# Copyright (C) 2013 Canonical Ltd. # # Author: Scott Moser # # This file is part of cloud-init. See LICENSE file for license information. import base64 import crypt import os import os.path import re import xml.etree.ElementTree as ET from collections import namedtuple from enum import Enum from functools import partial from time import sleep, time from typing import Optional from xml.dom import minidom import requests from cloudinit import dmi from cloudinit import log as logging from cloudinit import net, sources, ssh_util, subp, util from cloudinit.event import EventScope, EventType from cloudinit.net import device_driver from cloudinit.reporting import events from cloudinit.sources.helpers import netlink from cloudinit.sources.helpers.azure import ( DEFAULT_REPORT_FAILURE_USER_VISIBLE_MESSAGE, EphemeralDHCPv4WithReporting, azure_ds_reporter, azure_ds_telemetry_reporter, build_minimal_ovf, get_boot_telemetry, get_metadata_from_fabric, get_system_info, is_byte_swapped, push_log_to_kvp, report_diagnostic_event, report_failure_to_fabric, ) from cloudinit.url_helper import UrlError, readurl, retry_on_url_exc LOG = logging.getLogger(__name__) DS_NAME = "Azure" DEFAULT_METADATA = {"instance-id": "iid-AZURE-NODE"} # azure systems will always have a resource disk, and 66-azure-ephemeral.rules # ensures that it gets linked to this path. RESOURCE_DISK_PATH = "/dev/disk/cloud/azure_resource" LEASE_FILE = "/var/lib/dhcp/dhclient.eth0.leases" DEFAULT_FS = "ext4" # DMI chassis-asset-tag is set static for all azure instances AZURE_CHASSIS_ASSET_TAG = "7783-7084-3265-9085-8269-3286-77" REPROVISION_MARKER_FILE = "/var/lib/cloud/data/poll_imds" REPROVISION_NIC_DETACHED_MARKER_FILE = "/var/lib/cloud/data/nic_detached" REPORTED_READY_MARKER_FILE = "/var/lib/cloud/data/reported_ready" AGENT_SEED_DIR = "/var/lib/waagent" DEFAULT_PROVISIONING_ISO_DEV = "/dev/sr0" # In the event where the IMDS primary server is not # available, it takes 1s to fallback to the secondary one IMDS_TIMEOUT_IN_SECONDS = 2 IMDS_URL = "http://169.254.169.254/metadata" IMDS_VER_MIN = "2019-06-01" IMDS_VER_WANT = "2021-08-01" IMDS_EXTENDED_VER_MIN = "2021-03-01" # This holds SSH key data including if the source was # from IMDS, as well as the SSH key data itself. SSHKeys = namedtuple("SSHKeys", ("keys_from_imds", "ssh_keys")) class MetadataType(Enum): ALL = "{}/instance".format(IMDS_URL) NETWORK = "{}/instance/network".format(IMDS_URL) REPROVISION_DATA = "{}/reprovisiondata".format(IMDS_URL) class PPSType(Enum): NONE = "None" RUNNING = "Running" SAVABLE = "Savable" UNKNOWN = "Unknown" PLATFORM_ENTROPY_SOURCE = "/sys/firmware/acpi/tables/OEM0" # List of static scripts and network config artifacts created by # stock ubuntu suported images. UBUNTU_EXTENDED_NETWORK_SCRIPTS = [ "/etc/netplan/90-hotplug-azure.yaml", "/usr/local/sbin/ephemeral_eth.sh", "/etc/udev/rules.d/10-net-device-added.rules", "/run/network/interfaces.ephemeral.d", ] # This list is used to blacklist devices that will be considered # for renaming or fallback interfaces. # # On Azure network devices using these drivers are automatically # configured by the platform and should not be configured by # cloud-init's network configuration. # # Note: # Azure Dv4 and Ev4 series VMs always have mlx5 hardware. # https://docs.microsoft.com/en-us/azure/virtual-machines/dv4-dsv4-series # https://docs.microsoft.com/en-us/azure/virtual-machines/ev4-esv4-series # Earlier D and E series VMs (such as Dv2, Dv3, and Ev3 series VMs) # can have either mlx4 or mlx5 hardware, with the older series VMs # having a higher chance of coming with mlx4 hardware. # https://docs.microsoft.com/en-us/azure/virtual-machines/dv2-dsv2-series # https://docs.microsoft.com/en-us/azure/virtual-machines/dv3-dsv3-series # https://docs.microsoft.com/en-us/azure/virtual-machines/ev3-esv3-series BLACKLIST_DRIVERS = ["mlx4_core", "mlx5_core"] def find_storvscid_from_sysctl_pnpinfo(sysctl_out, deviceid): # extract the 'X' from dev.storvsc.X. if deviceid matches """ dev.storvsc.1.%pnpinfo: classid=32412632-86cb-44a2-9b5c-50d1417354f5 deviceid=00000000-0001-8899-0000-000000000000 """ for line in sysctl_out.splitlines(): if re.search(r"pnpinfo", line): fields = line.split() if len(fields) >= 3: columns = fields[2].split("=") if ( len(columns) >= 2 and columns[0] == "deviceid" and columns[1].startswith(deviceid) ): comps = fields[0].split(".") return comps[2] return None def find_busdev_from_disk(camcontrol_out, disk_drv): # find the scbusX from 'camcontrol devlist -b' output # if disk_drv matches the specified disk driver, i.e. blkvsc1 """ scbus0 on ata0 bus 0 scbus1 on ata1 bus 0 scbus2 on blkvsc0 bus 0 scbus3 on blkvsc1 bus 0 scbus4 on storvsc2 bus 0 scbus5 on storvsc3 bus 0 scbus-1 on xpt0 bus 0 """ for line in camcontrol_out.splitlines(): if re.search(disk_drv, line): items = line.split() return items[0] return None def find_dev_from_busdev(camcontrol_out, busdev): # find the daX from 'camcontrol devlist' output # if busdev matches the specified value, i.e. 'scbus2' """ at scbus1 target 0 lun 0 (cd0,pass0) at scbus2 target 0 lun 0 (da0,pass1) at scbus3 target 1 lun 0 (da1,pass2) """ for line in camcontrol_out.splitlines(): if re.search(busdev, line): items = line.split("(") if len(items) == 2: dev_pass = items[1].split(",") return dev_pass[0] return None def execute_or_debug(cmd, fail_ret=None): try: return subp.subp(cmd)[0] except subp.ProcessExecutionError: LOG.debug("Failed to execute: %s", " ".join(cmd)) return fail_ret def get_dev_storvsc_sysctl(): return execute_or_debug(["sysctl", "dev.storvsc"], fail_ret="") def get_camcontrol_dev_bus(): return execute_or_debug(["camcontrol", "devlist", "-b"]) def get_camcontrol_dev(): return execute_or_debug(["camcontrol", "devlist"]) def get_resource_disk_on_freebsd(port_id): g0 = "00000000" if port_id > 1: g0 = "00000001" port_id = port_id - 2 g1 = "000" + str(port_id) g0g1 = "{0}-{1}".format(g0, g1) # search 'X' from # 'dev.storvsc.X.%pnpinfo: # classid=32412632-86cb-44a2-9b5c-50d1417354f5 # deviceid=00000000-0001-8899-0000-000000000000' sysctl_out = get_dev_storvsc_sysctl() storvscid = find_storvscid_from_sysctl_pnpinfo(sysctl_out, g0g1) if not storvscid: LOG.debug("Fail to find storvsc id from sysctl") return None camcontrol_b_out = get_camcontrol_dev_bus() camcontrol_out = get_camcontrol_dev() # try to find /dev/XX from 'blkvsc' device blkvsc = "blkvsc{0}".format(storvscid) scbusx = find_busdev_from_disk(camcontrol_b_out, blkvsc) if scbusx: devname = find_dev_from_busdev(camcontrol_out, scbusx) if devname is None: LOG.debug("Fail to find /dev/daX") return None return devname # try to find /dev/XX from 'storvsc' device storvsc = "storvsc{0}".format(storvscid) scbusx = find_busdev_from_disk(camcontrol_b_out, storvsc) if scbusx: devname = find_dev_from_busdev(camcontrol_out, scbusx) if devname is None: LOG.debug("Fail to find /dev/daX") return None return devname return None # update the FreeBSD specific information if util.is_FreeBSD(): LEASE_FILE = "/var/db/dhclient.leases.hn0" DEFAULT_FS = "freebsd-ufs" res_disk = get_resource_disk_on_freebsd(1) if res_disk is not None: LOG.debug("resource disk is not None") RESOURCE_DISK_PATH = "/dev/" + res_disk else: LOG.debug("resource disk is None") # TODO Find where platform entropy data is surfaced PLATFORM_ENTROPY_SOURCE = None BUILTIN_DS_CONFIG = { "data_dir": AGENT_SEED_DIR, "disk_aliases": {"ephemeral0": RESOURCE_DISK_PATH}, "dhclient_lease_file": LEASE_FILE, "apply_network_config": True, # Use IMDS published network configuration } # RELEASE_BLOCKER: Xenial and earlier apply_network_config default is False BUILTIN_CLOUD_EPHEMERAL_DISK_CONFIG = { "disk_setup": { "ephemeral0": { "table_type": "gpt", "layout": [100], "overwrite": True, }, }, "fs_setup": [{"filesystem": DEFAULT_FS, "device": "ephemeral0.1"}], } DS_CFG_PATH = ["datasource", DS_NAME] DS_CFG_KEY_PRESERVE_NTFS = "never_destroy_ntfs" DEF_EPHEMERAL_LABEL = "Temporary Storage" # The redacted password fails to meet password complexity requirements # so we can safely use this to mask/redact the password in the ovf-env.xml DEF_PASSWD_REDACTION = "REDACTED" class DataSourceAzure(sources.DataSource): dsname = "Azure" default_update_events = { EventScope.NETWORK: { EventType.BOOT_NEW_INSTANCE, EventType.BOOT, } } _negotiated = False _metadata_imds = sources.UNSET _ci_pkl_version = 1 def __init__(self, sys_cfg, distro, paths): sources.DataSource.