int
zfs_sb_create(const char *osname, zfs_mntopts_t *zmo, zfs_sb_t **zsbp)
{
objset_t *os;
zfs_sb_t *zsb;
uint64_t zval;
int i, error;
uint64_t sa_obj;
zsb = kmem_zalloc(sizeof (zfs_sb_t), KM_SLEEP);
/*
* We claim to always be readonly so we can open snapshots;
* other ZPL code will prevent us from writing to snapshots.
*/
error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zsb, &os);
if (error) {
kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
/*
* Optional temporary mount options, free'd in zfs_sb_free().
*/
zsb->z_mntopts = (zmo ? zmo : zfs_mntopts_alloc());
/*
* Initialize the zfs-specific filesystem structure.
* Should probably make this a kmem cache, shuffle fields,
* and just bzero up to z_hold_mtx[].
*/
zsb->z_sb = NULL;
zsb->z_parent = zsb;
zsb->z_max_blksz = SPA_OLD_MAXBLOCKSIZE;
zsb->z_show_ctldir = ZFS_SNAPDIR_VISIBLE;
zsb->z_os = os;
error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zsb->z_version);
if (error) {
goto out;
} else if (zsb->z_version > ZPL_VERSION) {
error = SET_ERROR(ENOTSUP);
goto out;
}
if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
goto out;
zsb->z_norm = (int)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
goto out;
zsb->z_utf8 = (zval != 0);
if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
goto out;
zsb->z_case = (uint_t)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_ACLTYPE, &zval)) != 0)
goto out;
zsb->z_acl_type = (uint_t)zval;
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
*/
if (zsb->z_case == ZFS_CASE_INSENSITIVE ||
zsb->z_case == ZFS_CASE_MIXED)
zsb->z_norm |= U8_TEXTPREP_TOUPPER;
zsb->z_use_fuids = USE_FUIDS(zsb->z_version, zsb->z_os);
zsb->z_use_sa = USE_SA(zsb->z_version, zsb->z_os);
if (zsb->z_use_sa) {
/* should either have both of these objects or none */
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1,
&sa_obj);
if (error)
goto out;
error = zfs_get_zplprop(os, ZFS_PROP_XATTR, &zval);
if ((error == 0) && (zval == ZFS_XATTR_SA))
zsb->z_xattr_sa = B_TRUE;
} else {
/*
* Pre SA versions file systems should never touch
* either the attribute registration or layout objects.
*/
sa_obj = 0;
}
error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END,
&zsb->z_attr_table);
if (error)
goto out;
if (zsb->z_version >= ZPL_VERSION_SA)
sa_register_update_callback(os, zfs_sa_upgrade);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
&zsb->z_root);
if (error)
goto out;
ASSERT(zsb->z_root != 0);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
&zsb->z_unlinkedobj);
if (error)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA],
8, 1, &zsb->z_userquota_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ,
zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA],
8, 1, &zsb->z_groupquota_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
&zsb->z_fuid_obj);
if (error && error != ENOENT)
goto out;
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1,
&zsb->z_shares_dir);
if (error && error != ENOENT)
goto out;
mutex_init(&zsb->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&zsb->z_lock, NULL, MUTEX_DEFAULT, NULL);
list_create(&zsb->z_all_znodes, sizeof (znode_t),
offsetof(znode_t, z_link_node));
rrm_init(&zsb->z_teardown_lock, B_FALSE);
rw_init(&zsb->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL);
rw_init(&zsb->z_fuid_lock, NULL, RW_DEFAULT, NULL);
zsb->z_hold_mtx = vmem_zalloc(sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ,
KM_SLEEP);
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_init(&zsb->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
*zsbp = zsb;
return (0);
out:
dmu_objset_disown(os, zsb);
*zsbp = NULL;
vmem_free(zsb->z_hold_mtx, sizeof (kmutex_t) * ZFS_OBJ_MTX_SZ);
kmem_free(zsb, sizeof (zfs_sb_t));
return (error);
}
/*
* zfs_init_fs - Initialize the zfsvfs struct and the file system
* incore "master" object. Verify version compatibility.
*/
int
zfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp)
{
objset_t *os = zfsvfs->z_os;
int i, error;
uint64_t fsid_guid;
uint64_t zval;
*zpp = NULL;
error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
if (error) {
return (error);
} else if (zfsvfs->z_version > ZPL_VERSION) {
(void) printf("Mismatched versions: File system "
"is version %llu on-disk format, which is "
"incompatible with this software version %lld!",
(u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
return (ENOTSUP);
}
if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
return (error);
zfsvfs->z_norm = (int)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
return (error);
zfsvfs->z_utf8 = (zval != 0);
if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0)
return (error);
zfsvfs->z_case = (uint_t)zval;
/*
* Fold case on file systems that are always or sometimes case
* insensitive.
*/
if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE ||
zfsvfs->z_case == ZFS_CASE_MIXED)
zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER;
/*
* The fsid is 64 bits, composed of an 8-bit fs type, which
* separates our fsid from any other filesystem types, and a
* 56-bit objset unique ID. The objset unique ID is unique to
* all objsets open on this system, provided by unique_create().
