# Copyright (C) 2006, 2013, 2014, 2015 Red Hat, Inc. # Copyright (C) 2006 Daniel P. Berrange # # This work is licensed under the GNU GPLv2 or later. # See the COPYING file in the top-level directory. import os import threading import time import traceback import libvirt import virtinst from virtinst import log from virtinst import pollhelpers from .lib import connectauth from .lib import testmock from .baseclass import vmmGObject from .lib.libvirtenummap import LibvirtEnumMap from .object.domain import vmmDomain from .object.network import vmmNetwork from .object.nodedev import vmmNodeDevice from .object.storagepool import vmmStoragePool from .lib.statsmanager import vmmStatsManager class _ObjectList(vmmGObject): """ Class that wraps our internal list of libvirt objects """ # pylint: disable=not-context-manager # pylint doesn't know that lock() has 'with' support BLACKLIST_COUNT = 3 def __init__(self): vmmGObject.__init__(self) self._objects = [] self._denylist = {} self._lock = threading.Lock() def _cleanup(self): self._objects = [] def _denylist_key(self, obj): return str(obj.__class__) + obj.get_name() def add_denylist(self, obj): """ Add an object to the denylist. Basically a list of objects we choose not to poll, because they threw an error at init time :param obj: vmmLibvirtObject to denylist :returns: number of added object to list """ key = self._denylist_key(obj) count = self._denylist.get(key, 0) self._denylist[key] = count + 1 return self._denylist[key] def remove_denylist(self, obj): """ :param obj: vmmLibvirtObject to remove from denylist :returns: True if object was denylisted or False otherwise. """ key = self._denylist_key(obj) return bool(self._denylist.pop(key, 0)) def in_denylist(self, obj): """ If an object is in list only once don't consider it denylisted, give it one more chance. :param obj: vmmLibvirtObject to check :returns: True if object is denylisted """ key = self._denylist_key(obj) return self._denylist.get(key, 0) >= _ObjectList.BLACKLIST_COUNT def remove(self, obj): """ Remove an object from the list. :param obj: vmmLibvirtObject to remove :returns: True if object removed, False if object was not found """ with self._lock: # Identity check is sufficient here, since we should never be # asked to remove an object that wasn't at one point in the list. if obj not in self._objects: return self.remove_denylist(obj) self._objects.remove(obj) return True def add(self, obj): """ Add an object to the list. :param obj: vmmLibvirtObject to add :returns: True if object added, False if object already in the list """ with self._lock: # We don't look up based on identity here, to prevent tick() # races from adding the same domain twice # # We don't use lookup_object here since we need to hold the # lock the whole time to prevent a 'time of check' issue for checkobj in self._objects: if (checkobj.__class__ == obj.__class__ and checkobj.get_name() == obj.get_name()): return False self._objects.append(obj) return True def get_objects_for_class(self, classobj): """ Return all objects over the passed vmmLibvirtObject class """ with self._lock: return [o for o in self._objects if o.__class__ is classobj] def lookup_object(self, classobj, name): """ Lookup an object with the passed classobj + name """ # Doesn't require locking, since get_objects_for_class covers us for obj in self.get_objects_for_class(classobj): if obj.get_name() == name: return obj return None def all_objects(self): with self._lock: return self._objects[:] class vmmConnection(vmmGObject): __gsignals__ = { "vm-added": (vmmGObject.RUN_FIRST, None, [object]), "vm-removed": (vmmGObject.RUN_FIRST, None, [object]), "net-added": (vmmGObject.RUN_FIRST, None, [object]), "net-removed": (vmmGObject.RUN_FIRST, None, [object]), "pool-added": (vmmGObject.RUN_FIRST, None, [object]), "pool-removed": (vmmGObject.RUN_FIRST, None, [object]), "nodedev-added": (vmmGObject.RUN_FIRST, None, [object]), "nodedev-removed": (vmmGObject.RUN_FIRST, None, [object]), "resources-sampled": (vmmGObject.RUN_FIRST, None, []), "state-changed": (vmmGObject.RUN_FIRST, None, []), "open-completed": (vmmGObject.RUN_FIRST, None, [object]), } (_STATE_DISCONNECTED, _STATE_CONNECTING, _STATE_ACTIVE) = range(1, 4) def __init__(self, uri): self._uri = uri vmmGObject.