void
zfs_unregister_callbacks(zfs_sb_t *zsb)
{
objset_t *os = zsb->z_os;
struct dsl_dataset *ds;
/*
* Unregister properties.
*/
if (!dmu_objset_is_snapshot(os)) {
ds = dmu_objset_ds(os);
VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb,
zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "aclinherit",
acl_inherit_changed_cb, zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "vscan",
vscan_changed_cb, zsb) == 0);
VERIFY(dsl_prop_unregister(ds, "nbmand",
nbmand_changed_cb, zsb) == 0);
}
}
static void
zvol_last_close(zvol_state_t *zv)
{
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
dmu_buf_rele(zv->zv_dbuf, zvol_tag);
zv->zv_dbuf = NULL;
/*
* Evict cached data
*/
if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) &&
!(zv->zv_flags & ZVOL_RDONLY))
txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0);
(void) dmu_objset_evict_dbufs(zv->zv_objset);
dmu_objset_disown(zv->zv_objset, zvol_tag);
zv->zv_objset = NULL;
}
/*
* Remove minor node for the specified volume.
*/
int
zvol_remove_minor(zfs_cmd_t *zc)
{
zvol_state_t *zv;
char namebuf[30];
mutex_enter(&zvol_state_lock);
if ((zv = zvol_minor_lookup(zc->zc_name)) == NULL) {
mutex_exit(&zvol_state_lock);
return (ENXIO);
}
if (zv->zv_total_opens != 0) {
mutex_exit(&zvol_state_lock);
return (EBUSY);
}
(void) sprintf(namebuf, "%uc,raw", zv->zv_minor);
ddi_remove_minor_node(zfs_dip, namebuf);
(void) sprintf(namebuf, "%uc", zv->zv_minor);
ddi_remove_minor_node(zfs_dip, namebuf);
VERIFY(dsl_prop_unregister(dmu_objset_ds(zv->zv_objset),
"readonly", zvol_readonly_changed_cb, zv) == 0);
zil_close(zv->zv_zilog);
zv->zv_zilog = NULL;
dmu_objset_close(zv->zv_objset);
zv->zv_objset = NULL;
ddi_soft_state_free(zvol_state, zv->zv_minor);
zvol_minors--;
mutex_exit(&zvol_state_lock);
return (0);
}
/*
* Close an intent log.
*/
void
zil_close(zilog_t *zilog)
{
/*
* If the log isn't already committed, mark the objset dirty
* (so zil_sync() will be called) and wait for that txg to sync.
*/
if (!zil_is_committed(zilog)) {
uint64_t txg;
dmu_tx_t *tx = dmu_tx_create(zilog->zl_os);
(void) dmu_tx_assign(tx, TXG_WAIT);
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
dmu_tx_commit(tx);
txg_wait_synced(zilog->zl_dmu_pool, txg);
}
taskq_destroy(zilog->zl_clean_taskq);
zilog->zl_clean_taskq = NULL;
zilog->zl_get_data = NULL;
zil_itx_clean(zilog);
ASSERT(list_head(&zilog->zl_itx_list) == NULL);
}
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);
}
/*
* Teardown the zfs_sb_t.
*
* Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
* and 'z_teardown_inactive_lock' held.
*/
int
zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting)
{
znode_t *zp;
/*
* If someone has not already unmounted this file system,
* drain the iput_taskq to ensure all active references to the
* zfs_sb_t have been handled only then can it be safely destroyed.
*/
if (zsb->z_os) {
/*
* If we're unmounting we have to wait for the list to
* drain completely.
*
* If we're not unmounting there's no guarantee the list
* will drain completely, but iputs run from the taskq
* may add the parents of dir-based xattrs to the taskq
* so we want to wait for these.
*
* We can safely read z_nr_znodes without locking because the
* VFS has already blocked operations which add to the
* z_all_znodes list and thus increment z_nr_znodes.