__init__(self, sys_cfg, distro, paths) self.seed_dir = os.path.join(paths.seed_dir, "azure") self.cfg = {} self.seed = None self.ds_cfg = util.mergemanydict( [util.get_cfg_by_path(sys_cfg, DS_CFG_PATH, {}), BUILTIN_DS_CONFIG] ) self.dhclient_lease_file = self.ds_cfg.get("dhclient_lease_file") self._network_config = None self._ephemeral_dhcp_ctx = None self.iso_dev = None def _unpickle(self, ci_pkl_version: int) -> None: super()._unpickle(ci_pkl_version) self._ephemeral_dhcp_ctx = None if not hasattr(self, "iso_dev"): self.iso_dev = None def __str__(self): root = sources.DataSource.__str__(self) return "%s [seed=%s]" % (root, self.seed) def _get_subplatform(self): """Return the subplatform metadata source details.""" if self.seed.startswith("/dev"): subplatform_type = "config-disk" elif self.seed.lower() == "imds": subplatform_type = "imds" else: subplatform_type = "seed-dir" return "%s (%s)" % (subplatform_type, self.seed) @azure_ds_telemetry_reporter def crawl_metadata(self): """Walk all instance metadata sources returning a dict on success. @return: A dictionary of any metadata content for this instance. @raise: InvalidMetaDataException when the expected metadata service is unavailable, broken or disabled. """ crawled_data = {} # azure removes/ejects the cdrom containing the ovf-env.xml # file on reboot. So, in order to successfully reboot we # need to look in the datadir and consider that valid ddir = self.ds_cfg["data_dir"] # The order in which the candidates are inserted matters here, because # it determines the value of ret. More specifically, the first one in # the candidate list determines the path to take in order to get the # metadata we need. ovf_is_accessible = False metadata_source = None md = {} userdata_raw = "" cfg = {} files = {} if os.path.isfile(REPROVISION_MARKER_FILE): metadata_source = "IMDS" report_diagnostic_event( "Reprovision marker file already present " "before crawling Azure metadata: %s" % REPROVISION_MARKER_FILE, logger_func=LOG.debug, ) else: for src in list_possible_azure_ds(self.seed_dir, ddir): try: if src.startswith("/dev/"): if util.is_FreeBSD(): md, userdata_raw, cfg, files = util.mount_cb( src, load_azure_ds_dir, mtype="udf" ) else: md, userdata_raw, cfg, files = util.mount_cb( src, load_azure_ds_dir ) # save the device for ejection later self.iso_dev = src else: md, userdata_raw, cfg, files = load_azure_ds_dir(src) ovf_is_accessible = True metadata_source = src break except NonAzureDataSource: report_diagnostic_event( "Did not find Azure data source in %s" % src, logger_func=LOG.debug, ) continue except util.MountFailedError: report_diagnostic_event( "%s was not mountable" % src, logger_func=LOG.debug ) md = {"local-hostname": ""} cfg = {"system_info": {"default_user": {"name": ""}}} metadata_source = "IMDS" continue except BrokenAzureDataSource as exc: msg = "BrokenAzureDataSource: %s" % exc report_diagnostic_event(msg, logger_func=LOG.error) raise sources.InvalidMetaDataException(msg) report_diagnostic_event( "Found provisioning metadata in %s" % metadata_source, logger_func=LOG.debug, ) imds_md = self.get_imds_data_with_api_fallback( self.fallback_interface, retries=10 ) # reset _fallback_interface so that if the code enters reprovisioning # flow, it will force re-evaluation of new fallback nic. self._fallback_interface = None if not imds_md and not ovf_is_accessible: msg = "No OVF or IMDS available" report_diagnostic_event(msg) raise sources.InvalidMetaDataException(msg) pps_type = self._determine_pps_type(cfg, imds_md) if pps_type != PPSType.NONE: if util.is_FreeBSD(): msg = "Free BSD is not supported for PPS VMs" report_diagnostic_event(msg, logger_func=LOG.error) raise sources.InvalidMetaDataException(msg) self._write_reprovision_marker() if pps_type == PPSType.SAVABLE: self._wait_for_all_nics_ready() md, userdata_raw, cfg, files = self._reprovision() # fetch metadata again as it has changed after reprovisioning imds_md = self.get_imds_data_with_api_fallback( self.fallback_interface, retries=10 ) self.seed = metadata_source crawled_data.update( { "cfg": cfg, "files": files, "metadata": util.mergemanydict([md, {"imds": imds_md}]), "userdata_raw": userdata_raw, } ) imds_username = _username_from_imds(imds_md) imds_hostname = _hostname_from_imds(imds_md) imds_disable_password = _disable_password_from_imds(imds_md) if imds_username: LOG.debug("Username retrieved from IMDS: %s", imds_username) cfg["system_info"]["default_user"]["name"] = imds_username if imds_hostname: LOG.debug("Hostname retrieved from IMDS: %s", imds_hostname) crawled_data["metadata"]["local-hostname"] = imds_hostname if imds_disable_password: LOG.debug( "Disable password retrieved from IMDS: %s", imds_disable_password, ) crawled_data["metadata"][ "disable_password" ] = imds_disable_password if metadata_source == "IMDS" and not crawled_data["files"]: try: contents = build_minimal_ovf( username=imds_username, hostname=imds_hostname, disableSshPwd=imds_disable_password, ) crawled_data["files"] = {"ovf-env.xml": contents} except Exception as e: report_diagnostic_event( "Failed to construct OVF from IMDS data %s" % e, logger_func=LOG.debug, ) # only use userdata from imds if OVF did not provide custom data # userdata provided by IMDS is always base64 encoded if not userdata_raw: imds_userdata = _userdata_from_imds(imds_md) if imds_userdata: LOG.debug("Retrieved userdata from IMDS") try: crawled_data["userdata_raw"] = base64.b64decode( "".join(imds_userdata.split()) ) except Exception: report_diagnostic_event( "Bad userdata in IMDS", logger_func=LOG.warning ) if not metadata_source: msg = "No Azure metadata found" report_diagnostic_event(msg, logger_func=LOG.error) raise sources.InvalidMetaDataException(msg) else: report_diagnostic_event( "found datasource in %s" % metadata_source, logger_func=LOG.debug, ) if metadata_source == ddir: report_diagnostic_event( "using files cached in %s" % ddir, logger_func=LOG.debug ) seed = _get_random_seed() if seed: crawled_data["metadata"]["random_seed"] = seed crawled_data["metadata"]["instance-id"] = self._iid() if pps_type != PPSType.NONE: LOG.info("Reporting ready to Azure after getting ReprovisionData") use_cached_ephemeral = ( self.distro.networking.is_up(self.fallback_interface) and self._ephemeral_dhcp_ctx and self._ephemeral_dhcp_ctx.lease ) if use_cached_ephemeral: self._report_ready(lease=self._ephemeral_dhcp_ctx.lease) self._ephemeral_dhcp_ctx.clean_network() # Teardown ephemeral else: try: with EphemeralDHCPv4WithReporting( azure_ds_reporter ) as lease: self._report_ready(lease=lease) except Exception as e: report_diagnostic_event( "exception while reporting ready: %s" % e, logger_func=LOG.error, ) raise return crawled_data def _is_platform_viable(self): """Check platform environment to report if this datasource may run.""" return _is_platform_viable(self.seed_dir) def clear_cached_attrs(self, attr_defaults=()): """Reset any cached class attributes to defaults.""" super(DataSourceAzure, self).clear_cached_attrs(attr_defaults) self._metadata_imds = sources.UNSET @azure_ds_telemetry_reporter def _get_data(self): """Crawl and process datasource metadata caching metadata as attrs. @return: True on success, False on error, invalid or disabled datasource. """ if not self._is_platform_viable(): return False try: get_boot_telemetry() except Exception as e: LOG.warning("Failed to get boot telemetry: %s", e) try: get_system_info() except Exception as e: LOG.warning("Failed to get system information: %s", e) self.distro.networking.blacklist_drivers = BLACKLIST_DRIVERS try: crawled_data = util.log_time( logfunc=LOG.debug, msg="Crawl of metadata service", func=self.crawl_metadata, ) except Exception as e: report_diagnostic_event( "Could not crawl Azure metadata: %s" % e, logger_func=LOG.error ) self._report_failure( description=DEFAULT_REPORT_FAILURE_USER_VISIBLE_MESSAGE ) return False if ( self.distro and self.distro.name == "ubuntu" and self.ds_cfg.get("apply_network_config") ): maybe_remove_ubuntu_network_config_scripts() # Process crawled data and augment with various config defaults # Only merge in default cloud config related to the ephemeral disk # if the ephemeral disk exists devpath = RESOURCE_DISK_PATH if