/*
* Safely mark an inode dirty. Inodes which are part of a read-only
* file system or snapshot may not be dirtied.
*/
void
zfs_mark_inode_dirty(struct inode *ip)
{
zfs_sb_t *zsb = ITOZSB(ip);
if (zfs_is_readonly(zsb) || dmu_objset_is_snapshot(zsb->z_os))
return;
mark_inode_dirty(ip);
}
/*
* Return a znode for the extended attribute directory for zp.
* ** If the directory does not already exist, it is created **
*
* IN: zp - znode to obtain attribute directory from
* cr - credentials of caller
* flags - flags from the VOP_LOOKUP call
*
* OUT: xipp - pointer to extended attribute znode
*
* RETURN: 0 on success
* error number on failure
*/
int
zfs_get_xattrdir(znode_t *zp, struct inode **xipp, cred_t *cr, int flags)
{
zfsvfs_t *zfsvfs = ZTOZSB(zp);
znode_t *xzp;
zfs_dirlock_t *dl;
vattr_t va;
int error;
top:
error = zfs_dirent_lock(&dl, zp, "", &xzp, ZXATTR, NULL, NULL);
if (error)
return (error);
if (xzp != NULL) {
*xipp = ZTOI(xzp);
zfs_dirent_unlock(dl);
return (0);
}
if (!(flags & CREATE_XATTR_DIR)) {
zfs_dirent_unlock(dl);
return (SET_ERROR(ENOENT));
}
if (zfs_is_readonly(zfsvfs)) {
zfs_dirent_unlock(dl);
return (SET_ERROR(EROFS));
}
/*
* The ability to 'create' files in an attribute
* directory comes from the write_xattr permission on the base file.
*
* The ability to 'search' an attribute directory requires
* read_xattr permission on the base file.
*
* Once in a directory the ability to read/write attributes
* is controlled by the permissions on the attribute file.
*/
va.va_mask = ATTR_MODE | ATTR_UID | ATTR_GID;
va.va_mode = S_IFDIR | S_ISVTX | 0777;
zfs_fuid_map_ids(zp, cr, &va.va_uid, &va.va_gid);
va.va_dentry = NULL;
error = zfs_make_xattrdir(zp, &va, xipp, cr);
zfs_dirent_unlock(dl);
if (error == ERESTART) {
/* NB: we already did dmu_tx_wait() if necessary */
goto top;
}
return (error);
}
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);
}
/*
* 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);
}
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);
}
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);
}