*/
int round = 0;
while (zsb->z_nr_znodes > 0) {
taskq_wait_outstanding(dsl_pool_iput_taskq(
dmu_objset_pool(zsb->z_os)), 0);
if (++round > 1 && !unmounting)
break;
}
}
rrm_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
if (!unmounting) {
/*
* We purge the parent filesystem's super block as the
* parent filesystem and all of its snapshots have their
* inode's super block set to the parent's filesystem's
* super block. Note, 'z_parent' is self referential
* for non-snapshots.
*/
shrink_dcache_sb(zsb->z_parent->z_sb);
}
/*
* Close the zil. NB: Can't close the zil while zfs_inactive
* threads are blocked as zil_close can call zfs_inactive.
*/
if (zsb->z_log) {
zil_close(zsb->z_log);
zsb->z_log = NULL;
}
rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER);
/*
* If we are not unmounting (ie: online recv) and someone already
* unmounted this file system while we were doing the switcheroo,
* or a reopen of z_os failed then just bail out now.
*/
if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
rw_exit(&zsb->z_teardown_inactive_lock);
rrm_exit(&zsb->z_teardown_lock, FTAG);
return (SET_ERROR(EIO));
}
/*
* At this point there are no VFS ops active, and any new VFS ops
* will fail with EIO since we have z_teardown_lock for writer (only
* relevant for forced unmount).
*
* Release all holds on dbufs.
*/
if (!unmounting) {
mutex_enter(&zsb->z_znodes_lock);
for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
zp = list_next(&zsb->z_all_znodes, zp)) {
if (zp->z_sa_hdl)
zfs_znode_dmu_fini(zp);
}
mutex_exit(&zsb->z_znodes_lock);
}
/*
* If we are unmounting, set the unmounted flag and let new VFS ops
* unblock. zfs_inactive will have the unmounted behavior, and all
* other VFS ops will fail with EIO.
*/
if (unmounting) {
zsb->z_unmounted = B_TRUE;
rrm_exit(&zsb->z_teardown_lock, FTAG);
rw_exit(&zsb->z_teardown_inactive_lock);
}
/*
* z_os will be NULL if there was an error in attempting to reopen
* zsb, so just return as the properties had already been
*
* unregistered and cached data had been evicted before.
*/
if (zsb->z_os == NULL)
return (0);
/*
* Unregister properties.
*/
zfs_unregister_callbacks(zsb);
/*
* Evict cached data
*/
if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) &&
!zfs_is_readonly(zsb))
txg_wait_synced(dmu_objset_pool(zsb->z_os), 0);
dmu_objset_evict_dbufs(zsb->z_os);
return (0);
}
uint64_t sa_obj; ASSERT3U(spa_version(dmu_objset_spa(zsb->z_os)), >=, SPA_VERSION_SA); sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, DMU_OT_NONE, 0, tx); error = zap_add(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); ASSERT0(error); VERIFY(0 == sa_set_sa_object(os, sa_obj)); sa_register_update_callback(os, zfs_sa_upgrade); } spa_history_log_internal_ds(dmu_objset_ds(os), "upgrade", tx, "from %llu to %llu", zsb->z_version, newvers); dmu_tx_commit(tx); zsb->z_version = newvers; zfs_set_fuid_feature(zsb); return (0); } EXPORT_SYMBOL(zfs_set_version); /* * Read a property stored within the master node. */
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);
}
/*
* Teardown the zfs_sb_t.
*
* Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock'
* and 'z_teardown_inactive_lock' held.
*/
int
zfs_sb_teardown(zfs_sb_t *zsb, boolean_t unmounting)
{
znode_t *zp;
rrw_enter(&zsb->z_teardown_lock, RW_WRITER, FTAG);
if (!unmounting) {
/*
* We purge the parent filesystem's super block as the
* parent filesystem and all of its snapshots have their
* inode's super block set to the parent's filesystem's
* super block. Note, 'z_parent' is self referential
* for non-snapshots.
*/
shrink_dcache_sb(zsb->z_parent->z_sb);
}
/*
* If someone has not already unmounted this file system,
* drain the iput_taskq to ensure all active references to the
* zfs_sb_t have been handled only then can it be safely destroyed.