os.path.exists(devpath): report_diagnostic_event( "Ephemeral resource disk '%s' exists. " "Merging default Azure cloud ephemeral disk configs." % devpath, logger_func=LOG.debug, ) self.cfg = util.mergemanydict( [crawled_data["cfg"], BUILTIN_CLOUD_EPHEMERAL_DISK_CONFIG] ) else: report_diagnostic_event( "Ephemeral resource disk '%s' does not exist. " "Not merging default Azure cloud ephemeral disk configs." % devpath, logger_func=LOG.debug, ) self.cfg = crawled_data["cfg"] self._metadata_imds = crawled_data["metadata"]["imds"] self.metadata = util.mergemanydict( [crawled_data["metadata"], DEFAULT_METADATA] ) self.userdata_raw = crawled_data["userdata_raw"] user_ds_cfg = util.get_cfg_by_path(self.cfg, DS_CFG_PATH, {}) self.ds_cfg = util.mergemanydict([user_ds_cfg, self.ds_cfg]) # walinux agent writes files world readable, but expects # the directory to be protected. write_files( self.ds_cfg["data_dir"], crawled_data["files"], dirmode=0o700 ) return True @azure_ds_telemetry_reporter def get_imds_data_with_api_fallback( self, fallback_nic, retries, md_type=MetadataType.ALL, exc_cb=retry_on_url_exc, infinite=False, ): """ Wrapper for get_metadata_from_imds so that we can have flexibility in which IMDS api-version we use. If a particular instance of IMDS does not have the api version that is desired, we want to make this fault tolerant and fall back to a good known minimum api version. """ for _ in range(retries): try: LOG.info("Attempting IMDS api-version: %s", IMDS_VER_WANT) return get_metadata_from_imds( fallback_nic=fallback_nic, retries=0, md_type=md_type, api_version=IMDS_VER_WANT, exc_cb=exc_cb, ) except UrlError as err: LOG.info("UrlError with IMDS api-version: %s", IMDS_VER_WANT) if err.code == 400: log_msg = "Fall back to IMDS api-version: {}".format( IMDS_VER_MIN ) report_diagnostic_event(log_msg, logger_func=LOG.info) break LOG.info("Using IMDS api-version: %s", IMDS_VER_MIN) return get_metadata_from_imds( fallback_nic=fallback_nic, retries=retries, md_type=md_type, api_version=IMDS_VER_MIN, exc_cb=exc_cb, infinite=infinite, ) def device_name_to_device(self, name): return self.ds_cfg["disk_aliases"].get(name) @azure_ds_telemetry_reporter def get_public_ssh_keys(self): """ Retrieve public SSH keys. """ return self._get_public_ssh_keys_and_source().ssh_keys def _get_public_ssh_keys_and_source(self): """ Try to get the ssh keys from IMDS first, and if that fails (i.e. IMDS is unavailable) then fallback to getting the ssh keys from OVF. The benefit to getting keys from IMDS is a large performance advantage, so this is a strong preference. But we must keep OVF as a second option for environments that don't have IMDS. """ LOG.debug("Retrieving public SSH keys") ssh_keys = [] keys_from_imds = True LOG.debug("Attempting to get SSH keys from IMDS") try: ssh_keys = [ public_key["keyData"] for public_key in self.metadata["imds"]["compute"][ "publicKeys" ] ] for key in ssh_keys: if not _key_is_openssh_formatted(key=key): keys_from_imds = False break if not keys_from_imds: log_msg = "Keys not in OpenSSH format, using OVF" else: log_msg = "Retrieved {} keys from IMDS".format( len(ssh_keys) if ssh_keys is not None else 0 ) except KeyError: log_msg = "Unable to get keys from IMDS, falling back to OVF" keys_from_imds = False finally: report_diagnostic_event(log_msg, logger_func=LOG.debug) if not keys_from_imds: LOG.debug("Attempting to get SSH keys from OVF") try: ssh_keys = self.metadata["public-keys"] log_msg = "Retrieved {} keys from OVF".format(len(ssh_keys)) except KeyError: log_msg = "No keys available from OVF" finally: report_diagnostic_event(log_msg, logger_func=LOG.debug) return SSHKeys(keys_from_imds=keys_from_imds, ssh_keys=ssh_keys) def get_config_obj(self): return self.cfg def check_instance_id(self, sys_cfg): # quickly (local check only) if self.instance_id is still valid return sources.instance_id_matches_system_uuid(self.get_instance_id()) def _iid(self, previous=None): prev_iid_path = os.path.join( self.paths.get_cpath("data"), "instance-id" ) # Older kernels than 4.15 will have UPPERCASE product_uuid. # We don't want Azure to react to an UPPER/lower difference as a new # instance id as it rewrites SSH host keys. # LP: #1835584 iid = dmi.read_dmi_data("system-uuid").lower() if os.path.exists(prev_iid_path): previous = util.load_file(prev_iid_path).strip() if previous.lower() == iid: # If uppercase/lowercase equivalent, return the previous value # to avoid new instance id. return previous if is_byte_swapped(previous.lower(), iid): return previous return iid @azure_ds_telemetry_reporter def setup(self, is_new_instance): if self._negotiated is False: LOG.debug( "negotiating for %s (new_instance=%s)", self.get_instance_id(), is_new_instance, ) fabric_data = self._negotiate() LOG.debug("negotiating returned %s", fabric_data) if fabric_data: self.metadata.update(fabric_data) self._negotiated = True else: LOG.debug( "negotiating already done for %s", self.get_instance_id() ) @azure_ds_telemetry_reporter def _wait_for_nic_detach(self, nl_sock): """Use the netlink socket provided to wait for nic detach event. NOTE: The function doesn't close the socket. The caller owns closing the socket and disposing it safely. """ try: ifname = None # Preprovisioned VM will only have one NIC, and it gets # detached immediately after deployment. with events.ReportEventStack( name="wait-for-nic-detach", description="wait for nic detach", parent=azure_ds_reporter, ): ifname = netlink.wait_for_nic_detach_event(nl_sock) if ifname is None: msg = ( "Preprovisioned nic not detached as expected. " "Proceeding without failing." ) report_diagnostic_event(msg, logger_func=LOG.warning) else: report_diagnostic_event( "The preprovisioned nic %s is detached" % ifname, logger_func=LOG.warning, ) path = REPROVISION_NIC_DETACHED_MARKER_FILE LOG.info("Creating a marker file for nic detached: %s", path) util.write_file( path, "{pid}: {time}\n".format(pid=os.getpid(), time=time()) ) except AssertionError as error: report_diagnostic_event(error, logger_func=LOG.error) raise @azure_ds_telemetry_reporter def wait_for_link_up(self, ifname): """In cases where the link state is still showing down after a nic is hot-attached, we can attempt to bring it up by forcing the hv_netvsc drivers to query the link state by unbinding and then binding the device. This function attempts infinitely until the link is up, because we cannot proceed further until we have a stable link.""" if self.distro.networking.try_set_link_up(ifname): report_diagnostic_event( "The link %s is already up." % ifname, logger_func=LOG.info ) return LOG.debug("Attempting to bring %s up", ifname) attempts = 0 LOG.info("Unbinding and binding the interface %s", ifname) while True: devicename = net.read_sys_net(ifname, "device/device_id").strip( "{}" ) util.write_file( "/sys/bus/vmbus/drivers/hv_netvsc/unbind", devicename ) util.write_file( "/sys/bus/vmbus/drivers/hv_netvsc/bind", devicename ) attempts = attempts + 1 if self.distro.networking.try_set_link_up(ifname): msg = "The link %s is up after %s attempts" % ( ifname, attempts, ) report_diagnostic_event(msg, logger_func=LOG.info) return if attempts % 10 == 0: msg = "Link is not up after %d attempts to rebind" % attempts report_diagnostic_event(msg, logger_func=LOG.info) LOG.info(msg) # It could take some time after rebind for the interface to be up. # So poll for the status for some time before attempting to rebind # again. sleep_duration = 0.5 max_status_polls = 20 LOG.debug( "Polling %d seconds for primary NIC link up after rebind.", sleep_duration * max_status_polls, ) for i in range(0, max_status_polls): if self.distro.networking.is_up(ifname): msg = ( "After %d attempts to rebind, link is up after " "polling the link status %d times" % (attempts, i) ) report_diagnostic_event(msg, logger_func=LOG.info) LOG.debug(msg) return else: sleep(sleep_duration) @azure_ds_telemetry_reporter def _create_report_ready_marker(self): path = REPORTED_READY_MARKER_FILE LOG.info("Creating a marker file to report ready: %s", path) util.write_file( path, "{pid}: {time}\n".format(pid=os.getpid(), time=time()) ) report_diagnostic_event( "Successfully created reported ready marker file " "while in the preprovisioning pool.", logger_func=LOG.debug, ) @azure_ds_telemetry_reporter def _report_ready_if_needed(self): """Report