__init__(self) self._state = self._STATE_DISCONNECTED self._backend = virtinst.VirtinstConnection(self._uri) self._closing = False # Error strings are stored here if open() fails self.connect_error = None self._init_object_count = None self._init_object_event = None self.using_domain_events = False self._domain_cb_ids = [] self.using_network_events = False self._network_cb_ids = [] self.using_storage_pool_events = False self._storage_pool_cb_ids = [] self.using_node_device_events = False self._node_device_cb_ids = [] self._xml_flags = {} self._objects = _ObjectList() self.statsmanager = vmmStatsManager() self._stats = [] self._hostinfo = None self.add_gsettings_handle( self._on_config_pretty_name_changed( self._config_pretty_name_changed_cb)) self._init_virtconn() @staticmethod def pretty_hv(gtype, domtype): """ Convert XML and bar into a more human relevant string. """ gtype = gtype.lower() domtype = domtype.lower() label = domtype if domtype == "xen": if gtype == "xen": label = "xen (paravirt)" elif gtype == "hvm": label = "xen (fullvirt)" elif domtype == "test": if gtype == "xen": label = "test (xen)" elif gtype == "hvm": label = "test (hvm)" elif domtype == "qemu": label = "QEMU TCG" elif domtype == "kvm": label = "KVM" return label def __repr__(self): # pylint: disable=arguments-differ return "<%s uri=%s id=%s>" % ( self.__class__.__name__, self.get_uri(), hex(id(self))) ################# # Init routines # ################# def _wait_for_condition(self, compare_cb, timeout=3): """ Wait for this object to emit the specified signal. Will not block the mainloop. """ from gi.repository import Gtk is_main_thread = (threading.current_thread().name == "MainThread") start_time = time.time() while True: cur_time = time.time() if compare_cb(): return if (cur_time - start_time) >= timeout: return # pragma: no cover if is_main_thread: if Gtk.events_pending(): Gtk.main_iteration_do(False) continue time.sleep(.1) def _init_virtconn(self): self._backend.cb_fetch_all_domains = ( lambda: [obj.get_xmlobj(refresh_if_nec=False) for obj in self.list_vms()]) self._backend.cb_fetch_all_pools = ( lambda: [obj.get_xmlobj(refresh_if_nec=False) for obj in self.list_pools()]) self._backend.cb_fetch_all_nodedevs = ( lambda: [obj.get_xmlobj(refresh_if_nec=False) for obj in self.list_nodedevs()]) def fetch_all_vols(): ret = [] for pool in self.list_pools(): for vol in pool.get_volumes(): try: ret.append(vol.get_xmlobj(refresh_if_nec=False)) except Exception as e: # pragma: no cover log.debug("Fetching volume XML failed: %s", e) return ret self._backend.cb_fetch_all_vols = fetch_all_vols def cache_new_pool(obj): if not self.is_active(): return name = obj.name() self.schedule_priority_tick(pollpool=True) def compare_cb(): return bool(self.get_pool_by_name(name)) self._wait_for_condition(compare_cb) self._backend.cb_cache_new_pool = cache_new_pool ######################## # General data getters # ######################## def get_uri(self): return self._uri def get_backend(self): return self._backend def invalidate_caps(self): return self._backend.invalidate_caps() caps = property(lambda self: getattr(self, "_backend").caps) def host_memory_size(self): if not self._backend.is_open() or self._hostinfo is None: return 0 return self._hostinfo[1] * 1024 def host_active_processor_count(self): if not self._backend.is_open() or self._hostinfo is None: return 0 # pragma: no cover return self._hostinfo[2] ########################## # URI + hostname helpers # ########################## def libvirt_gethostname(self): return self._backend.getHostname() get_uri_hostname = property(lambda s: getattr(s, "_backend").get_uri_hostname) get_uri_username = property(lambda s: getattr(s, "_backend").get_uri_username) get_uri_transport = property(lambda s: getattr(s, "_backend").get_uri_transport) get_uri_port = property(lambda s: getattr(s, "_backend").get_uri_port) get_driver = property(lambda s: getattr(s, "_backend").get_uri_driver) is_container_only = property( lambda s: getattr(s, "_backend").is_container_only) is_lxc = property(lambda s: getattr(s, "_backend").is_lxc) is_vz = property(lambda s: getattr(s, "_backend").is_vz) is_xen = property(lambda s: getattr(s, "_backend").is_xen) is_remote = property(lambda s: getattr(s, "_backend").is_remote) is_qemu = property(lambda s: getattr(s, "_backend").is_qemu) is_qemu_privileged = property(lambda s: getattr(s, "_backend").is_qemu_privileged) is_qemu_unprivileged = property(lambda s: getattr(s, "_backend").is_qemu_unprivileged) is_test = property(lambda s: getattr(s, "_backend").is_test) is_unprivileged = property(lambda s: getattr(s, "_backend").is_unprivileged) def get_cache_dir(self): uri = self.get_uri().replace("/", "_") ret = os.path.join(self._backend.get_app_cache_dir(), uri) if not os.path.exists(ret): os.makedirs(ret, 0o755) # pragma: no cover return ret def get_default_storage_format(self): raw = self.config.get_default_storage_format(raw=True) if raw != "default": return raw # pragma: no cover fmt = self.config.get_default_storage_format() if fmt != "qcow2": return fmt # pragma: no cover if self.support.conn_default_qcow2(): return