*/
if (zsb->z_os)
taskq_wait(dsl_pool_iput_taskq(dmu_objset_pool(zsb->z_os)));
/*
* Close the zil. NB: Can't close the zil while zfs_inactive
* threads are blocked as zil_close can call zfs_inactive.
*/
if (zsb->z_log) {
zil_close(zsb->z_log);
zsb->z_log = NULL;
}
rw_enter(&zsb->z_teardown_inactive_lock, RW_WRITER);
/*
* If we are not unmounting (ie: online recv) and someone already
* unmounted this file system while we were doing the switcheroo,
* or a reopen of z_os failed then just bail out now.
*/
if (!unmounting && (zsb->z_unmounted || zsb->z_os == NULL)) {
rw_exit(&zsb->z_teardown_inactive_lock);
rrw_exit(&zsb->z_teardown_lock, FTAG);
return (EIO);
}
/*
* At this point there are no VFS ops active, and any new VFS ops
* will fail with EIO since we have z_teardown_lock for writer (only
* relevant for forced unmount).
*
* Release all holds on dbufs.
*/
mutex_enter(&zsb->z_znodes_lock);
for (zp = list_head(&zsb->z_all_znodes); zp != NULL;
zp = list_next(&zsb->z_all_znodes, zp)) {
if (zp->z_sa_hdl) {
ASSERT(atomic_read(&ZTOI(zp)->i_count) > 0);
zfs_znode_dmu_fini(zp);
}
}
mutex_exit(&zsb->z_znodes_lock);
/*
* If we are unmounting, set the unmounted flag and let new VFS ops
* unblock. zfs_inactive will have the unmounted behavior, and all
* other VFS ops will fail with EIO.
*/
if (unmounting) {
zsb->z_unmounted = B_TRUE;
rrw_exit(&zsb->z_teardown_lock, FTAG);
rw_exit(&zsb->z_teardown_inactive_lock);
}
/*
* z_os will be NULL if there was an error in attempting to reopen
* zsb, so just return as the properties had already been
*
* unregistered and cached data had been evicted before.
*/
if (zsb->z_os == NULL)
return (0);
/*
* Unregister properties.
*/
zfs_unregister_callbacks(zsb);
/*
* Evict cached data
*/
if (dsl_dataset_is_dirty(dmu_objset_ds(zsb->z_os)) &&
!zfs_is_readonly(zsb))
txg_wait_synced(dmu_objset_pool(zsb->z_os), 0);
dmu_objset_evict_dbufs(zsb->z_os);
return (0);
}
static int
zfs_register_callbacks(vfs_t *vfsp)
{
struct dsl_dataset *ds = NULL;
objset_t *os = NULL;
zfsvfs_t *zfsvfs = NULL;
int do_readonly = FALSE, readonly;
int do_setuid = FALSE, setuid;
int do_exec = FALSE, exec;
int do_devices = FALSE, devices;
int error = 0;
ASSERT(vfsp);
zfsvfs = vfsp->vfs_data;
ASSERT(zfsvfs);
os = zfsvfs->z_os;
/*
* 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
* restore them after we register the callbacks.
*/
if (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) {
readonly = B_TRUE;
do_readonly = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
readonly = B_FALSE;
do_readonly = B_TRUE;
}
if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
devices = B_FALSE;
setuid = B_FALSE;
do_devices = B_TRUE;
do_setuid = B_TRUE;
} else {
if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) {
devices = B_FALSE;
do_devices = B_TRUE;
} else if (vfs_optionisset(vfsp,
MNTOPT_DEVICES, NULL)) {
devices = B_TRUE;
do_devices = B_TRUE;
}
if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
setuid = B_FALSE;
do_setuid = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
setuid = B_TRUE;
do_setuid = B_TRUE;
}
}
if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
exec = B_FALSE;
do_exec = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
exec = B_TRUE;
do_exec = 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, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"recordsize", blksz_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"readonly", readonly_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"devices", devices_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"setuid", setuid_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"exec", exec_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"snapdir", snapdir_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclmode", acl_mode_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclinherit", acl_inherit_changed_cb, zfsvfs);
if (error)
goto unregister;
/*
* Invoke our callbacks to restore temporary mount options.
