUEUE_FLAG_NONROT, zv->zv_queue);
#endif
#ifdef QUEUE_FLAG_ADD_RANDOM
queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, zv->zv_queue);
#endif
if (spa_writeable(dmu_objset_spa(os))) {
if (zil_replay_disable)
zil_destroy(dmu_objset_zil(os), B_FALSE);
else
zil_replay(os, zv, zvol_replay_vector);
}
/*
* When udev detects the addition of the device it will immediately
* invoke blkid(8) to determine the type of content on the device.
* Prefetching the blocks commonly scanned by blkid(8) will speed
* up this process.
*/
len = MIN(MAX(zvol_prefetch_bytes, 0), SPA_MAXBLOCKSIZE);
if (len > 0) {
dmu_prefetch(os, ZVOL_OBJ, 0, 0, len, ZIO_PRIORITY_SYNC_READ);
dmu_prefetch(os, ZVOL_OBJ, 0, volsize - len, len,
ZIO_PRIORITY_SYNC_READ);
}
zv->zv_objset = NULL;
out_dmu_objset_disown:
dmu_objset_disown(os, zvol_tag);
out_doi:
kmem_free(doi, sizeof (dmu_object_info_t));
out:
if (error == 0) {
zvol_insert(zv);
/*
* Drop the lock to prevent deadlock with sys_open() ->
* zvol_open(), which first takes bd_disk->bd_mutex and then
* takes zvol_state_lock, whereas this code path first takes
* zvol_state_lock, and then takes bd_disk->bd_mutex.
*/
mutex_exit(&zvol_state_lock);
add_disk(zv->zv_disk);
} else {
mutex_exit(&zvol_state_lock);
}
return (SET_ERROR(error));
}
int
zfs_register_callbacks(zfs_sb_t *zsb)
{
struct dsl_dataset *ds = NULL;
objset_t *os = zsb->z_os;
boolean_t do_readonly = B_FALSE;
int error = 0;
if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
do_readonly = B_TRUE;
/*
* Register property callbacks.
*
* It would probably be fine to just check for i/o error from
* the first prop_register(), but I guess I like to go
* overboard...
*/
ds = dmu_objset_ds(os);
dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
error = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_ATIME), atime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_XATTR), xattr_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_RECORDSIZE), blksz_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_READONLY), readonly_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_DEVICES), devices_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SETUID), setuid_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_EXEC), exec_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_SNAPDIR), snapdir_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_ACLINHERIT), acl_inherit_changed_cb,
zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_VSCAN), vscan_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zsb);
dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
if (error)
goto unregister;
if (do_readonly)
readonly_changed_cb(zsb, B_TRUE);
return (0);
unregister:
/*
* We may attempt to unregister some callbacks that are not
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ATIME),
atime_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_XATTR),
xattr_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE),
blksz_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_READONLY),
readonly_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_DEVICES),
devices_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SETUID),
setuid_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_EXEC),
exec_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_SNAPDIR),
snapdir_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_ACLINHERIT),
acl_inherit_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_VSCAN),
vscan_changed_cb, zsb);
(void) dsl_prop_unregister(ds, zfs_prop_to_name(ZFS_PROP_NBMAND),
nbmand_changed_cb, zsb);
return (error);
}
/*
* Read out the command history.
*/
int
spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
{
objset_t *mos = spa->spa_meta_objset;
dmu_buf_t *dbp;
uint64_t read_len, phys_read_off, phys_eof;
uint64_t leftover = 0;
spa_history_phys_t *shpp;
int err;
/*
* If the command history doesn't exist (older pool),
* that's ok, just return ENOENT.
*/
if (!spa->spa_history)
return (ENOENT);
/*
* The history is logged asynchronously, so when they request
* the first chunk of history, make sure everything has been
* synced to disk so that we get it.
*/
if (*offp == 0 && spa_writeable(spa))
txg_wait_synced(spa_get_dsl(spa), 0);
if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0)
return (err);
shpp = dbp->db_data;
#ifdef ZFS_DEBUG
{
dmu_object_info_t doi;
dmu_object_info_from_db(dbp, &doi);
ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
}
#endif
mutex_enter(&spa->spa_history_lock);
phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
if (*offp < shpp->sh_pool_create_len) {
/* read in just the zpool create history */
phys_read_off = *offp;
read_len = MIN(*len, shpp->sh_pool_create_len -
phys_read_off);
} else {
/*
* Need to reset passed in offset to BOF if the passed in
* offset has since been overwritten.
*/
*offp = MAX(*offp, shpp->sh_bof);
phys_read_off = spa_history_log_to_phys(*offp, shpp);
/*
* Read up to the minimum of what the user passed down or
* the EOF (physical or logical). If we hit physical EOF,
* use 'leftover' to read from the physical BOF.
*/
if (phys_read_off <= phys_eof) {
read_len = MIN(*len, phys_eof - phys_read_off);
} else {
read_len = MIN(*len,
shpp->sh_phys_max_off - phys_read_off);
if (phys_read_off + *len > shpp->sh_phys_max_off) {
leftover = MIN(*len - read_len,
phys_eof - shpp->sh_pool_create_len);
}
}
}
/* offset for consumer to use next */
*offp += read_len + leftover;
/* tell the consumer how much you actually read */
*len = read_len + leftover;
if (read_len == 0) {
mutex_exit(&spa->spa_history_lock);
dmu_buf_rele(dbp, FTAG);
return (0);
}
err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
DMU_READ_PREFETCH);
if (leftover && err == 0) {
err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
leftover, buf + read_len, DMU_READ_PREFETCH);
}
mutex_exit(&spa->spa_history_lock);
dmu_buf_rele(dbp, FTAG);
return (err);
}
int
zfs_register_callbacks(zfs_sb_t *zsb)
{
struct dsl_dataset *ds = NULL;
objset_t *os = zsb->z_os;
int error = 0;
if (zfs_is_readonly(zsb) || !spa_writeable(dmu_objset_spa(os)))
readonly_changed_cb(zsb, B_TRUE);
/*
* Register property callbacks.
*
* It would probably be fine to just check for i/o error from
* the first prop_register(), but I guess I like to go
* overboard...
*/
ds = dmu_objset_ds(os);
error = dsl_prop_register(ds,
"atime", atime_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"xattr", xattr_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"recordsize", blksz_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"readonly", readonly_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"devices", devices_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"setuid", setuid_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"exec", exec_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"snapdir", snapdir_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"aclinherit", acl_inherit_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"vscan", vscan_changed_cb, zsb);
error = error ? error : dsl_prop_register(ds,
"nbmand", nbmand_changed_cb, zsb);
if (error)
goto unregister;
return (0);
unregister:
/*
* We may attempt to unregister some callbacks that are not
* registered, but this is OK; it will simply return ENOMSG,
* which we will ignore.
*/
(void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
zsb);
(void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zsb);
(void) dsl_prop_unregister(ds, "nbmand", nbmand_changed_cb, zsb);
return (error);
}