fmt return None # pragma: no cover #################################### # Connection pretty print routines # #################################### def get_pretty_desc(self): """ Return a pretty label for use in the manager view, and various connection lists. """ if self._get_config_pretty_name(): return self._get_config_pretty_name() pretty_map = { "esx": "ESX", "lxc": "LXC", "openvz": "OpenVZ", "qemu": "QEMU/KVM", "vbox": "Virtualbox", "vmware": "VMWare", "xen": "Xen", } hv = pretty_map.get(self.get_driver(), self.get_driver()) hostname = self.get_uri_hostname() path = self.get_backend().get_uri_path() ret = hv if path == "/session": ret += " " + _("User session") elif (path and path != "/system" and os.path.basename(path)): # Used by test URIs to report what XML file they are using ret += " %s" % os.path.basename(path) if hostname: ret += ": %s" % hostname return ret ####################### # API support helpers # ####################### @property def support(self): return self._backend.support def _get_flags_helper(self, obj, key, check_func): ignore = obj flags_dict = self._xml_flags.get(key) if flags_dict is None: # Flags already set inact, act = check_func() flags_dict = {} flags_dict["active"] = act flags_dict["inactive"] = inact self._xml_flags[key] = flags_dict active_flags = flags_dict["active"] inactive_flags = flags_dict["inactive"] return (inactive_flags, active_flags) def get_dom_flags(self, vm): key = "domain" def check_func(): act = 0 inact = 0 if self.support.domain_xml_inactive(vm): inact = libvirt.VIR_DOMAIN_XML_INACTIVE else: # pragma: no cover log.debug("Domain XML inactive flag not supported.") if self.support.domain_xml_secure(vm): inact |= libvirt.VIR_DOMAIN_XML_SECURE act = libvirt.VIR_DOMAIN_XML_SECURE else: # pragma: no cover log.debug("Domain XML secure flag not supported.") return inact, act return self._get_flags_helper(vm, key, check_func) def get_default_pool(self): poolxml = virtinst.StoragePool.lookup_default_pool(self.get_backend()) if poolxml: for p in self.list_pools(): if p.get_name() == poolxml.name: return p return None def get_vol_by_path(self, path): for pool in self.list_pools(): for vol in pool.get_volumes(): try: if vol.get_target_path() == path: return vol except Exception as e: # pragma: no cover # Errors can happen if the volume disappeared, bug 1092739 log.debug("Error looking up volume from path=%s: %s", path, e) return None ################################### # Connection state getter/setters # ################################### def _change_state(self, newstate): if self._state != newstate: self._state = newstate log.debug("conn=%s changed to state=%s", self.get_uri(), self.get_state_text()) self.emit("state-changed") def is_active(self): return self._state == self._STATE_ACTIVE def is_disconnected(self): return self._state == self._STATE_DISCONNECTED def is_connecting(self): return self._state == self._STATE_CONNECTING def get_state_text(self): if self.is_disconnected(): return _("Disconnected") elif self.is_connecting(): return _("Connecting") return _("Active") ################################# # Libvirt object lookup methods # ################################# def get_vm_by_name(self, name): return self._objects.lookup_object(vmmDomain, name) def list_vms(self): return self._objects.get_objects_for_class(vmmDomain) def get_net_by_name(self, name): return self._objects.lookup_object(vmmNetwork, name) def list_nets(self): return self._objects.get_objects_for_class(vmmNetwork) def get_pool_by_name(self, name): return self._objects.lookup_object(vmmStoragePool, name) def list_pools(self): return self._objects.get_objects_for_class(vmmStoragePool) def get_nodedev_by_name(self, name): return self._objects.lookup_object(vmmNodeDevice, name) def list_nodedevs(self): return self._objects.get_objects_for_class(vmmNodeDevice) ############################ # nodedev helper functions # ############################ def filter_nodedevs(self, devtype): retdevs = [] for dev in self.list_nodedevs(): try: xmlobj = dev.get_xmlobj() except libvirt.libvirtError as e: # pragma: no cover # Libvirt nodedev XML fetching can be busted # https://bugzilla.redhat.com/show_bug.cgi?id=1225771 if e.get_error_code() != libvirt.VIR_ERR_NO_NODE_DEVICE: log.debug("Error fetching nodedev XML", exc_info=True) continue if devtype and xmlobj.device_type != devtype: continue retdevs.append(dev) return retdevs ################################### # Libvirt object creation methods # ################################### def define_domain(self, xml): return self._backend.defineXML(xml) def define_network(self, xml): return self._backend.networkDefineXML(xml) def define_pool(self, xml): return self._backend.storagePoolDefineXML(xml, 0) def rename_object(self, obj, origxml, newxml): if obj.is_domain(): define_cb = self.define_domain elif obj.is_pool(): define_cb = self.define_pool elif obj.is_network(): define_cb = self.define_network else: raise virtinst.xmlutil.DevError("rename_object " "helper doesn't support object class %s" % obj.