*/
if (do_readonly)
readonly_changed_cb(zfsvfs, readonly);
if (do_setuid)
setuid_changed_cb(zfsvfs, setuid);
if (do_exec)
exec_changed_cb(zfsvfs, exec);
if (do_devices)
devices_changed_cb(zfsvfs, devices);
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, zfsvfs);
(void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
zfsvfs);
return (error);
}
/*
* Create a minor node for the specified volume.
*/
int
zvol_create_minor(zfs_cmd_t *zc)
{
char *name = zc->zc_name;
dev_t dev = zc->zc_dev;
zvol_state_t *zv;
objset_t *os;
uint64_t volsize;
minor_t minor = 0;
struct pathname linkpath;
int ds_mode = DS_MODE_PRIMARY;
vnode_t *vp = NULL;
char *devpath;
size_t devpathlen = strlen(ZVOL_FULL_DEV_DIR) + 1 + strlen(name) + 1;
char chrbuf[30], blkbuf[30];
int error;
mutex_enter(&zvol_state_lock);
if ((zv = zvol_minor_lookup(name)) != NULL) {
mutex_exit(&zvol_state_lock);
return (EEXIST);
}
if (strchr(name, '@') != 0)
ds_mode |= DS_MODE_READONLY;
error = dmu_objset_open(name, DMU_OST_ZVOL, ds_mode, &os);
if (error) {
mutex_exit(&zvol_state_lock);
return (error);
}
error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize);
if (error) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (error);
}
/*
* If there's an existing /dev/zvol symlink, try to use the
* same minor number we used last time.
*/
devpath = kmem_alloc(devpathlen, KM_SLEEP);
(void) sprintf(devpath, "%s/%s", ZVOL_FULL_DEV_DIR, name);
error = lookupname(devpath, UIO_SYSSPACE, NO_FOLLOW, NULL, &vp);
kmem_free(devpath, devpathlen);
if (error == 0 && vp->v_type != VLNK)
error = EINVAL;
if (error == 0) {
pn_alloc(&linkpath);
error = pn_getsymlink(vp, &linkpath, kcred);
if (error == 0) {
char *ms = strstr(linkpath.pn_path, ZVOL_PSEUDO_DEV);
if (ms != NULL) {
ms += strlen(ZVOL_PSEUDO_DEV);
minor = stoi(&ms);
}
}
pn_free(&linkpath);
}
if (vp != NULL)
VN_RELE(vp);
/*
* If we found a minor but it's already in use, we must pick a new one.
*/
if (minor != 0 && ddi_get_soft_state(zvol_state, minor) != NULL)
minor = 0;
if (minor == 0)
minor = zvol_minor_alloc();
if (minor == 0) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (ENXIO);
}
if (ddi_soft_state_zalloc(zvol_state, minor) != DDI_SUCCESS) {
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
(void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, name);
(void) sprintf(chrbuf, "%uc,raw", minor);
if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR,
minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
ddi_soft_state_free(zvol_state, minor);
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
(void) sprintf(blkbuf, "%uc", minor);
if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK,
minor, DDI_PSEUDO, 0) == DDI_FAILURE) {
ddi_remove_minor_node(zfs_dip, chrbuf);
ddi_soft_state_free(zvol_state, minor);
dmu_objset_close(os);
mutex_exit(&zvol_state_lock);
return (EAGAIN);
}
zv = ddi_get_soft_state(zvol_state, minor);
(void) strcpy(zv->zv_name, name);
zv->zv_min_bs = DEV_BSHIFT;
zv->zv_minor = minor;
zv->zv_volsize = volsize;
zv->zv_objset = os;
zv->zv_mode = ds_mode;
zv->zv_zilog = zil_open(os, NULL);
rw_init(&zv->zv_dslock, NULL, RW_DEFAULT, NULL);
zil_replay(os, zv, &zv->zv_txg_assign, zvol_replay_vector, NULL);
zvol_size_changed(zv, dev);
/* XXX this should handle the possible i/o error */
VERIFY(dsl_prop_register(dmu_objset_ds(zv->zv_objset),
"readonly", zvol_readonly_changed_cb, zv) == 0);
zvol_minors++;
mutex_exit(&zvol_state_lock);
return (0);
}
static int
zfs_register_callbacks(vfs_t *vfsp)
{
struct dsl_dataset *ds = NULL;
objset_t *os = NULL;
zfsvfs_t *zfsvfs = NULL;
uint64_t nbmand;
int readonly, do_readonly = FALSE;
int setuid, do_setuid = FALSE;
int exec, do_exec = FALSE;
int xattr, do_xattr = FALSE;
int atime, do_atime = FALSE;
int error = 0;
ASSERT(vfsp);
zfsvfs = vfsp->vfs_data;
ASSERT(zfsvfs);
os = zfsvfs->z_os;
/*
* This function can be called for a snapshot when we update snapshot's
* mount point, which isn't really supported.