ready to the platform if the marker file is not present, and create the marker file. """ have_not_reported_ready = not os.path.isfile( REPORTED_READY_MARKER_FILE ) if have_not_reported_ready: report_diagnostic_event( "Reporting ready before nic detach", logger_func=LOG.info ) try: with EphemeralDHCPv4WithReporting(azure_ds_reporter) as lease: self._report_ready(lease=lease) except Exception as e: report_diagnostic_event( "Exception reporting ready during " "preprovisioning before nic detach: %s" % e, logger_func=LOG.error, ) raise self._create_report_ready_marker() else: report_diagnostic_event( "Already reported ready before nic detach." " The marker file already exists: %s" % REPORTED_READY_MARKER_FILE, logger_func=LOG.error, ) @azure_ds_telemetry_reporter def _check_if_nic_is_primary(self, ifname): """Check if a given interface is the primary nic or not. If it is the primary nic, then we also get the expected total nic count from IMDS. IMDS will process the request and send a response only for primary NIC. """ is_primary = False expected_nic_count = -1 imds_md = None metadata_poll_count = 0 metadata_logging_threshold = 1 expected_errors_count = 0 # For now, only a VM's primary NIC can contact IMDS and WireServer. If # DHCP fails for a NIC, we have no mechanism to determine if the NIC is # primary or secondary. In this case, the desired behavior is to fail # VM provisioning if there is any DHCP failure when trying to determine # the primary NIC. try: dhcp_ctx = EphemeralDHCPv4WithReporting( azure_ds_reporter, iface=ifname ) dhcp_ctx.obtain_lease() except Exception as e: report_diagnostic_event( "Giving up. Failed to obtain dhcp lease " "for %s when attempting to determine " "primary NIC during reprovision due to %s" % (ifname, e), logger_func=LOG.error, ) raise # Retry polling network metadata for a limited duration only when the # calls fail due to network unreachable error or timeout. # This is because the platform drops packets going towards IMDS # when it is not a primary nic. If the calls fail due to other issues # like 410, 503 etc, then it means we are primary but IMDS service # is unavailable at the moment. Retry indefinitely in those cases # since we cannot move on without the network metadata. In the future, # all this will not be necessary, as a new dhcp option would tell # whether the nic is primary or not. def network_metadata_exc_cb(msg, exc): nonlocal expected_errors_count, metadata_poll_count nonlocal metadata_logging_threshold metadata_poll_count = metadata_poll_count + 1 # Log when needed but back off exponentially to avoid exploding # the log file. if metadata_poll_count >= metadata_logging_threshold: metadata_logging_threshold *= 2 report_diagnostic_event( "Ran into exception when attempting to reach %s " "after %d polls." % (msg, metadata_poll_count), logger_func=LOG.error, ) if isinstance(exc, UrlError): report_diagnostic_event( "poll IMDS with %s failed. Exception: %s and code: %s" % (msg, exc.cause, exc.code), logger_func=LOG.error, ) # Retry up to a certain limit for both timeout and network # unreachable errors. if exc.cause and isinstance( exc.cause, (requests.Timeout, requests.ConnectionError) ): expected_errors_count = expected_errors_count + 1 return expected_errors_count <= 10 return True # Primary nic detection will be optimized in the future. The fact that # primary nic is being attached first helps here. Otherwise each nic # could add several seconds of delay. try: imds_md = self.get_imds_data_with_api_fallback( ifname, 0, MetadataType.NETWORK, network_metadata_exc_cb, True ) except Exception as e: LOG.warning( "Failed to get network metadata using nic %s. Attempt to " "contact IMDS failed with error %s. Assuming this is not the " "primary nic.", ifname, e, ) finally: # If we are not the primary nic, then clean the dhcp context. if imds_md is None: dhcp_ctx.clean_network() if imds_md is not None: # Only primary NIC will get a response from IMDS. LOG.info("%s is the primary nic", ifname) is_primary = True # If primary, set ephemeral dhcp ctx so we can report ready self._ephemeral_dhcp_ctx = dhcp_ctx # Set the expected nic count based on the response received. expected_nic_count = len(imds_md["interface"]) report_diagnostic_event( "Expected nic count: %d" % expected_nic_count, logger_func=LOG.info, ) return is_primary, expected_nic_count @azure_ds_telemetry_reporter def _wait_for_hot_attached_nics(self, nl_sock): """Wait until all the expected nics for the vm are hot-attached. The expected nic count is obtained by requesting the network metadata from IMDS. """ LOG.info("Waiting for nics to be hot-attached") try: # Wait for nics to be attached one at a time, until we know for # sure that all nics have been attached. nics_found = [] primary_nic_found = False expected_nic_count = -1 # Wait for netlink nic attach events. After the first nic is # attached, we are already in the customer vm deployment path and # so eerything from then on should happen fast and avoid # unnecessary delays wherever possible. while True: ifname = None with events.ReportEventStack( name="wait-for-nic-attach", description=( "wait for nic attach after %d nics have been attached" % len(nics_found) ), parent=azure_ds_reporter, ): ifname = netlink.wait_for_nic_attach_event( nl_sock, nics_found ) # wait_for_nic_attach_event guarantees that ifname it not None nics_found.append(ifname) report_diagnostic_event( "Detected nic %s attached." % ifname, logger_func=LOG.info ) # Attempt to bring the interface's operating state to # UP in case it is not already. self.wait_for_link_up(ifname) # If primary nic is not found, check if this is it. The # platform will attach the primary nic first so we # won't be in primary_nic_found = false state for long. if not primary_nic_found: LOG.info("Checking if %s is the primary nic", ifname) ( primary_nic_found, expected_nic_count, ) = self._check_if_nic_is_primary(ifname) # Exit criteria: check if we've discovered all nics if ( expected_nic_count != -1 and len(nics_found) >= expected_nic_count ): LOG.info("Found all the nics for this VM.") break except AssertionError as error: report_diagnostic_event(error, logger_func=LOG.error) @azure_ds_telemetry_reporter def _wait_for_all_nics_ready(self): """Wait for nic(s) to be hot-attached. There may be multiple nics depending on the customer request. But only primary nic would be able to communicate with wireserver and IMDS. So we detect and save the primary nic to be used later. """ nl_sock = None try: nl_sock = netlink.create_bound_netlink_socket() report_ready_marker_present = bool( os.path.isfile(REPORTED_READY_MARKER_FILE) ) # Report ready if the marker file is not already present. # The nic of the preprovisioned vm gets hot-detached as soon as # we report ready. So no need to save the dhcp context. self._report_ready_if_needed() has_nic_been_detached = bool( os.path.isfile(REPROVISION_NIC_DETACHED_MARKER_FILE) ) if not has_nic_been_detached: LOG.info("NIC has not been detached yet.") self._wait_for_nic_detach(nl_sock) # If we know that the preprovisioned nic has been detached, and we # still have a fallback nic, then it means the VM must have # rebooted as part of customer assignment, and all the nics have # already been attached by the Azure platform. So there is no need # to wait for nics to be hot-attached. if not self.fallback_interface: self._wait_for_hot_attached_nics(nl_sock) else: report_diagnostic_event( "Skipping waiting for nic attach " "because we already have a fallback " "interface. Report Ready marker " "present before detaching nics: %s" % report_ready_marker_present, logger_func=LOG.info, ) except netlink.NetlinkCreateSocketError as e: report_diagnostic_event(e, logger_func=LOG.warning) raise finally: if nl_sock: nl_sock.close() @azure_ds_telemetry_reporter def _poll_imds(self): """Poll IMDS for the new provisioning data until we get a valid response. Then return the returned JSON object.""" url = "{}?api-version={}".format( MetadataType.REPROVISION_DATA.value, IMDS_VER_MIN ) headers = {"Metadata": "true"} nl_sock = None report_ready = bool(not os.path.isfile(REPORTED_READY_MARKER_FILE)) self.imds_logging_threshold = 1 self.imds_poll_counter = 1 dhcp_attempts = 0 vnet_switched = False return_val = None def exc_cb(msg, exception): if isinstance(exception, UrlError): if