* The 8-bit fs type must be put in the low bits of fsid[1]
* because that's where other Solaris filesystems put it.
*/
fsid_guid = dmu_objset_fsid_guid(os);
ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0);
zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid;
zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) |
zfsvfs->z_vfs->mnt_vfc->vfc_typenum & 0xFF;
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1,
&zfsvfs->z_root);
if (error)
return (error);
ASSERT(zfsvfs->z_root != 0);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1,
&zfsvfs->z_unlinkedobj);
if (error)
return (error);
/*
* Initialize zget mutex's
*/
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL);
error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp);
if (error) {
/*
* On error, we destroy the mutexes here since it's not
* possible for the caller to determine if the mutexes were
* initialized properly.
*/
for (i = 0; i != ZFS_OBJ_MTX_SZ; i++)
mutex_destroy(&zfsvfs->z_hold_mtx[i]);
return (error);
}
ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root);
error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1,
&zfsvfs->z_fuid_obj);
if (error == ENOENT)
error = 0;
return (0);
}
/*
* Reopen zfs_sb_t and release VFS ops.
*/
int
zfs_resume_fs(zfs_sb_t *zsb, const char *osname)
{
int err, err2;
znode_t *zp;
uint64_t sa_obj = 0;
ASSERT(RRM_WRITE_HELD(&zsb->z_teardown_lock));
ASSERT(RW_WRITE_HELD(&zsb->z_teardown_inactive_lock));
/*
* We already own this, so just hold and rele it to update the
* objset_t, as the one we had before may have been evicted.
*/
VERIFY0(dmu_objset_hold(osname, zsb, &zsb->z_os));
VERIFY3P(zsb->z_os->os_dsl_dataset->ds_owner, ==, zsb);
VERIFY(dsl_dataset_long_held(zsb->z_os->os_dsl_dataset));
dmu_objset_rele(zsb->z_os, zsb);
/*
* Make sure version hasn't changed
*/
err = zfs_get_zplprop(zsb->z_os, ZFS_PROP_VERSION,
&zsb->z_version);
if (err)
goto bail;
err = zap_lookup(zsb->z_os, MASTER_NODE_OBJ,
ZFS_SA_ATTRS, 8, 1, &sa_obj);
if (err && zsb->z_version >= ZPL_VERSION_SA)
goto bail;
if ((err = sa_setup(zsb->z_os, sa_obj,
zfs_attr_table, ZPL_END, &zsb->z_attr_table)) != 0)
goto bail;
if (zsb->z_version >= ZPL_VERSION_SA)
sa_register_update_callback(zsb->z_os,
zfs_sa_upgrade);
VERIFY(zfs_sb_setup(zsb, B_FALSE) == 0);
zfs_set_fuid_feature(zsb);
zsb->z_rollback_time = jiffies;
/*
* Attempt to re-establish all the active inodes with their
* dbufs. If a zfs_rezget() fails, then we unhash the inode
* and mark it stale. This prevents a collision if a new
* inode/object is created which must use the same inode
* number. The stale inode will be be released when the
* VFS prunes the dentry holding the remaining references
* on the stale inode.
*/
mutex_enter(&zsb->z_znodes_lock);
for (zp = list_head(&zsb->z_all_znodes); zp;
zp = list_next(&zsb->z_all_znodes, zp)) {
err2 = zfs_rezget(zp);
if (err2) {
remove_inode_hash(ZTOI(zp));
zp->z_is_stale = B_TRUE;
}
}
mutex_exit(&zsb->z_znodes_lock);
bail:
/* release the VFS ops */
rw_exit(&zsb->z_teardown_inactive_lock);
rrm_exit(&zsb->z_teardown_lock, FTAG);
if (err) {
/*
* Since we couldn't setup the sa framework, try to force
* unmount this file system.
*/
if (zsb->z_os)
(void) zfs_umount(zsb->z_sb);
}
return (err);
}
error = dmu_tx_assign(tx, TXG_WAIT);
ASSERT3U(error, ==, 0);
if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID)
zpl_version = ZPL_VERSION;
else
zpl_version = ZPL_VERSION_FUID - 1;
VERIFY(nvlist_alloc(&zprops, NV_UNIQUE_NAME, KM_SLEEP) == 0);
VERIFY(nvlist_add_uint64(zprops,
zfs_prop_to_name(ZFS_PROP_VERSION), zpl_version) == 0);
zfs_create_fs(os, cr, zprops, tx);
nvlist_free(zprops);
dmu_tx_commit(tx);
}
error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version);
if (error) {
return (error);
} else if (zfsvfs->z_version > ZPL_VERSION) {
(void) printf("Mismatched versions: File system "
"is version %llu on-disk format, which is "
"incompatible with this software version %lld!",
(u_longlong_t)zfsvfs->z_version, ZPL_VERSION);
return (ENOTSUP);
}
if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0)
return (error);
zfsvfs->z_norm = (int)zval;
if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0)
return (error);
