The virt-qemu-sev-validate program will compare a reported SEV/SEV-ES domain launch measurement, to a computed launch measurement. This determines whether the domain has been tampered with during launch. This initial implementation requires all inputs to be provided explicitly, and as such can run completely offline, without any connection to libvirt. The tool is placed in the libvirt-client-qemu sub-RPM since it is specific to the QEMU driver. Reviewed-by: Ján Tomko <jtomko(a)redhat.com&gt; Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com&gt; --- docs/manpages/meson.build | 1 + docs/manpages/virt-qemu-sev-validate.rst | 207 ++++++++++++++++++ libvirt.spec.in | 2 + tools/meson.build | 5 + tools/virt-qemu-sev-validate | 263 +++++++++++++++++++++++ 5 files changed, 478 insertions(+) create mode 100644 docs/manpages/virt-qemu-sev-validate.rst create mode 100755 tools/virt-qemu-sev-validate diff --git a/docs/manpages/meson.build b/docs/manpages/meson.build index b5556996a4..84b2e247e9 100644 --- a/docs/manpages/meson.build +++ b/docs/manpages/meson.build @@ -20,6 +20,7 @@ docs_man_files = [ { 'name': 'virt-qemu-run', 'section': '1', 'install': conf.has('WITH_QEMU') }, { 'name': 'virt-qemu-qmp-proxy', 'section': '1', 'install': conf.has('WITH_QEMU') }, { 'name': 'virt-xml-validate', 'section': '1', 'install': true }, + { 'name': 'virt-qemu-sev-validate', 'section': '1', 'install': conf.has('WITH_QEMU') }, { 'name': 'libvirt-guests', 'section': '8', 'install': conf.has('WITH_LIBVIRTD') }, { 'name': 'libvirtd', 'section': '8', 'install': conf.has('WITH_LIBVIRTD') }, diff --git a/docs/manpages/virt-qemu-sev-validate.rst b/docs/manpages/virt-qemu-sev-validate.rst new file mode 100644 index 0000000000..0f6c64ba90 --- /dev/null +++ b/docs/manpages/virt-qemu-sev-validate.rst @@ -0,0 +1,207 @@ +====================== +virt-qemu-sev-validate +====================== + +-------------------------------------------- +validate a domain AMD SEV launch measurement +-------------------------------------------- + +:Manual section: 1 +:Manual group: Virtualization Support + +.. contents:: + +SYNOPSIS +======== + + +``virt-qemu-sev-validate`` [*OPTIONS*] + + +DESCRIPTION +=========== + +This program validates the reported measurement for a domain launched with AMD +SEV. If the program exits with a status of zero, the guest owner can be +confident that their guest OS is running under the protection offered by the +SEV / SEV-ES platform. + +Note that the level of protection varies depending on the AMD SEV platform +generation and describing the differences is outside the scope of this +document. + +For the results of this program to be considered trustworthy, it is required to +be run on a machine that is already trusted by the guest owner. This could be a +machine that the guest owner has direct physical control over, or it could be +another virtual machine protected by AMD SEV that has already had its launch +measurement validated. Running this program on the virtualization host will not +produce an answer that can be trusted. + +OPTIONS +======= + +Common options +-------------- + +``-h``, ``--help`` + +Display command line help usage then exit. + +``-d``, ``--debug`` + +Show debug information while running + +``-q``, ``--quiet`` + +Don't print information about the attestation result. + +Guest state options +------------------- + +These options provide information about the state of the guest that needs its +boot attested. + +``--measurement BASE64-STRING`` + +The launch measurement reported by the hypervisor of the domain to be validated. +The measurement must be 48 bytes of binary data encoded as a base64 string. + +``--api-major VERSION`` + +The SEV API major version of the hypervisor the domain is running on. + +``--api-minor VERSION`` + +The SEV API major version of the hypervisor the domain is running on. + +``--build-id