exception.code in (404, 410): if self.imds_poll_counter == self.imds_logging_threshold: # Reducing the logging frequency as we are polling IMDS self.imds_logging_threshold *= 2 LOG.debug( "Backing off logging threshold for the same " "exception to %d", self.imds_logging_threshold, ) report_diagnostic_event( "poll IMDS with %s failed. " "Exception: %s and code: %s" % (msg, exception.cause, exception.code), logger_func=LOG.debug, ) self.imds_poll_counter += 1 return True else: # If we get an exception while trying to call IMDS, we call # DHCP and setup the ephemeral network to acquire a new IP. report_diagnostic_event( "poll IMDS with %s failed. Exception: %s and code: %s" % (msg, exception.cause, exception.code), logger_func=LOG.warning, ) return False report_diagnostic_event( "poll IMDS failed with an unexpected exception: %s" % exception, logger_func=LOG.warning, ) return False # When the interface is hot-attached, we would have already # done dhcp and set the dhcp context. In that case, skip # the attempt to do dhcp. is_ephemeral_ctx_present = self._ephemeral_dhcp_ctx is not None msg = ( "Unexpected error. Dhcp context is not expected to be already " "set when we need to wait for vnet switch" ) if is_ephemeral_ctx_present and report_ready: report_diagnostic_event(msg, logger_func=LOG.error) raise RuntimeError(msg) while True: try: # Since is_ephemeral_ctx_present is set only once, this ensures # that with regular reprovisioning, dhcp is always done every # time the loop runs. if not is_ephemeral_ctx_present: # Save our EphemeralDHCPv4 context to avoid repeated dhcp # later when we report ready self._ephemeral_dhcp_ctx = EphemeralDHCPv4WithReporting( azure_ds_reporter ) lease = self._ephemeral_dhcp_ctx.obtain_lease() if vnet_switched: dhcp_attempts += 1 if report_ready: try: nl_sock = netlink.create_bound_netlink_socket() except netlink.NetlinkCreateSocketError as e: report_diagnostic_event( "Failed to create bound netlink socket: %s" % e, logger_func=LOG.warning, ) self._ephemeral_dhcp_ctx.clean_network() break report_ready_succeeded = self._report_ready(lease=lease) if not report_ready_succeeded: msg = ( "Failed reporting ready while in " "the preprovisioning pool." ) report_diagnostic_event(msg, logger_func=LOG.error) self._ephemeral_dhcp_ctx.clean_network() raise sources.InvalidMetaDataException(msg) self._create_report_ready_marker() report_ready = False LOG.debug("Wait for vnetswitch to happen") with events.ReportEventStack( name="wait-for-media-disconnect-connect", description="wait for vnet switch", parent=azure_ds_reporter, ): try: netlink.wait_for_media_disconnect_connect( nl_sock, lease["interface"] ) except AssertionError as e: report_diagnostic_event( "Error while waiting for vnet switch: %s" % e, logger_func=LOG.error, ) break vnet_switched = True self._ephemeral_dhcp_ctx.clean_network() else: with events.ReportEventStack( name="get-reprovision-data-from-imds", description="get reprovision data from imds", parent=azure_ds_reporter, ): return_val = readurl( url, timeout=IMDS_TIMEOUT_IN_SECONDS, headers=headers, exception_cb=exc_cb, infinite=True, log_req_resp=False, ).contents break except UrlError: # Teardown our EphemeralDHCPv4 context on failure as we retry self._ephemeral_dhcp_ctx.clean_network() # Also reset this flag which determines if we should do dhcp # during retries. is_ephemeral_ctx_present = False finally: if nl_sock: nl_sock.close() if vnet_switched: report_diagnostic_event( "attempted dhcp %d times after reuse" % dhcp_attempts, logger_func=LOG.debug, ) report_diagnostic_event( "polled imds %d times after reuse" % self.imds_poll_counter, logger_func=LOG.debug, ) return return_val @azure_ds_telemetry_reporter def _report_failure(self, description: Optional[str] = None) -> bool: """Tells the Azure fabric that provisioning has failed. @param description: A description of the error encountered. @return: The success status of sending the failure signal. """ unknown_245_key = "unknown-245" try: if ( self.distro.networking.is_up(self.fallback_interface) and self._ephemeral_dhcp_ctx and self._ephemeral_dhcp_ctx.lease and unknown_245_key in self._ephemeral_dhcp_ctx.lease ): report_diagnostic_event( "Using cached ephemeral dhcp context " "to report failure to Azure", logger_func=LOG.debug, ) report_failure_to_fabric( dhcp_opts=self._ephemeral_dhcp_ctx.lease[unknown_245_key], description=description, ) self._ephemeral_dhcp_ctx.clean_network() # Teardown ephemeral return True except Exception as e: report_diagnostic_event( "Failed to report failure using " "cached ephemeral dhcp context: %s" % e, logger_func=LOG.error, ) try: report_diagnostic_event( "Using new ephemeral dhcp to report failure to Azure", logger_func=LOG.debug, ) with EphemeralDHCPv4WithReporting(azure_ds_reporter) as lease: report_failure_to_fabric( dhcp_opts=lease[unknown_245_key], description=description ) return True except Exception as e: report_diagnostic_event( "Failed to report failure using new ephemeral dhcp: %s" % e, logger_func=LOG.debug, ) return False def _report_ready(self, lease: dict) -> bool: """Tells the fabric provisioning has completed. @param lease: dhcp lease to use for sending the ready signal. @return: The success status of sending the ready signal. """ try: get_metadata_from_fabric( fallback_lease_file=None, dhcp_opts=lease["unknown-245"], iso_dev=self.iso_dev, ) return True except Exception as e: report_diagnostic_event( "Error communicating with Azure fabric; You may experience " "connectivity issues: %s" % e, logger_func=LOG.warning, ) return False def _ppstype_from_imds(self, imds_md: dict = None) -> str: try: return imds_md["extended"]["compute"]["ppsType"] except Exception as e: report_diagnostic_event( "Could not retrieve pps configuration from IMDS: %s" % e, logger_func=LOG.debug, ) return None def _determine_pps_type(self, ovf_cfg: dict, imds_md: dict) -> PPSType: """Determine PPS type using OVF, IMDS data, and reprovision marker.""" if os.path.isfile(REPROVISION_MARKER_FILE): pps_type = PPSType.UNKNOWN elif ( ovf_cfg.get("PreprovisionedVMType", None) == PPSType.SAVABLE.value or self._ppstype_from_imds(imds_md) == PPSType.SAVABLE.value ): pps_type = PPSType.SAVABLE elif ( ovf_cfg.get("PreprovisionedVm") is True or ovf_cfg.get("PreprovisionedVMType", None) == PPSType.RUNNING.value or self._ppstype_from_imds(imds_md) == PPSType.RUNNING.value ): pps_type = PPSType.RUNNING else: pps_type = PPSType.NONE report_diagnostic_event( "PPS type: %s" % pps_type.value, logger_func=LOG.info ) return pps_type def _write_reprovision_marker(self): """Write reprovision marker file in case system is rebooted.""" LOG.info( "Creating a marker file to poll imds: %s", REPROVISION_MARKER_FILE ) util.write_file( REPROVISION_MARKER_FILE, "{pid}: {time}\n".format(pid=os.getpid(), time=time()), ) def _reprovision(self): """Initiate the reprovisioning workflow.""" contents = self._poll_imds() with events.ReportEventStack( name="reprovisioning-read-azure-ovf", description="read azure ovf during reprovisioning", parent=azure_ds_reporter, ): md, ud, cfg = read_azure_ovf(contents) return (md, ud, cfg, {"ovf-env.xml": contents}) @azure_ds_telemetry_reporter def _negotiate(self): """Negotiate with fabric and return data from it. On success, returns a dictionary including 'public_keys'. On failure, returns False. """ pubkey_info = None ssh_keys_and_source = self._get_public_ssh_keys_and_source() if not ssh_keys_and_source.keys_from_imds: pubkey_info = self.cfg.get("_pubkeys", None) log_msg = "Retrieved {} fingerprints from OVF".format( len(pubkey_info) if pubkey_info is not None else 0 ) report_diagnostic_event(log_msg, logger_func=LOG.debug) metadata_func = partial( get_metadata_from_fabric, fallback_lease_file=self.dhclient_lease_file, pubkey_info=pubkey_info, ) LOG.debug("negotiating with fabric") try: fabric_data = metadata_func() except Exception as e: report_diagnostic_event( "Error communicating with Azure fabric; You may experience " "connectivity issues: %s" % e, logger_func=LOG.warning, ) return False util.del_file(REPORTED_READY_MARKER_FILE) util.del_file(REPROVISION_MARKER_FILE) util.del_file(REPROVISION_NIC_DETACHED_MARKER_FILE) return fabric_data @azure_ds_telemetry_reporter def activate(self, cfg, is_new_instance): try: address_ephemeral_resize( is_new_instance=is_new_instance, preserve_ntfs=self.ds_cfg.get(DS_CFG_KEY_PRESERVE_NTFS, False), ) finally: push_log_to_kvp(self.sys_cfg["def_log_file"]) return @property def availability_zone(self): return ( self.metadata.get("imds", {}) .get("compute", {}) .get("platformFaultDomain") ) @property def network_config(self): """Generate a network config like net.generate_fallback_network() with the following exceptions. 