*/
if (dmu_objset_is_snapshot(os))
return (EOPNOTSUPP);
/*
* 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 (vfs_optionisset(vfsp, MNTOPT_RO, NULL)) {
readonly = B_TRUE;
do_readonly = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) {
readonly = B_FALSE;
do_readonly = B_TRUE;
}
if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) {
setuid = B_FALSE;
do_setuid = B_TRUE;
} else {
if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) {
setuid = B_FALSE;
do_setuid = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) {
setuid = B_TRUE;
do_setuid = B_TRUE;
}
}
if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) {
exec = B_FALSE;
do_exec = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) {
exec = B_TRUE;
do_exec = B_TRUE;
}
if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) {
xattr = B_FALSE;
do_xattr = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) {
xattr = B_TRUE;
do_xattr = B_TRUE;
}
if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) {
atime = B_FALSE;
do_atime = B_TRUE;
} else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) {
atime = B_TRUE;
do_atime = B_TRUE;
}
/*
* nbmand is a special property. It can only be changed at
* mount time.
*
* This is weird, but it is documented to only be changeable
* at mount time.
*/
if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) {
nbmand = B_FALSE;
} else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) {
nbmand = B_TRUE;
} else {
char osname[MAXNAMELEN];
dmu_objset_name(os, osname);
if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand,
NULL)) {
return (error);
}
}
/*
* 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, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"xattr", xattr_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"recordsize", blksz_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"readonly", readonly_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"setuid", setuid_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"exec", exec_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"snapdir", snapdir_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclmode", acl_mode_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"aclinherit", acl_inherit_changed_cb, zfsvfs);
error = error ? error : dsl_prop_register(ds,
"vscan", vscan_changed_cb, zfsvfs);
if (error)
goto unregister;
/*
* Invoke our callbacks to restore temporary mount options.
*/
if (do_readonly)
readonly_changed_cb(zfsvfs, readonly);
if (do_setuid)
setuid_changed_cb(zfsvfs, setuid);
if (do_exec)
exec_changed_cb(zfsvfs, exec);
if (do_xattr)
xattr_changed_cb(zfsvfs, xattr);
if (do_atime)
atime_changed_cb(zfsvfs, atime);
nbmand_changed_cb(zfsvfs, 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, "atime", atime_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs);
(void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb,
zfsvfs);
(void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs);
return (error);
}
/*
* Start a log block write and advance to the next log block.
* Calls are serialized.
*/
static lwb_t *
zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb)
{
lwb_t *nlwb;
zil_trailer_t *ztp = (zil_trailer_t *)(lwb->lwb_buf + lwb->lwb_sz) - 1;
spa_t *spa = zilog->zl_spa;
blkptr_t *bp = &ztp->zit_next_blk;
uint64_t txg;
uint64_t zil_blksz;
int error;
ASSERT(lwb->lwb_nused <= ZIL_BLK_DATA_SZ(lwb));
/*
* Allocate the next block and save its address in this block
* before writing it in order to establish the log chain.