ID`` + +The SEV build ID of the hypervisor the domain is running on. + +``--policy POLiCY`` + +The policy bitmask associated with the session launch data of the domain to be +validated. + +Guest config options +-------------------- + +These options provide items needed to calculate the expected domain launch +measurement. This will then be compared to the reported launch measurement. + +``-f PATH``, ``--firmware=PATH`` + +Path to the firmware loader binary. This is the EDK2 build that knows how to +initialize AMD SEV. For the validation to be trustworthy it important that the +firmware build used has no support for loading non-volatile variables from +NVRAM, even if NVRAM is expose to the guest. + +``--tik PATH`` + +TIK file for domain. This file must be exactly 16 bytes in size and contains the +unique transport integrity key associated with the domain session launch data. +This is mutually exclusive with the ``--tk`` argument. + +``--tek PATH`` + +TEK file for domain. This file must be exactly 16 bytes in size and contains the +unique transport encryption key associated with the domain session launch data. +This is mutually exclusive with the ``--tk`` argument. + +``--tk PATH`` + +TEK/TIK combined file for the domain. This file must be exactly 32 bytes in +size, with the first 16 bytes containing the TEK and the last 16 bytes +containing the TIK. This is mutually exclusive with the ``--tik`` and ``--tek`` +arguments. + +EXAMPLES +======== + +Fully offline execution +----------------------- + +This scenario allows a measurement to be securely validated in a completely +offline state without any connection to the hypervisor host. All required +data items must be provided as command line parameters. This usage model is +considered secure, because all input data is provided by the user. + +Validate the measurement of a SEV guest booting from disk: + +:: + + # virt-qemu-sev-validate \ + --firmware OVMF.sev.fd \ + --tk this-guest-tk.bin \ + --measurement Zs2pf19ubFSafpZ2WKkwquXvACx9Wt/BV+eJwQ/taO8jhyIj/F8swFrybR1fZ2ID \ + --api-major 0 \ + --api-minor 24 \ + --build-id 13 \ + --policy 3 + +EXIT STATUS +=========== + +Upon successful attestation of the launch measurement, an exit status of 0 will +be set. + +Upon failure to attest the launch measurement one of the following codes will +be set: + +* **1** - *Guest measurement did not validate* + + Assuming the inputs to this program are correct, the virtual machine launch + has been compromised and it should not be trusted henceforth. + +* **2** - *Usage scenario cannot be supported* + + The way in which this program has been invoked prevent it from being able to + validate the launch measurement. + +* **3** - *unexpected error occurred in the code* + + A logic flaw in this program means it is unable to complete the validation of + the measurement. This is a bug which should be reported to the maintainers. + +AUTHOR +====== + +Daniel P. Berrangé + + +BUGS +==== + +Please report all bugs you discover. This should be done via either: + +#. the mailing list + + `https://libvirt.org/contact.html <https://libvirt.org/contact.html>`_ + +#. the bug tracker + + `https://libvirt.org/bugs.html <https://libvirt.org/bugs.html>`_ + +Alternatively, you may report bugs to your software distributor / vendor. + + +COPYRIGHT +========= + +Copyright (C) 2022 by Red Hat, Inc. + + +LICENSE +======= + +``virt-qemu-sev-validate`` is distributed under the terms of the GNU LGPL v2.1+. +This is free software; see the source for copying conditions. There +is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR +PURPOSE + + +SEE ALSO +======== + +virsh(1), `SEV launch security usage <https://libvirt.org/kbase/launch_security_sev.html>`_, +`https://www.libvirt.org/ <https://www.libvirt.org/>`_ diff --git a/libvirt.spec.in b/libvirt.spec.in index 5ea9ef2912..a24e4d1159 100644 --- a/libvirt.spec.in +++ b/libvirt.spec.in @@ -2179,7 +2179,9 @@ exit 0 %if %{with_qemu} %files client-qemu %{_mandir}/man1/virt-qemu-qmp-proxy.1* +%{_mandir}/man1/virt-qemu-sev-validate.1* %{_bindir}/virt-qemu-qmp-proxy +%{_bindir}/virt-qemu-sev-validate %endif %files libs -f %{name}.lang diff --git a/tools/meson.build b/tools/meson.build index 20509906af..c41c619af4 100644 --- a/tools/meson.build +++ b/tools/meson.build @@ -299,6 +299,11 @@ if conf.has('WITH_SANLOCK') ) endif +if conf.has('WITH_QEMU') + install_data('virt-qemu-sev-validate', + install_dir: bindir) +endif + if conf.has('WITH_LIBVIRTD') configure_file( input: 'libvirt-guests.sh.in', diff --git a/tools/virt-qemu-sev-validate b/tools/virt-qemu-sev-validate new file mode 100755 index 0000000000..7ff54e7623 --- /dev/null +++ b/tools/virt-qemu-sev-validate @@ -0,0 +1,263 @@ +#!/usr/bin/python3 +# +# SPDX-License-Identifier: LGPL-2.1-or-later +# +# Validates a guest AMD SEV launch measurement +# +# A general principle in writing this tool is that it must calculate the +# expected measurement based entirely on information it receives on the CLI +# from the guest owner. +# +# It cannot generally trust information obtained from the guest XML or from the +# virtualization host OS. The main exceptions are: +# +# - The guest measurement +# +# This is a result of cryptographic operation using a shared secret known +# only to the guest owner and SEV platform, not the host OS. +# +# - The guest policy +# +# This is encoded in the launch session blob that is encrypted with a shared +# secret known only to the guest owner and SEV platform, not the host OS. It +# is impossible for the host OS to maliciously launch a guest with different +# policy and the user provided launch session blob. +# +# CAVEAT: the user must ALWAYS create a launch blob with freshly generated +# TIK/TEK for every new VM. Re-use of the same TIK/TEK for multiple VMs +# is insecure. +# +# - The SEV API version / build ID +# +# This does not have an impact on the security of the measurement, unless +# the guest owner needs a guarantee that the host is not using specific +# firmware versions with known flaws. +# + +import argparse +from base64 import b64decode +from hashlib import sha256 +import hmac +import logging +import sys +import traceback + +log = logging.getLogger() + + +class AttestationFailedException(Exception): + pass + + +class UnsupportedUsageException(Exception): + pass + + +class ConfidentialVM(object): + + def __init__(self, + measurement=None, + api_major=None, + api_minor=None, + build_id=None, + policy=None): + self.measurement = measurement + self.api_major = api_major + self.api_minor = api_minor + self.build_id = build_id + self.policy = policy + + self.firmware = None + self.tik = None + self.tek = None + + def load_tik_tek(self, tik_path, tek_path): + with open(tik_path, 'rb') as fh: + self.tik = fh.read() + log.debug("TIK(hex): %s", self.tik.hex()) + + if len(self.tik) != 16: + raise UnsupportedUsageException( + "Expected 16 bytes in TIK file, but got %d" % len(self.tik)) + + with open(tek_path, 'rb') as fh: + self.tek = fh.read() + log.debug("TEK(hex): %s", self.tek.hex()) + + if len(self.tek) != 16: + raise UnsupportedUsageException( + "Expected 16 bytes in TEK file, but got %d" % len(self.tek)) + + def load_tk(self, tk_path): + with open(tk_path, 'rb') as fh: + tk = fh.read() + + if len(tk) != 32: + raise UnsupportedUsageException( + "Expected 32 bytes in TIK/TEK file, but got %d" % len(tk)) + + self.tek = tk[0:16] + self.tik = tk[16:32] + log.debug("TIK(hex): %s", self.tik.hex()) + log.debug("TEK(hex): %s", self.tek.hex()) + + def load_firmware(self, firmware_path): + with open(firmware_path, 'rb') as fh: + self.firmware = fh.read() + log.debug("Firmware(sha256): %s", sha256(self.firmware).hexdigest()) + + # Get the full set of measured launch data for the domain + # + # The measured data that the guest is initialized with is the concatenation + # of the following: + # + # - The firmware blob + def get_measured_data(self): + measured_data = self.firmware + log.debug("Measured-data(sha256): %s", + sha256(measured_data).hexdigest()) + return measured_data + + # Get the reported and computed launch measurements for the domain + # + # AMD Secure Encrypted Virtualization API , section 6.5: + # + # measurement = HMAC(0x04 || API_MAJOR || API_MINOR || BUILD || + # GCTX.POLICY || GCTX.LD || MNONCE; GCTX.TIK) + # + # Where GCTX.LD covers all the measured data the guest is initialized with + # per get_measured_data(). + def get_measurements(self): + measurement = b64decode(self.measurement) + reported = measurement[0:32] + nonce = measurement[32:48] + + measured_data = self.get_measured_data() + msg = ( + bytes([0x4]) + + self.api_major.to_bytes(1, 'little') + + self.api_minor.to_bytes(1, 'little') + + self.build_id.to_bytes(1, 'little') + + self.policy.to_bytes(4, 'little') + + sha256(measured_data).digest() + + nonce + ) + log.debug("Measured-msg(hex): %s", msg.hex()) + + computed = hmac.new(self.tik, msg, 'sha256').digest() + + log.debug("Measurement reported(hex): %s", reported.hex()) + log.debug("Measurement computed(hex): %s", computed.hex()) + + return reported, computed + + def attest(self): + reported, computed = self.get_measurements() + + if reported != computed: + raise AttestationFailedException( + "Measurement does not match, VM is not trustworthy") + + +def parse_command_line(): + parser = argparse.ArgumentParser( + description='Validate guest AMD SEV launch measurement') + parser.add_argument('--debug', '-d', action='store_true', + help='Show debug information') + parser.add_argument('--quiet', '-q', action='store_true', + help='Do not display status') + + # Arguments related to the state of the launched guest + vmstate = parser.add_argument_group("Virtual machine launch state") + vmstate.add_argument('--measurement', '-m', required=True, + help='Measurement for the running domain') + vmstate.add_argument('--api-major', type=int, required=True, + help='SEV API major version for the running domain') + vmstate.add_argument('--api-minor', type=int, required=True, + help='SEV API minor version for the running domain') + vmstate.add_argument('--build-id', type=int, required=True, + help='SEV build ID for the running domain') + vmstate.add_argument('--policy', type=int, required=True, + help='SEV policy for the running domain') + + # Arguments related to calculation of the expected launch measurement + vmconfig = parser.add_argument_group("Virtual machine config") + vmconfig.add_argument('--firmware', '-f', required=True, + help='Path to the firmware binary') + vmconfig.add_argument('--tik', + help='TIK file for domain') + vmconfig.add_argument('--tek', + help='TEK file for domain') + vmconfig.add_argument('--tk', + help='TEK/TIK combined file for domain') + + return parser.parse_args() + + +# Sanity check the set of CLI args specified provide enough info for us to do +# the job +def check_usage(args): + if args.tk is not None: + if args.tik is not None or args.tek is not None: + raise UnsupportedUsageException( + "--tk is mutually exclusive with --tek/--tik") + else: + if args.tik is None or args.tek is None: + raise UnsupportedUsageException( + "Either --tk or both of --tek/--tik are required") + + +def attest(args): + cvm = ConfidentialVM(measurement=args.measurement, + api_major=args.api_major, + api_minor=args.api_minor, + build_id=args.build_id, + policy=args.policy) + + cvm.load_firmware(args.firmware) + + if args.tk is not None: + cvm.load_tk(args.tk) + else: + cvm.load_tik_tek(args.tik, args.tek) + + cvm.attest() + + if not args.quiet: + print("OK: Looks good to me") + +def main(): + args = parse_command_line() + if args.debug: + logging.basicConfig(level="DEBUG") + formatter = logging.Formatter("[%(levelname)s]: %(message)s") + handler = log.handlers[0] + handler.setFormatter(formatter) + + try: + check_usage(args) + + attest(args) + + sys.exit(0) + except AttestationFailedException as e: + if args.debug: + traceback.print_tb(e.