1. Probe the drivers of the net-devices present and inject them in the network configuration under params: driver: value 2. Generate a fallback network config that does not include any of the blacklisted devices. """ if not self._network_config or self._network_config == sources.UNSET: if self.ds_cfg.get("apply_network_config"): nc_src = self._metadata_imds else: nc_src = None self._network_config = parse_network_config(nc_src) return self._network_config @property def region(self): return self.metadata.get("imds", {}).get("compute", {}).get("location") def _username_from_imds(imds_data): try: return imds_data["compute"]["osProfile"]["adminUsername"] except KeyError: return None def _userdata_from_imds(imds_data): try: return imds_data["compute"]["userData"] except KeyError: return None def _hostname_from_imds(imds_data): try: return imds_data["compute"]["osProfile"]["computerName"] except KeyError: return None def _disable_password_from_imds(imds_data): try: return ( imds_data["compute"]["osProfile"]["disablePasswordAuthentication"] == "true" ) except KeyError: return None def _key_is_openssh_formatted(key): """ Validate whether or not the key is OpenSSH-formatted. """ # See https://bugs.launchpad.net/cloud-init/+bug/1910835 if "\r\n" in key.strip(): return False parser = ssh_util.AuthKeyLineParser() try: akl = parser.parse(key) except TypeError: return False return akl.keytype is not None def _partitions_on_device(devpath, maxnum=16): # return a list of tuples (ptnum, path) for each part on devpath for suff in ("-part", "p", ""): found = [] for pnum in range(1, maxnum): ppath = devpath + suff + str(pnum) if os.path.exists(ppath): found.append((pnum, os.path.realpath(ppath))) if found: return found return [] @azure_ds_telemetry_reporter def _has_ntfs_filesystem(devpath): ntfs_devices = util.find_devs_with("TYPE=ntfs", no_cache=True) LOG.debug("ntfs_devices found = %s", ntfs_devices) return os.path.realpath(devpath) in ntfs_devices @azure_ds_telemetry_reporter def can_dev_be_reformatted(devpath, preserve_ntfs): """Determine if the ephemeral drive at devpath should be reformatted. A fresh ephemeral disk is formatted by Azure and will: a.) have a partition table (dos or gpt) b.) have 1 partition that is ntfs formatted, or have 2 partitions with the second partition ntfs formatted. (larger instances with >2TB ephemeral disk have gpt, and will have a microsoft reserved partition as part 1. LP: #1686514) c.) the ntfs partition will have no files other than possibly 'dataloss_warning_readme.txt' User can indicate that NTFS should never be destroyed by setting DS_CFG_KEY_PRESERVE_NTFS in dscfg. If data is found on NTFS, user is warned to set DS_CFG_KEY_PRESERVE_NTFS to make sure cloud-init does not accidentally wipe their data. If cloud-init cannot mount the disk to check for data, destruction will be allowed, unless the dscfg key is set.""" if preserve_ntfs: msg = "config says to never destroy NTFS (%s.%s), skipping checks" % ( ".".join(DS_CFG_PATH), DS_CFG_KEY_PRESERVE_NTFS, ) return False, msg if not os.path.exists(devpath): return False, "device %s does not exist" % devpath LOG.debug( "Resolving realpath of %s -> %s", devpath, os.path.realpath(devpath) ) # devpath of /dev/sd[a-z] or /dev/disk/cloud/azure_resource # where partitions are "1" or "-part1" or "p1" partitions = _partitions_on_device(devpath) if len(partitions) == 0: return False, "device %s was not partitioned" % devpath elif len(partitions) > 2: msg = "device %s had 3 or more partitions: %s" % ( devpath, " ".join([p[1] for p in partitions]), ) return False, msg elif len(partitions) == 2: cand_part, cand_path = partitions[1] else: cand_part, cand_path = partitions[0] if not _has_ntfs_filesystem(cand_path): msg = "partition %s (%s) on device %s was not ntfs formatted" % ( cand_part, cand_path, devpath, ) return False, msg @azure_ds_telemetry_reporter def count_files(mp): ignored = set(["dataloss_warning_readme.txt"]) return len([f for f in os.listdir(mp) if f.lower() not in ignored]) bmsg = "partition %s (%s) on device %s was ntfs formatted" % ( cand_part, cand_path, devpath, ) with events.ReportEventStack( name="mount-ntfs-and-count", description="mount-ntfs-and-count", parent=azure_ds_reporter, ) as evt: try: file_count = util.mount_cb( cand_path, count_files, mtype="ntfs", update_env_for_mount={"LANG": "C"}, ) except util.MountFailedError as e: evt.description = "cannot mount ntfs" if "unknown filesystem type 'ntfs'" in str(e): return ( True, ( bmsg + " but this system cannot mount NTFS," " assuming there are no important files." " Formatting allowed." ), ) return False, bmsg + " but mount of %s failed: %s" % (cand_part, e) if file_count != 0: evt.description = "mounted and counted %d files" % file_count LOG.warning( "it looks like you're using NTFS on the ephemeral" " disk, to ensure that filesystem does not get wiped," " set %s.%s in config", ".".join(DS_CFG_PATH), DS_CFG_KEY_PRESERVE_NTFS, ) return False, bmsg + " but had %d files on it." % file_count return True, bmsg + " and had no important files. Safe for reformatting." @azure_ds_telemetry_reporter def address_ephemeral_resize( devpath=RESOURCE_DISK_PATH, is_new_instance=False, preserve_ntfs=False ): if not os.path.exists(devpath): report_diagnostic_event( "Ephemeral resource disk '%s' does not exist." % devpath, logger_func=LOG.debug, ) return else: report_diagnostic_event( "Ephemeral resource disk '%s' exists." % devpath, logger_func=LOG.debug, ) result = False msg = None if is_new_instance: result, msg = (True, "First instance boot.") else: result, msg = can_dev_be_reformatted(devpath, preserve_ntfs) LOG.debug("reformattable=%s: %s", result, msg) if not result: return for mod in ["disk_setup", "mounts"]: sempath = "/var/lib/cloud/instance/sem/config_" + mod bmsg = 'Marker "%s" for module "%s"' % (sempath, mod) if os.path.exists(sempath): try: os.unlink(sempath) LOG.debug("%s removed.", bmsg) except Exception as e: # python3 throws FileNotFoundError, python2 throws OSError LOG.warning("%s: remove failed! (%s)", bmsg, e) else: LOG.debug("%s did not exist.", bmsg) return @azure_ds_telemetry_reporter def write_files(datadir, files, dirmode=None): def _redact_password(cnt, fname): """Azure provides the UserPassword in plain text. So we redact it""" try: root = ET.fromstring(cnt) for elem in root.iter(): if ( "UserPassword" in elem.tag and elem.text != DEF_PASSWD_REDACTION ): elem.text = DEF_PASSWD_REDACTION return ET.tostring(root) except Exception: LOG.critical("failed to redact userpassword in %s", fname) return cnt if not datadir: return if not files: files = {} util.ensure_dir(datadir, dirmode) for (name, content) in files.items(): fname = os.path.join(datadir, name) if "ovf-env.xml" in name: content = _redact_password(content, fname) util.write_file(filename=fname, content=content, mode=0o600) def find_child(node, filter_func): ret = [] if not node.hasChildNodes(): return ret for child in node.childNodes: if filter_func(child): ret.append(child) return ret @azure_ds_telemetry_reporter def load_azure_ovf_pubkeys(sshnode): # This parses a 'SSH' node formatted like below, and returns # an array of dicts. # [{'fingerprint': '6BE7A7C3C8A8F4B123CCA5D0C2F1BE4CA7B63ED7', # 'path': '/where/to/go'}] # # # ABC/x/y/z # ... # # Under some circumstances, there may be a element along with the # Fingerprint and Path. Pass those along if they appear. results = find_child(sshnode, lambda n: n.localName == "PublicKeys") if len(results) == 0: return [] if len(results) > 1: raise BrokenAzureDataSource( "Multiple 'PublicKeys'(%s) in SSH node" % len(results) ) pubkeys_node = results[0] pubkeys = find_child(pubkeys_node, lambda n: n.localName == "PublicKey") if len(pubkeys) == 0: return [] found = [] text_node = minidom.Document.TEXT_NODE for pk_node in pubkeys: if not pk_node.hasChildNodes(): continue cur = {"fingerprint": "", "path": "", "value": ""} for child in