__class__) # Undefine the original object obj.delete(force=False) newobj = None try: # Redefine new domain newobj = define_cb(newxml) except Exception as renameerr: try: log.debug("Error defining new name %s XML", obj.class_name(), exc_info=True) newobj = define_cb(origxml) except Exception as fixerr: # pragma: no cover log.debug("Failed to redefine original %s!", obj.class_name(), exc_info=True) msg = _("%(object)s rename failed. Attempting to recover also " "failed.\n" "\n" "Original error: %(origerror)s\n" "\n" "Recover error: %(recovererror)s") % { "object": obj.class_name(), "origerror": str(renameerr), "recovererror": str(fixerr), } raise RuntimeError(msg) from None raise finally: if newobj: # Reinsert handle into new obj obj.change_name_backend(newobj) ######################### # Domain event handling # ######################### # Our strategy here isn't the most efficient: since we need to keep the # poll helpers around for compat with old libvirt, switching to a fully # event driven setup is hard, so we end up doing more polling than # necessary on most events. def _domain_xml_misc_event(self, conn, domain, *args): # Just trigger a domain XML refresh for hotplug type events ignore = conn args = list(args) eventstr = args.pop(-1) name = domain.name() log.debug("domain xmlmisc event: domain=%s event=%s args=%s", name, eventstr, args) obj = self.get_vm_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) def _domain_lifecycle_event(self, conn, domain, state, reason, userdata): ignore = conn ignore = userdata name = domain.name() log.debug("domain lifecycle event: domain=%s %s", name, LibvirtEnumMap.domain_lifecycle_str(state, reason)) obj = self.get_vm_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) else: self.schedule_priority_tick(pollvm=True, force=True) def _domain_agent_lifecycle_event(self, conn, domain, state, reason, userdata): ignore = conn ignore = userdata name = domain.name() log.debug("domain agent lifecycle event: domain=%s %s", name, LibvirtEnumMap.domain_agent_lifecycle_str(state, reason)) obj = self.get_vm_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) else: self.schedule_priority_tick(pollvm=True, force=True) # pragma: no cover def _network_lifecycle_event(self, conn, network, state, reason, userdata): ignore = conn ignore = userdata name = network.name() log.debug("network lifecycle event: network=%s %s", name, LibvirtEnumMap.network_lifecycle_str(state, reason)) obj = self.get_net_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) else: self.schedule_priority_tick(pollnet=True, force=True) def _storage_pool_lifecycle_event(self, conn, pool, state, reason, userdata): ignore = conn ignore = userdata name = pool.name() log.debug("storage pool lifecycle event: pool=%s %s", name, LibvirtEnumMap.storage_lifecycle_str(state, reason)) obj = self.get_pool_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) else: self.schedule_priority_tick(pollpool=True, force=True) def _storage_pool_refresh_event(self, conn, pool, userdata): ignore = conn ignore = userdata name = pool.name() log.debug("storage pool refresh event: pool=%s", name) obj = self.get_pool_by_name(name) if not obj: return self.idle_add(obj.refresh_pool_cache_from_event_loop) def _node_device_lifecycle_event(self, conn, dev, state, reason, userdata): ignore = conn ignore = userdata name = dev.name() log.debug("node device lifecycle event: nodedev=%s %s", name, LibvirtEnumMap.nodedev_lifecycle_str(state, reason)) self.schedule_priority_tick(pollnodedev=True, force=True) def _node_device_update_event(self, conn, dev, userdata): ignore = conn ignore = userdata name = dev.name() log.debug("node device update event: nodedev=%s", name) obj = self.get_nodedev_by_name(name) if obj: self.idle_add(obj.recache_from_event_loop) def _add_conn_events(self): if not self.support.conn_working_xen_events(): return # pragma: no cover def _check_events_disabled(): if self.config.CLITestOptions.no_events: raise RuntimeError("events disabled via cli") try: _check_events_disabled() self._domain_cb_ids.append( self.get_backend().domainEventRegisterAny( None, libvirt.VIR_DOMAIN_EVENT_ID_LIFECYCLE, self._domain_lifecycle_event, None)) self.using_domain_events = True log.debug("Using domain events") except Exception as e: self.using_domain_events = False log.debug("Error registering domain events: %s", e) def _add_domain_xml_event(eventname, eventval, cb=None): if not self.using_domain_events: return if not cb: cb = self._domain_xml_misc_event try: eventid = getattr(libvirt, eventname, eventval) self._domain_cb_ids.append( self.get_backend().domainEventRegisterAny( None, eventid, cb, eventname)) if (eventname == "VIR_DOMAIN_EVENT_ID_AGENT_LIFECYCLE" and self.config.CLITestOptions.fake_agent_event): testmock.schedule_fake_agent_event(self, cb) except Exception as e: # pragma: no cover log.debug("Error registering %s event: %s", eventname, e) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_BALLOON_CHANGE", 13) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_TRAY_CHANGE", 10) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_DEVICE_REMOVED", 15) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_DEVICE_ADDED", 19) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_AGENT_LIFECYCLE", 18, self._domain_agent_lifecycle_event) _add_domain_xml_event("VIR_DOMAIN_EVENT_ID_METADATA_CHANGE", 23) try: _check_events_disabled() eventid = getattr(libvirt, "VIR_NETWORK_EVENT_ID_LIFECYCLE", 0) self._network_cb_ids.append( self.get_backend().networkEventRegisterAny( None, eventid, self._network_lifecycle_event, None)) self.using_network_events = True log.debug("Using network events") except Exception as e: self.using_network_events = False log.debug("Error registering network events: %s", e) try: _check_events_disabled() eventid = getattr(libvirt, "VIR_STORAGE_POOL_EVENT_ID_LIFECYCLE", 0) refreshid = getattr(libvirt, "VIR_STORAGE_POOL_EVENT_ID_REFRESH", 1) self._storage_pool_cb_ids.append( self.get_backend().storagePoolEventRegisterAny( None, eventid, self._storage_pool_lifecycle_event, None)) self._storage_pool_cb_ids.append( self.get_backend().storagePoolEventRegisterAny( None, refreshid, self._storage_pool_refresh_event, None)) self.using_storage_pool_events = True log.debug("Using storage pool events") except Exception as e: self.using_storage_pool_events = False log.debug("Error registering storage pool events: %s", e) try: _check_events_disabled() eventid = getattr(libvirt, "VIR_NODE_DEVICE_EVENT_ID_LIFECYCLE", 0) updateid = getattr(libvirt, "VIR_NODE_DEVICE_EVENT_ID_UPDATE", 1) lifecycle_cb = self._node_device_lifecycle_event update_cb = self._node_device_update_event self._node_device_cb_ids.append( self.get_backend().nodeDeviceEventRegisterAny( None, eventid, lifecycle_cb, None)) self._node_device_cb_ids.append( self.get_backend().nodeDeviceEventRegisterAny( None, updateid, update_cb, None)) if self.config.CLITestOptions.fake_nodedev_event: testmock.schedule_fake_nodedev_event(self, lifecycle_cb, update_cb) self.using_node_device_events = True log.debug("Using node device events") except Exception as e: self.using_network_events = False log.debug("Error registering node device events: %s", e) ###################################### # Connection closing/opening methods # ###################################### def _schedule_close(self): self._closing = True self.idle_add(self.close) def close(self): if not self.is_disconnected(): log.debug("conn.close() uri=%s", self.get_uri()) self._closing = True try: if self._backend.is_open(): for eid in self._domain_cb_ids: self._backend.domainEventDeregisterAny(eid) for eid in self._network_cb_ids: self._backend.networkEventDeregisterAny(eid) for eid in self._storage_pool_cb_ids: self._backend.storagePoolEventDeregisterAny(eid) for eid in self._node_device_cb_ids: self._backend.nodeDeviceEventDeregisterAny(eid) except Exception: # pragma: no cover log.debug("Failed to deregister events in conn cleanup", exc_info=True) finally: self._domain_cb_ids = [] self._network_cb_ids = [] self._storage_pool_cb_ids = [] self._node_device_cb_ids = [] self._stats = [] if self._init_object_event: self._init_object_event.clear() # pragma: no cover for obj in self._objects.all_objects(): self._objects.remove(obj) try: self._remove_object_signal(obj) obj.cleanup() except Exception as e: # pragma: no cover log.debug("Failed to cleanup %s: %s", obj, e) self._objects.cleanup() self._objects = _ObjectList() closeret = self._backend.close() if closeret == 1: log.debug( # pragma: no cover "LEAK: conn close() returned 1, " "meaning refs may have leaked.") self._change_state(self._STATE_DISCONNECTED) self._closing = False def _cleanup(self): self.close() self._objects = None self._backend.cb_fetch_all_domains = None self._backend.cb_fetch_all_pools = None self._backend.cb_fetch_all_nodedevs = None self._backend.cb_fetch_all_vols = None self._backend.cb_cache_new_pool = None self.statsmanager.cleanup() self.statsmanager = None def open(self): if not self.is_disconnected(): return # pragma: no