__traceback__) + if not args.quiet: + print("ERROR: %s" % e, file=sys.stderr) + sys.exit(1) + except UnsupportedUsageException as e: + if args.debug: + traceback.print_tb(e.__traceback__) + if not args.quiet: + print("ERROR: %s" % e, file=sys.stderr) + sys.exit(2) + except Exception as e: + if args.debug: + traceback.print_tb(e.__traceback__) + if not args.quiet: + print("ERROR: %s" % e, file=sys.stderr) + sys.exit(3) + +if __name__ == "__main__": + main() -- 2.37.3

When doing direct kernel boot we need to include the kernel, initrd and cmdline in the measurement. Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com&gt; --- docs/manpages/virt-qemu-sev-validate.rst | 43 +++++++++ tools/virt-qemu-sev-validate | 113 ++++++++++++++++++++++- 2 files changed, 155 insertions(+), 1 deletion(-) diff --git a/docs/manpages/virt-qemu-sev-validate.rst b/docs/manpages/virt-qemu-sev-validate.rst index e2c4672a05..e8a868f5a8 100644 --- a/docs/manpages/virt-qemu-sev-validate.rst +++ b/docs/manpages/virt-qemu-sev-validate.rst @@ -102,6 +102,20 @@ initialize AMD SEV. For the validation to be trustworthy it important that the firmware build used has no support for loading non-volatile variables from NVRAM, even if NVRAM is expose to the guest. +``-k PATH``, ``--kernel=PATH`` + +Path to the kernel binary if doing direct kernel boot. + +``-r PATH``, ``--initrd=PATH`` + +Path to the initrd binary if doing direct kernel boot. Defaults to zero length +content if omitted. + +``-e STRING``, ``--cmdline=STRING`` + +String containing any kernel command line parameters used during boot of the +domain. Defaults to the empty string if omitted. + ``--tik PATH`` TIK file for domain. This file must be exactly 16 bytes in size and contains the @@ -182,6 +196,22 @@ Validate the measurement of a SEV guest booting from disk: --build-id 13 \ --policy 3 +Validate the measurement of a SEV guest with direct kernel boot: + +:: + + # virt-dom-sev-validate \ + --firmware OVMF.sev.fd \ + --kernel vmlinuz-5.11.12 \ + --initrd initramfs-5.11.12 \ + --cmdline "root=/dev/vda1" \ + --tk this-guest-tk.bin \ + --measurement Zs2pf19ubFSafpZ2WKkwquXvACx9Wt/BV+eJwQ/taO8jhyIj/F8swFrybR1fZ2ID \ + --api-major 0 \ + --api-minor 24 \ + --build-id 13 \ + --policy 3 + Fetch from remote libvirt ------------------------- @@ -202,6 +232,19 @@ Validate the measurement of a SEV guest booting from disk: --tk this-guest-tk.bin \ --domain fedora34x86_64 +Validate the measurement of a SEV guest with direct kernel boot: + +:: + + # virt-dom-sev-validate \ + --connect qemu+ssh://root@some.remote.host/system \ + --firmware OVMF.sev.fd \ + --kernel vmlinuz-5.11.12 \ + --initrd initramfs-5.11.12 \ + --cmdline "root=/dev/vda1" \ + --tk this-guest-tk.bin \ + --domain fedora34x86_64 + Fetch from local libvirt ------------------------ diff --git a/tools/virt-qemu-sev-validate b/tools/virt-qemu-sev-validate index 31c739c10f..b978c3eb3d 100755 --- a/tools/virt-qemu-sev-validate +++ b/tools/virt-qemu-sev-validate @@ -34,6 +34,7 @@ # firmware versions with known flaws. # +import abc import argparse from base64 import b64decode from hashlib import sha256 @@ -43,6 +44,7 @@ import re import socket import sys import traceback +from uuid import UUID from lxml import etree import libvirt @@ -70,6 +72,91 @@ class InvalidStateException(Exception): pass +class GUIDTable(abc.ABC): + GUID_LEN = 16 + + def __init__(self, guid, lenlen=2): + self.guid = guid + self.lenlen = lenlen + + @abc.abstractmethod + def entries(self): + pass + + def build_entry(self, guid, payload, lenlen): + dummylen = int(0).to_bytes(lenlen, 