from pipes import quote from bundlewrap.items import Item from bundlewrap.utils.text import mark_for_translation as _ def create(node, path, options): option_list = [] for option, value in sorted(options.items()): # We must exclude the 'mounted' property here because it's a # read-only "informational" property. if option != 'mounted' and value is not None: option_list.append("-o {}={}".format(quote(option), quote(value))) option_args = " ".join(option_list) node.run( "zfs create {} {}".format( option_args, quote(path), ), may_fail=True, ) if options['mounted'] == 'no': set_option(node, path, 'mounted', 'no') def does_exist(node, path): status_result = node.run( "zfs list {}".format(quote(path)), may_fail=True, ) return status_result.return_code == 0 def get_option(node, path, option): cmd = "zfs get -Hp -o value {} {}".format(quote(option), quote(path)) # We always expect this to succeed since we don't call this function # if we have already established that the dataset does not exist. status_result = node.run(cmd) return status_result.stdout.decode('utf-8').strip() def set_option(node, path, option, value): if option == 'mounted': # 'mounted' is a read-only property that can not be altered by # 'set'. We need to call 'zfs mount tank/foo'. node.run( "zfs {} {}".format( "mount" if value == 'yes' else "unmount", quote(path), ), may_fail=True, ) else: node.run( "zfs set {}={} {}".format( quote(option), quote(value), quote(path), ), may_fail=True, ) class ZFSDataset(Item): """ Creates ZFS datasets and manages their options. """ BUNDLE_ATTRIBUTE_NAME = "zfs_datasets" ITEM_ATTRIBUTES = { 'atime': None, 'acltype': None, 'compression': None, 'mountpoint': None, 'quota': None, 'recordsize': None, 'dedup': None, } ITEM_TYPE_NAME = "zfs_dataset" def __repr__(self): return f"" def cdict(self): cdict = {} for option, value in self.attributes.items(): if option == 'mountpoint' and value is None: value = "none" if value is not None: cdict[option] = value cdict['mounted'] = 'no' if cdict.get('mountpoint') in (None, "none") else 'yes' return cdict def fix(self, status): if status.must_be_created: create(self.node, self.name, status.cdict) else: for option in status.keys_to_fix: set_option(self.node, self.name, option, status.cdict[option]) def get_auto_deps(self, items): pool = self.name.split("/")[0] pool_item = "zfs_pool:{}".format(pool) pool_item_found = False for item in items: if item.ITEM_TYPE_NAME == "zfs_pool" and item.name == pool: # Add dependency to the pool this dataset resides on. pool_item_found = True yield pool_item elif ( item.ITEM_TYPE_NAME == "zfs_dataset" and self.name != item.name and self.name.startswith(item.name + "/") ): # Find all other datasets that are parents of this # dataset. # XXX Could be optimized by finding the "largest" # parent only. yield item.id if not pool_item_found: raise Exception(_( "ZFS dataset {dataset} resides on pool {pool} but item " "{dep} does not exist" ).format( dataset=self.name, pool=pool, dep=pool_item, )) def sdict(self): if not does_exist(self.node, self.name): return None sdict = {} for option, value in self.attributes.items(): sdict[option] = get_option(self.node, self.name, option) sdict['mounted'] = get_option(self.node, self.name, 'mounted') return sdict