pk_node.childNodes: if child.nodeType == text_node or not child.localName: continue name = child.localName.lower() if name not in cur.keys(): continue if ( len(child.childNodes) != 1 or child.childNodes[0].nodeType != text_node ): continue cur[name] = child.childNodes[0].wholeText.strip() found.append(cur) return found @azure_ds_telemetry_reporter def read_azure_ovf(contents): try: dom = minidom.parseString(contents) except Exception as e: error_str = "Invalid ovf-env.xml: %s" % e report_diagnostic_event(error_str, logger_func=LOG.warning) raise BrokenAzureDataSource(error_str) from e results = find_child( dom.documentElement, lambda n: n.localName == "ProvisioningSection" ) if len(results) == 0: raise NonAzureDataSource("No ProvisioningSection") if len(results) > 1: raise BrokenAzureDataSource( "found '%d' ProvisioningSection items" % len(results) ) provSection = results[0] lpcs_nodes = find_child( provSection, lambda n: n.localName == "LinuxProvisioningConfigurationSet", ) if len(lpcs_nodes) == 0: raise NonAzureDataSource("No LinuxProvisioningConfigurationSet") if len(lpcs_nodes) > 1: raise BrokenAzureDataSource( "found '%d' %ss" % (len(lpcs_nodes), "LinuxProvisioningConfigurationSet") ) lpcs = lpcs_nodes[0] if not lpcs.hasChildNodes(): raise BrokenAzureDataSource("no child nodes of configuration set") md_props = "seedfrom" md = {"azure_data": {}} cfg = {} ud = "" password = None username = None for child in lpcs.childNodes: if child.nodeType == dom.TEXT_NODE or not child.localName: continue name = child.localName.lower() simple = False value = "" if ( len(child.childNodes) == 1 and child.childNodes[0].nodeType == dom.TEXT_NODE ): simple = True value = child.childNodes[0].wholeText attrs = dict([(k, v) for k, v in child.attributes.items()]) # we accept either UserData or CustomData. If both are present # then behavior is undefined. if name == "userdata" or name == "customdata": if attrs.get("encoding") in (None, "base64"): ud = base64.b64decode("".join(value.split())) else: ud = value elif name == "username": username = value elif name == "userpassword": password = value elif name == "hostname": md["local-hostname"] = value elif name == "dscfg": if attrs.get("encoding") in (None, "base64"): dscfg = base64.b64decode("".join(value.split())) else: dscfg = value cfg["datasource"] = {DS_NAME: util.load_yaml(dscfg, default={})} elif name == "ssh": cfg["_pubkeys"] = load_azure_ovf_pubkeys(child) elif name == "disablesshpasswordauthentication": cfg["ssh_pwauth"] = util.is_false(value) elif simple: if name in md_props: md[name] = value else: md["azure_data"][name] = value defuser = {} if username: defuser["name"] = username if password: defuser["lock_passwd"] = False if DEF_PASSWD_REDACTION != password: defuser["passwd"] = cfg["password"] = encrypt_pass(password) if defuser: cfg["system_info"] = {"default_user": defuser} if "ssh_pwauth" not in cfg and password: cfg["ssh_pwauth"] = True preprovisioning_cfg = _get_preprovisioning_cfgs(dom) cfg = util.mergemanydict([cfg, preprovisioning_cfg]) return (md, ud, cfg) @azure_ds_telemetry_reporter def _get_preprovisioning_cfgs(dom): """Read the preprovisioning related flags from ovf and populates a dict with the info. Two flags are in use today: PreprovisionedVm bool and PreprovisionedVMType enum. In the long term, the PreprovisionedVm bool will be deprecated in favor of PreprovisionedVMType string/enum. Only these combinations of values are possible today: - PreprovisionedVm=True and PreprovisionedVMType=Running - PreprovisionedVm=False and PreprovisionedVMType=Savable - PreprovisionedVm is missing and PreprovisionedVMType=Running/Savable - PreprovisionedVm=False and PreprovisionedVMType is missing More specifically, this will never happen: - PreprovisionedVm=True and PreprovisionedVMType=Savable """ cfg = {"PreprovisionedVm": False, "PreprovisionedVMType": None} platform_settings_section = find_child( dom.documentElement, lambda n: n.localName == "PlatformSettingsSection" ) if not platform_settings_section or len(platform_settings_section) == 0: LOG.debug("PlatformSettingsSection not found") return cfg platform_settings = find_child( platform_settings_section[0], lambda n: n.localName == "PlatformSettings", ) if not platform_settings or len(platform_settings) == 0: LOG.debug("PlatformSettings not found") return cfg # Read the PreprovisionedVm bool flag. This should be deprecated when the # platform has removed PreprovisionedVm and only surfaces # PreprovisionedVMType. cfg["PreprovisionedVm"] = _get_preprovisionedvm_cfg_value( platform_settings ) cfg["PreprovisionedVMType"] = _get_preprovisionedvmtype_cfg_value( platform_settings ) return cfg @azure_ds_telemetry_reporter def _get_preprovisionedvm_cfg_value(platform_settings): preprovisionedVm = False # Read the PreprovisionedVm bool flag. This should be deprecated when the # platform has removed PreprovisionedVm and only surfaces # PreprovisionedVMType. preprovisionedVmVal = find_child( platform_settings[0], lambda n: n.localName == "PreprovisionedVm" ) if not preprovisionedVmVal or len(preprovisionedVmVal) == 0: LOG.debug("PreprovisionedVm not found") return preprovisionedVm preprovisionedVm = util.translate_bool( preprovisionedVmVal[0].firstChild.nodeValue ) report_diagnostic_event( "PreprovisionedVm: %s" % preprovisionedVm, logger_func=LOG.info ) return preprovisionedVm @azure_ds_telemetry_reporter def _get_preprovisionedvmtype_cfg_value(platform_settings): preprovisionedVMType = None # Read the PreprovisionedVMType value from the ovf. It can be # 'Running' or 'Savable' or not exist. This enum value is intended to # replace PreprovisionedVm bool flag in the long term. # A Running VM is the same as preprovisioned VMs of today. This is # equivalent to having PreprovisionedVm=True. # A Savable VM is one whose nic is hot-detached immediately after it # reports ready the first time to free up the network resources. # Once assigned to customer, the customer-requested nics are # hot-attached to it and reprovision happens like today. preprovisionedVMTypeVal = find_child( platform_settings[0], lambda n: n.localName == "PreprovisionedVMType" ) if ( not preprovisionedVMTypeVal or len(preprovisionedVMTypeVal) == 0 or preprovisionedVMTypeVal[0].firstChild is None ): LOG.debug("PreprovisionedVMType not found") return preprovisionedVMType preprovisionedVMType = preprovisionedVMTypeVal[0].firstChild.nodeValue report_diagnostic_event( "PreprovisionedVMType: %s" % preprovisionedVMType, logger_func=LOG.info ) return preprovisionedVMType def encrypt_pass(password, salt_id="$6$"): return crypt.crypt(password, salt_id + util.rand_str(strlen=16)) @azure_ds_telemetry_reporter def _check_freebsd_cdrom(cdrom_dev): """Return boolean indicating path to cdrom device has content.""" try: with open(cdrom_dev) as fp: fp.read(1024) return True except IOError: LOG.debug("cdrom (%s) is not configured", cdrom_dev) return False @azure_ds_telemetry_reporter def _get_random_seed(source=PLATFORM_ENTROPY_SOURCE): """Return content random seed file if available, otherwise, return None.""" # azure / hyper-v provides random data here # now update ds_cfg to reflect contents pass in config if source is None: return None seed = util.load_file(source, quiet=True, decode=False) # The seed generally contains non-Unicode characters. load_file puts # them into a str (in python 2) or bytes (in python 3). In python 2, # bad octets in a str cause util.json_dumps() to throw an exception. In # python 3, bytes is a non-serializable type, and the handler load_file # uses applies b64 encoding *again* to handle it. The simplest solution # is to just b64encode the data and then decode it to a serializable # string. Same number of bits of entropy, just with 25% more zeroes. # There's no need to undo this base64-encoding when the random seed is # actually used in cc_seed_random.py. seed = base64.b64encode(seed).decode() return seed @azure_ds_telemetry_reporter def list_possible_azure_ds(seed, cache_dir): yield seed yield DEFAULT_PROVISIONING_ISO_DEV if util.is_FreeBSD(): cdrom_dev = "/dev/cd0" if _check_freebsd_cdrom(cdrom_dev): yield cdrom_dev else: for fstype in ("iso9660", "udf"): yield from util.find_devs_with("TYPE=%s" % fstype) if cache_dir: yield cache_dir @azure_ds_telemetry_reporter def load_azure_ds_dir(source_dir): ovf_file = os.path.join(source_dir, "ovf-env.xml") if not os.path.isfile(ovf_file): raise NonAzureDataSource("No ovf-env file found") with open(ovf_file, "rb") as fp: contents = fp.read() md, ud, cfg = read_azure_ovf(contents) return (md, ud, cfg, {"ovf-env.xml": contents}) @azure_ds_telemetry_reporter def parse_network_config(imds_metadata) -> dict: """Convert imds_metadata dictionary to network v2 configuration. Parses network configuration from imds metadata if present or generate fallback network config excluding mlx4_core devices. @param: imds_metadata: Dict of content read from IMDS network service. @return: Dictionary containing network version 2 standard configuration. """ if imds_metadata != sources.UNSET and imds_metadata: try: return _generate_network_config_from_imds_metadata(imds_metadata) except Exception as e: LOG.error( "Failed generating network config " "from IMDS network metadata: %s", str(e), ) try: return _generate_network_config_from_fallback_config() except Exception as e: LOG.error("Failed generating fallback network config: %s", str(e)) return {} @azure_ds_telemetry_reporter def _generate_network_config_from_imds_metadata(imds_metadata) -> dict: """Convert imds_metadata dictionary to network v2 configuration. Parses network configuration from imds metadata. @param: imds_metadata: Dict of content read from IMDS network service. @return: Dictionary containing network version 2 standard configuration. """ netconfig = {"version": 2, "ethernets": {}} network_metadata = imds_metadata["network"] for idx, intf in enumerate(network_metadata["interface"]): has_ip_address = False # First IPv4 and/or IPv6 address will be obtained via DHCP. # Any additional IPs of each type will be set as static # addresses. nicname = "eth{idx}".format(idx=idx) dhcp_override = {"route-metric": (idx + 1) * 100} dev_config = { "dhcp4": True, "dhcp4-overrides": dhcp_override, "dhcp6": False, } for addr_type in ("ipv4", "ipv6"): addresses = intf.get(addr_type, {}).get("ipAddress", []) # If there are no available IP addresses, then we don't # want to add this interface to the generated config. if not addresses: continue has_ip_address = True if addr_type == "ipv4": default_prefix = "24" else: default_prefix = "128" if addresses: dev_config["dhcp6"] = True # non-primary interfaces should have a higher # route-metric (cost) so default routes prefer # primary nic due to lower route-metric value dev_config["dhcp6-overrides"] = dhcp_override for addr in addresses[1:]: # Append static address config for ip > 1 netPrefix = intf[addr_type]["subnet"][0].get( "prefix", default_prefix ) privateIp = addr["privateIpAddress"] if not dev_config.get("addresses"): dev_config["addresses"] = [] dev_config["addresses"].append( "{ip}/{prefix}".format(ip=privateIp, prefix=netPrefix) ) if dev_config and has_ip_address: mac = ":".join(re.findall(r"..", intf["macAddress"])) dev_config.update( {"match": {"macaddress": mac.lower()}, "set-name": nicname} ) # With netvsc, we can get two interfaces that # share the same MAC, so we need to make sure # our match condition also contains the driver driver = device_driver(nicname) if driver and driver == "hv_netvsc": dev_config["match"]["driver"] = driver netconfig["ethernets"][nicname] = dev_config return netconfig @azure_ds_telemetry_reporter def _generate_network_config_from_fallback_config() -> dict: """Generate fallback network config excluding blacklisted devices. @return: Dictionary containing network version 2 standard configuration. """ return net.generate_fallback_config( blacklist_drivers=BLACKLIST_DRIVERS, config_driver=True ) @azure_ds_telemetry_reporter def get_metadata_from_imds( fallback_nic, retries, md_type=MetadataType.ALL, api_version=IMDS_VER_MIN, exc_cb=retry_on_url_exc, infinite=False, ): """Query Azure's instance metadata service, returning a dictionary. If network is not up, setup ephemeral dhcp on fallback_nic to talk to the IMDS. For more info on IMDS: https://docs.microsoft.com/en-us/azure/virtual-machines/windows/instance-metadata-service @param fallback_nic: String. The name of the nic which requires active network in order to query IMDS. @param retries: The number of retries of the IMDS_URL. @param md_type: Metadata type for IMDS request. @param api_version: IMDS api-version to use in the request. @return: A dict of instance metadata containing compute and network info. """ kwargs = { "logfunc": LOG.debug, "msg": "Crawl of Azure Instance Metadata Service (IMDS)", "func": _get_metadata_from_imds, "args": (retries, exc_cb, md_type, api_version, infinite), } if net.is_up(fallback_nic): return util.log_time(**kwargs) else: try: with EphemeralDHCPv4WithReporting(azure_ds_reporter, fallback_nic): return util.log_time(**kwargs) except Exception as e: report_diagnostic_event( "exception while getting metadata: %s" % e, logger_func=LOG.warning, ) raise @azure_ds_telemetry_reporter def _get_metadata_from_imds( retries, exc_cb, md_type=MetadataType.ALL, api_version=IMDS_VER_MIN, infinite=False, ): url = "{}?api-version={}".format(md_type.value, api_version) headers = {"Metadata": "true"} # support for extended metadata begins with 2021-03-01 if api_version >= IMDS_EXTENDED_VER_MIN and md_type == MetadataType.ALL: url = url + "&extended=true" try: response = readurl( url, timeout=IMDS_TIMEOUT_IN_SECONDS, headers=headers, retries=retries, exception_cb=exc_cb, infinite=infinite, ) except Exception as e: # pylint:disable=no-member if isinstance(e, UrlError) and e.code == 400: raise else: report_diagnostic_event( "Ignoring IMDS instance metadata. " "Get metadata from IMDS failed: %s" % e, logger_func=LOG.warning, ) return {} try: from json.decoder import JSONDecodeError json_decode_error = JSONDecodeError except ImportError: json_decode_error = ValueError try: return util.load_json(str(response)) except json_decode_error as e: report_diagnostic_event( "Ignoring non-json IMDS instance metadata response: %s. " "Loading non-json IMDS response failed: %s" % (str(response), e), logger_func=LOG.warning, ) return {} @azure_ds_telemetry_reporter def maybe_remove_ubuntu_network_config_scripts(paths=None): """Remove Azure-specific ubuntu network config for non-primary nics. @param paths: List of networking scripts or directories to remove when present. In certain supported ubuntu images, static udev rules or netplan yaml config is delivered in the base ubuntu image to support dhcp on any additional interfaces which get attached by a customer at some point after initial boot. Since the Azure datasource can now regenerate network configuration as metadata reports these new devices, we no longer want the udev rules or netplan's 90-hotplug-azure.yaml to configure networking on eth1 or greater as it might collide with cloud-init's configuration. Remove the any existing extended network scripts if the datasource is enabled to write network per-boot. """ if not paths: paths = UBUNTU_EXTENDED_NETWORK_SCRIPTS logged = False for path in paths: if os.path.exists(path): if not logged: LOG.info( "Removing Ubuntu extended network scripts because" " cloud-init updates Azure network configuration on the" " following events: %s.", [EventType.BOOT.value, EventType.BOOT_LEGACY.value], ) logged = True if os.path.isdir(path): util.del_dir(path) else: util.del_file(path) def _is_platform_viable(seed_dir): """Check platform environment to report if this datasource may run.""" with events.ReportEventStack( name="check-platform-viability", description="found azure asset tag", parent=azure_ds_reporter, ) as evt: asset_tag = dmi.read_dmi_data("chassis-asset-tag") if asset_tag == AZURE_CHASSIS_ASSET_TAG: return True msg = "Non-Azure DMI asset tag '%s' discovered." % asset_tag evt.description = msg report_diagnostic_event(msg, logger_func=LOG.debug) if os.path.exists(os.path.join(seed_dir, "ovf-env.xml")): return True return False class BrokenAzureDataSource(Exception): pass class NonAzureDataSource(Exception): pass # Legacy: Must be present in case we load an old pkl object DataSourceAzureNet = DataSourceAzure # Used to match classes to dependencies datasources = [ (DataSourceAzure, (sources.DEP_FILESYSTEM,)), ] # Return a list of data sources that match this set of dependencies def get_datasource_list(depends): return sources.list_from_depends(depends, datasources) # vi: ts=4 expandtab