'little') + entry = bytearray(guid + dummylen + payload) + + lenle = len(entry).to_bytes(lenlen, 'little') + entry[self.GUID_LEN:(self.GUID_LEN + lenlen)] = lenle + + return bytes(entry) + + def build(self): + payload = self.entries() + + if len(payload) == 0: + return bytes([]) + + dummylen = int(0).to_bytes(self.lenlen, 'little') + table = bytearray(self.guid + dummylen + payload) + + guidlen = len(table).to_bytes(self.lenlen, 'little') + table[self.GUID_LEN:(self.GUID_LEN + self.lenlen)] = guidlen + + pad = 16 - (len(table) % 16) + table += bytes([0]) * pad + + log.debug("Table(hex): %s", bytes(table).hex()) + return bytes(table) + + +class KernelTable(GUIDTable): + + TABLE_GUID = UUID('{9438d606-4f22-4cc9-b479-a793-d411fd21}').bytes_le + KERNEL_GUID = UUID('{4de79437-abd2-427f-b835-d5b1-72d2045b}').bytes_le + INITRD_GUID = UUID('{44baf731-3a2f-4bd7-9af1-41e2-9169781d}').bytes_le + CMDLINE_GUID = UUID('{97d02dd8-bd20-4c94-aa78-e771-4d36ab2a}').bytes_le + + def __init__(self): + super().__init__(guid=self.TABLE_GUID, + lenlen=2) + + self.kernel = None + self.initrd = None + self.cmdline = None + + def load_kernel(self, path): + with open(path, "rb") as fh: + self.kernel = sha256(fh.read()).digest() + + def load_initrd(self, path): + with open(path, "rb") as fh: + self.initrd = sha256(fh.read()).digest() + + def load_cmdline(self, val): + self.cmdline = sha256(val.encode("utf8") + bytes([0])).digest() + + def entries(self): + entries = bytes([]) + if self.kernel is None: + return entries + + if self.initrd is None: + self.initrd = sha256(bytes([])).digest() + if self.cmdline is None: + self.cmdline = sha256(bytes([0])).digest() + + log.debug("Kernel(sha256): %s", self.kernel.hex()) + log.debug("Initrd(sha256): %s", self.initrd.hex()) + log.debug("Cmdline(sha256): %s", self.cmdline.hex()) + entries += self.build_entry(self.CMDLINE_GUID, self.cmdline, 2) + entries += self.build_entry(self.INITRD_GUID, self.initrd, 2) + entries += self.build_entry(self.KERNEL_GUID, self.kernel, 2) + + return entries + + class ConfidentialVM(object): def __init__(self, @@ -88,6 +175,8 @@ class ConfidentialVM(object): self.tik = None self.tek = None + self.kernel_table = KernelTable() + def load_tik_tek(self, tik_path, tek_path): with open(tik_path, 'rb') as fh: self.tik = fh.read() @@ -129,8 +218,10 @@ class ConfidentialVM(object): # of the following: # # - The firmware blob + # - The kernel GUID table def get_measured_data(self): - measured_data = self.firmware + measured_data = (self.firmware + + self.kernel_table.build()) log.debug("Measured-data(sha256): %s", sha256(measured_data).hexdigest()) return measured_data @@ -303,6 +394,12 @@ def parse_command_line(): vmconfig = parser.add_argument_group("Virtual machine config") vmconfig.add_argument('--firmware', '-f', help='Path to the firmware binary') + vmconfig.add_argument('--kernel', '-k', + help='Path to the kernel binary') + vmconfig.add_argument('--initrd', '-r', + help='Path to the initrd binary') + vmconfig.add_argument('--cmdline', '-e', + help='Cmdline string booted with') vmconfig.add_argument('--tik', help='TIK file for domain') vmconfig.add_argument('--tek', @@ -361,6 +458,11 @@ def check_usage(args): raise UnsupportedUsageException( "Either --firmware or --domain is required") + if args.kernel is None: + if args.initrd is not None or args.cmdline is not None: + raise UnsupportedUsageException( + "--initrd/--cmdline require --kernel") + def attest(args): if args.domain is None: @@ -384,6 +486,15 @@ def attest(args): else: cvm.load_tik_tek(args.tik, args.tek) + if args.kernel is not None: + cvm.kernel_table.load_kernel(args.kernel) + + if args.initrd is not None: + cvm.kernel_table.load_initrd(args.initrd) + + if args.cmdline is not None: + cvm.kernel_table.load_cmdline(args.cmdline) + if args.domain is not None: cvm.load_domain(args.connect, args.domain, -- 2.37.3