* Note that if the allocation of nlwb synced before we wrote
* the block that points at it (lwb), we'd leak it if we crashed.
* Therefore, we don't do txg_rele_to_sync() until zil_lwb_write_done().
*/
txg = txg_hold_open(zilog->zl_dmu_pool, &lwb->lwb_txgh);
txg_rele_to_quiesce(&lwb->lwb_txgh);
/*
* Pick a ZIL blocksize. We request a size that is the
* maximum of the previous used size, the current used size and
* the amount waiting in the queue.
*/
zil_blksz = MAX(zilog->zl_prev_used,
zilog->zl_cur_used + sizeof (*ztp));
zil_blksz = MAX(zil_blksz, zilog->zl_itx_list_sz + sizeof (*ztp));
zil_blksz = P2ROUNDUP_TYPED(zil_blksz, ZIL_MIN_BLKSZ, uint64_t);
if (zil_blksz > ZIL_MAX_BLKSZ)
zil_blksz = ZIL_MAX_BLKSZ;
BP_ZERO(bp);
/* pass the old blkptr in order to spread log blocks across devs */
error = zio_alloc_blk(spa, zil_blksz, bp, &lwb->lwb_blk, txg);
if (error) {
dmu_tx_t *tx = dmu_tx_create_assigned(zilog->zl_dmu_pool, txg);
/*
* We dirty the dataset to ensure that zil_sync() will
* be called to remove this lwb from our zl_lwb_list.
* Failing to do so, may leave an lwb with a NULL lwb_buf
* hanging around on the zl_lwb_list.
*/
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
dmu_tx_commit(tx);
/*
* Since we've just experienced an allocation failure so we
* terminate the current lwb and send it on its way.
*/
ztp->zit_pad = 0;
ztp->zit_nused = lwb->lwb_nused;
ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum;
zio_nowait(lwb->lwb_zio);
/*
* By returning NULL the caller will call tx_wait_synced()
*/
return (NULL);
}
ASSERT3U(bp->blk_birth, ==, txg);
ztp->zit_pad = 0;
ztp->zit_nused = lwb->lwb_nused;
ztp->zit_bt.zbt_cksum = lwb->lwb_blk.blk_cksum;
bp->blk_cksum = lwb->lwb_blk.blk_cksum;
bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++;
/*
* Allocate a new log write buffer (lwb).
*/
nlwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
nlwb->lwb_zilog = zilog;
nlwb->lwb_blk = *bp;
nlwb->lwb_nused = 0;
nlwb->lwb_sz = BP_GET_LSIZE(&nlwb->lwb_blk);
nlwb->lwb_buf = zio_buf_alloc(nlwb->lwb_sz);
nlwb->lwb_max_txg = txg;
nlwb->lwb_zio = NULL;
/*
* Put new lwb at the end of the log chain
*/
mutex_enter(&zilog->zl_lock);
list_insert_tail(&zilog->zl_lwb_list, nlwb);
mutex_exit(&zilog->zl_lock);
/* Record the block for later vdev flushing */
zil_add_block(zilog, &lwb->lwb_blk);
/*
* kick off the write for the old log block
*/
dprintf_bp(&lwb->lwb_blk, "lwb %p txg %llu: ", lwb, txg);
ASSERT(lwb->lwb_zio);
zio_nowait(lwb->lwb_zio);
return (nlwb);
}
int
zil_claim(char *osname, void *txarg)
{
dmu_tx_t *tx = txarg;
uint64_t first_txg = dmu_tx_get_txg(tx);
zilog_t *zilog;
zil_header_t *zh;
objset_t *os;
int error;
error = dmu_objset_open(osname, DMU_OST_ANY, DS_MODE_USER, &os);
if (error) {
cmn_err(CE_WARN, "can't open objset for %s", osname);
return (0);
}
zilog = dmu_objset_zil(os);
zh = zil_header_in_syncing_context(zilog);
if (zilog->zl_spa->spa_log_state == SPA_LOG_CLEAR) {
if (!BP_IS_HOLE(&zh->zh_log))
zio_free_blk(zilog->zl_spa, &zh->zh_log, first_txg);
BP_ZERO(&zh->zh_log);
dsl_dataset_dirty(dmu_objset_ds(os), tx);
}
/*
* Record here whether the zil has any records to replay.