cover self._change_state(self._STATE_CONNECTING) log.debug("Scheduling background open thread for %s", self.get_uri()) self._start_thread(self._open_thread, "Connect %s" % self.get_uri()) def _do_open(self): warnconsole = False libvirt_error_code = None libvirt_error_message = None exc = None try: cb = connectauth.creds_dialog data = self if self.config.CLITestOptions.fake_openauth: testmock.fake_openauth(self, cb, data) if self.config.CLITestOptions.fake_session_error: lerr = libvirt.libvirtError("fake session error") lerr.err = [libvirt.VIR_ERR_AUTH_FAILED, None, "fake session error not authorized"] raise lerr self._backend.open(cb, data) return True, None except Exception as e: exc = e tb = "".join(traceback.format_exc()) if isinstance(exc, libvirt.libvirtError): # pylint: disable=no-member libvirt_error_code = exc.get_error_code() libvirt_error_message = exc.get_error_message() if (libvirt_error_code == getattr(libvirt, "VIR_ERR_AUTH_CANCELLED", None)): # pragma: no cover log.debug("User cancelled auth, not raising any error.") return False, None if (libvirt_error_code == libvirt.VIR_ERR_AUTH_FAILED and "not authorized" in libvirt_error_message.lower()): log.debug("Looks like we might have failed policykit " "auth. Checking to see if we have a valid " "console session") if not self.is_remote(): warnconsole = bool(not connectauth.do_we_have_session()) if self.config.CLITestOptions.fake_session_error: warnconsole = True ConnectError = connectauth.connect_error( self, str(exc), tb, warnconsole) return False, ConnectError def _populate_initial_state(self): log.debug("libvirt version=%s", self._backend.local_libvirt_version()) log.debug("daemon version=%s", self._backend.daemon_version()) log.debug("conn version=%s", self._backend.conn_version()) log.debug("%s capabilities:\n%s", self.get_uri(), self.caps.get_xml()) if not self.support.conn_domain(): # pragma: no cover raise RuntimeError("Connection does not support required " "domain listing APIs") if not self.support.conn_storage(): # pragma: no cover log.debug("Connection doesn't seem to support storage APIs.") if not self.support.conn_network(): # pragma: no cover log.debug("Connection doesn't seem to support network APIs.") if not self.support.conn_nodedev(): # pragma: no cover log.debug("Connection doesn't seem to support nodedev APIs.") self._add_conn_events() try: self._backend.setKeepAlive(20, 1) except Exception as e: log.debug("Failed to setKeepAlive: %s", str(e)) # The initial tick will set up a threading event that will only # trigger after all the polled libvirt objects are fully initialized. # That way we only report the connection is open when everything is # nicely setup for the rest of the app. self._init_object_event = threading.Event() self._init_object_count = 0 self.schedule_priority_tick(stats_update=True, pollvm=True, pollnet=True, pollpool=True, pollnodedev=True, force=True, initial_poll=True) self._init_object_event.wait() self._init_object_event = None self._init_object_count = None # Try to create the default storage pool # We need this after events setup so we can determine if the default # pool already exists try: virtinst.StoragePool.build_default_pool(self.get_backend()) except Exception as e: # pragma: no cover log.debug("Building default pool failed: %s", str(e)) def _open_thread(self): try: is_active, ConnectError = self._do_open() if is_active: self._populate_initial_state() except Exception as e: # pragma: no cover is_active = False ConnectError = connectauth.connect_error(self, str(e), "".join(traceback.format_exc()), False) if is_active: self.idle_add(self._change_state, self._STATE_ACTIVE) else: self._schedule_close() self.idle_emit("open-completed", ConnectError) ####################### # Tick/Update methods # ####################### def _remove_object_signal(self, obj): if obj.is_domain(): self.emit("vm-removed", obj) elif obj.is_network(): self.emit("net-removed", obj) elif obj.is_pool(): self.emit("pool-removed", obj) elif obj.is_nodedev(): self.emit("nodedev-removed", obj) def _gone_object_signals(self, gone_objects): """ Responsible for signaling the UI for any updates. All possible UI updates need to go here to enable threading that doesn't block the app with long tick operations. """ if not self._backend.is_open(): return # pragma: no cover for obj in gone_objects: class_name = obj.class_name() name = obj.get_name() if not self._objects.remove(obj): log.debug("Requested removal of %s=%s, but it's " "not in our object list.", class_name, name) continue log.debug("%s=%s removed", class_name, name) self._remove_object_signal(obj) obj.cleanup() def _new_object_cb(self, obj, initialize_failed): if not self._backend.is_open(): return # pragma: no cover try: class_name = obj.class_name() if initialize_failed: log.debug("Blacklisting %s=%s", class_name, obj.get_name()) count = self._objects.add_denylist(obj) log.debug("Object added in denylist, count=%d", count) return self._objects.remove_denylist(obj) if not self._objects.add(obj): log.debug("New %s=%s requested, but it's already tracked.", class_name, obj.get_name()) obj.cleanup() return if not obj.is_nodedev(): # Skip nodedev logging since it's noisy and not interesting log.debug("%s=%s status=%s added", class_name, obj.get_name(), obj.run_status()) if obj.is_domain(): self.emit("vm-added", obj) elif obj.is_network(): self.emit("net-added", obj) elif obj.is_pool(): self.emit("pool-added", obj) elif obj.is_nodedev(): self.emit("nodedev-added", obj) finally: if self._init_object_event: self._init_object_count -= 1 if self._init_object_count <= 0: self._init_object_event.set() def _poll(self, initial_poll, pollvm, pollnet, pollpool, pollnodedev): """ Helper called from tick() to do necessary polling and return the relevant object lists """ gone_objects = [] preexisting_objects = [] def _process_objects(ptype): if ptype == "nets": dopoll = pollnet objs = self.list_nets() cls = vmmNetwork pollcb = pollhelpers.fetch_nets elif ptype == "pools": dopoll = pollpool objs = self.list_pools() cls = vmmStoragePool pollcb = pollhelpers.fetch_pools elif ptype == "nodedevs": dopoll = pollnodedev objs = self.list_nodedevs() cls = vmmNodeDevice pollcb = pollhelpers.fetch_nodedevs else: dopoll = pollvm objs = self.list_vms() cls = vmmDomain pollcb = pollhelpers.fetch_vms keymap = dict((o.get_name(), o) for o in objs) def cb(obj, name): return cls(self, obj, name) if dopoll: gone, new, master = pollcb(self._backend, keymap, cb) else: gone, new, master = [], [], list(keymap.values()) if initial_poll: self._init_object_count += len(new) gone_objects.extend(gone) preexisting_objects.extend([o for o in master if o not in new]) new = [n for n in new if not self._objects.in_denylist(n)] return new new_vms = _process_objects("vms") new_nets = _process_objects("nets") new_pools = _process_objects("pools") new_nodedevs = _process_objects("nodedevs") # Kick off one thread per object type to handle the initial # XML fetching. Going any more fine grained then this probably # won't be that useful due to libvirt's locking structure. # # Would prefer to start refreshing some objects before all polling # is complete, but we need init_object_count to be fully accurate # before we start initializing objects if initial_poll and self._init_object_count == 0: # If the connection doesn't have any objects, new_object_cb # is never called and the event is never set, so let's do it here self._init_object_event.set() for newlist in [new_vms, new_nets, new_pools, new_nodedevs]: if not newlist: continue def cb(lst): for obj in lst: obj.connect_once("initialized", self._new_object_cb) obj.init_libvirt_state() self._start_thread(cb, "refreshing xml for new %s" % newlist[0].class_name(), args=(newlist,)) return gone_objects, preexisting_objects def _tick(self, stats_update=False, pollvm=False, pollnet=False, pollpool=False, pollnodedev=False, force=False, initial_poll=False): """ main update function: polls for new objects, updates stats, ... :param force: Perform the requested polling even if async events are in use. """ if self._closing: return # pragma: no cover if self.is_disconnected(): return # pragma: no cover if self.is_connecting() and not force: return # pragma: no cover # We need to set this before the event check, since stats polling # is independent of events if not pollvm: stats_update = False if self.using_domain_events and not force: pollvm = False if self.using_network_events and not force: pollnet = False if self.using_storage_pool_events and not force: pollpool = False if self.using_node_device_events and not force: pollnodedev = False self._hostinfo = self._backend.getInfo() if stats_update: self.statsmanager.cache_all_stats(self) gone_objects, preexisting_objects = self._poll( initial_poll, pollvm, pollnet, pollpool, pollnodedev) self.idle_add(self._gone_object_signals, gone_objects) # Only tick() pre-existing objects, since new objects will be # initialized asynchronously and tick() would be redundant for obj in preexisting_objects: try: if obj.reports_stats() and stats_update: pass elif obj.is_domain() and not pollvm: continue elif obj.is_network() and not pollnet: continue elif obj.is_pool() and not pollpool: continue elif obj.is_nodedev() and not pollnodedev: continue if self.config.CLITestOptions.conn_crash: self._backend.close() e = libvirt.libvirtError("fake error") e.err = [libvirt.VIR_ERR_SYSTEM_ERROR] raise