* If the header block pointer is null or the block points
* to the stubby then we know there are no valid log records.
* We use the header to store this state as the the zilog gets
* freed later in dmu_objset_close().
* The flags (and the rest of the header fields) are cleared in
* zil_sync() as a result of a zil_destroy(), after replaying the log.
*
* Note, the intent log can be empty but still need the
* stubby to be claimed.
*/
if (!zil_empty(zilog)) {
zh->zh_flags |= ZIL_REPLAY_NEEDED;
dsl_dataset_dirty(dmu_objset_ds(os), tx);
}
/*
* Claim all log blocks if we haven't already done so, and remember
* the highest claimed sequence number. This ensures that if we can
* read only part of the log now (e.g. due to a missing device),
* but we can read the entire log later, we will not try to replay
* or destroy beyond the last block we successfully claimed.
*/
ASSERT3U(zh->zh_claim_txg, <=, first_txg);
if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) {
zh->zh_claim_txg = first_txg;
zh->zh_claim_seq = zil_parse(zilog, zil_claim_log_block,
zil_claim_log_record, tx, first_txg);
dsl_dataset_dirty(dmu_objset_ds(os), tx);
}
ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1));
dmu_objset_close(os);
return (0);
}
/*
* Create an on-disk intent log.
*/
static void
zil_create(zilog_t *zilog)
{
const zil_header_t *zh = zilog->zl_header;
lwb_t *lwb;
uint64_t txg = 0;
dmu_tx_t *tx = NULL;
blkptr_t blk;
int error = 0;
/*
* Wait for any previous destroy to complete.
*/
txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg);
ASSERT(zh->zh_claim_txg == 0);
ASSERT(zh->zh_replay_seq == 0);
blk = zh->zh_log;
/*
* If we don't already have an initial log block or we have one
* but it's the wrong endianness then allocate one.
*/
if (BP_IS_HOLE(&blk) || BP_SHOULD_BYTESWAP(&blk)) {
tx = dmu_tx_create(zilog->zl_os);
(void) dmu_tx_assign(tx, TXG_WAIT);
dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx);
txg = dmu_tx_get_txg(tx);
if (!BP_IS_HOLE(&blk)) {
zio_free_blk(zilog->zl_spa, &blk, txg);
BP_ZERO(&blk);
}
error = zio_alloc_blk(zilog->zl_spa, ZIL_MIN_BLKSZ, &blk,
NULL, txg);
if (error == 0)
zil_init_log_chain(zilog, &blk);
}
/*
* Allocate a log write buffer (lwb) for the first log block.
*/
if (error == 0) {
lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP);
lwb->lwb_zilog = zilog;
lwb->lwb_blk = blk;
lwb->lwb_nused = 0;
lwb->lwb_sz = BP_GET_LSIZE(&lwb->lwb_blk);
lwb->lwb_buf = zio_buf_alloc(lwb->lwb_sz);
lwb->lwb_max_txg = txg;
lwb->lwb_zio = NULL;
mutex_enter(&zilog->zl_lock);
list_insert_tail(&zilog->zl_lwb_list, lwb);
mutex_exit(&zilog->zl_lock);
}
/*
* If we just allocated the first log block, commit our transaction
* and wait for zil_sync() to stuff the block poiner into zh_log.
* (zh is part of the MOS, so we cannot modify it in open context.)
*/
if (tx != NULL) {
dmu_tx_commit(tx);
txg_wait_synced(zilog->zl_dmu_pool, txg);
}
ASSERT(bcmp(&blk, &zh->zh_log, sizeof (blk)) == 0);
}
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);
}