e obj.tick(stats_update=stats_update) except Exception as e: log.exception("Tick for %s failed", obj) if (isinstance(e, libvirt.libvirtError) and (getattr(e, "get_error_code")() == libvirt.VIR_ERR_SYSTEM_ERROR)): # Try a simple getInfo call to see if conn was dropped self._backend.getInfo() log.debug( # pragma: no cover "vm tick raised system error but " "connection doesn't seem to have dropped. " "Ignoring.") if stats_update: self._recalculate_stats( [o for o in preexisting_objects if o.reports_stats()]) self.idle_emit("resources-sampled") def _recalculate_stats(self, vms): if not self._backend.is_open(): return # pragma: no cover now = time.time() expected = self.config.get_stats_history_length() current = len(self._stats) if current > expected: del self._stats[expected:current] # pragma: no cover mem = 0 cpuTime = 0 rdRate = 0 wrRate = 0 rxRate = 0 txRate = 0 diskMaxRate = self.disk_io_max_rate() or 10.0 netMaxRate = self.network_traffic_max_rate() or 10.0 for vm in vms: if not vm.is_active(): continue cpuTime += vm.cpu_time() mem += vm.stats_memory() rdRate += vm.disk_read_rate() wrRate += vm.disk_write_rate() rxRate += vm.network_rx_rate() txRate += vm.network_tx_rate() netMaxRate = max(netMaxRate, vm.network_traffic_max_rate()) diskMaxRate = max(diskMaxRate, vm.disk_io_max_rate()) pcentHostCpu = 0 pcentMem = mem * 100.0 / self.host_memory_size() if len(self._stats) > 0: prevTimestamp = self._stats[0]["timestamp"] host_cpus = self.host_active_processor_count() pcentHostCpu = ((cpuTime) * 100.0 / ((now - prevTimestamp) * 1000.0 * 1000.0 * 1000.0 * host_cpus)) pcentHostCpu = max(0.0, min(100.0, pcentHostCpu)) pcentMem = max(0.0, min(100.0, pcentMem)) newStats = { "timestamp": now, "memory": mem, "memoryPercent": pcentMem, "cpuTime": cpuTime, "cpuHostPercent": pcentHostCpu, "diskRdRate": rdRate, "diskWrRate": wrRate, "netRxRate": rxRate, "netTxRate": txRate, "diskMaxRate": diskMaxRate, "netMaxRate": netMaxRate, } self._stats.insert(0, newStats) def schedule_priority_tick(self, **kwargs): from .engine import vmmEngine vmmEngine.get_instance().schedule_priority_tick(self, kwargs) def tick_from_engine(self, *args, **kwargs): try: self._tick(*args, **kwargs) except Exception: self._schedule_close() raise ######################## # Stats getter methods # ######################## def _get_record_helper(self, record_name): if len(self._stats) == 0: return 0 return self._stats[0][record_name] def _vector_helper(self, record_name, limit, ceil=100.0): vector = [] statslen = self.config.get_stats_history_length() + 1 if limit is not None: statslen = min(statslen, limit) # pragma: no cover for i in range(statslen): if i < len(self._stats): vector.append(self._stats[i][record_name] / ceil) else: vector.append(0) return vector def stats_memory_vector(self, limit=None): return self._vector_helper("memoryPercent", limit) def host_cpu_time_vector(self, limit=None): return self._vector_helper("cpuHostPercent", limit) def stats_memory(self): return self._get_record_helper("memory") def host_cpu_time_percentage(self): return self._get_record_helper("cpuHostPercent") def guest_cpu_time_percentage(self): return self.host_cpu_time_percentage() def network_traffic_rate(self): return (self._get_record_helper("netRxRate") + self._get_record_helper("netTxRate")) def disk_io_rate(self): return (self._get_record_helper("diskRdRate") + self._get_record_helper("diskWrRate")) def network_traffic_max_rate(self): return self._get_record_helper("netMaxRate") def disk_io_max_rate(self): return self._get_record_helper("diskMaxRate") ########################### # Per-conn config helpers # ########################### def get_autoconnect(self): return self.config.get_conn_autoconnect(self.get_uri()) def set_autoconnect(self, val): self.config.set_conn_autoconnect(self.get_uri(), val) def set_config_pretty_name(self, value): cfgname = self._get_config_pretty_name() if value == cfgname: return # pragma: no cover if not cfgname and value == self.get_pretty_desc(): # Don't encode the default connection value into gconf right # away, require the user to edit it first return self.config.set_perconn(self.get_uri(), "/pretty-name", value) def _get_config_pretty_name(self): return self.config.get_perconn(self.get_uri(), "/pretty-name") def _on_config_pretty_name_changed(self, *args, **kwargs): return self.config.listen_perconn(self.get_uri(), "/pretty-name", *args, **kwargs) def _config_pretty_name_changed_cb(self): self.emit("state-changed") def set_details_window_size(self, w, h): self.config.set_perconn(self.get_uri(), "/window-size", (w, h)) def get_details_window_size(self): ret = self.config.get_perconn(self.get_uri(), "/window-size") return ret