Skip to content

feat(migtd): support tdinfo_init in MigtdMigrationInformation#825

Merged
sgrams merged 1 commit into
intel:mainfrom
MichalTarnacki:issue822
May 14, 2026
Merged

feat(migtd): support tdinfo_init in MigtdMigrationInformation#825
sgrams merged 1 commit into
intel:mainfrom
MichalTarnacki:issue822

Conversation

@MichalTarnacki
Copy link
Copy Markdown
Contributor

@MichalTarnacki MichalTarnacki commented May 6, 2026 •
edited
Loading

Per GHCI 1.5, the VMM must put migpolicy.policy_key and migpolicy.policy_svn in TDINFO.MROWNER and TDINFO.MROWNERCONFIG respectively, and may provide initMigtdData in the StartMigration request for subsequent migrations.

This change implements:

a) Allow VMM to pass initMigtdData via the StartMigration request.
MigtdMigrationInformation now has has_init_data field at offset 9
and the parser accepts variable-length payload with init data.

b) MigTD startup self-check: verify that own TDINFO.MROWNER equals
SHA384(policy signing public key) and TDINFO.MROWNERCONFIG equals
own policy_svn. Reject with InvalidPolicyError on mismatch.

c) Migration flow verification: when VMM provides initMigtdData,
verify that initMigtdData.MROWNER matches own policy signer key
hash and initMigtdData.MROWNERCONFIG <= own policy_svn before
proceeding with key exchange.

d) SPDM requester for migration uses VMM-provided initMigtdData
(when available) instead of always generating local TDINFO,
enabling correct attestation for non-initial migrations.

Closes: #822 #823

@haitaohuang
Copy link
Copy Markdown
Contributor

haitaohuang commented May 12, 2026

LGTM, thanks

@MichalTarnacki MichalTarnacki marked this pull request as ready for review May 13, 2026 09:32
@MichalTarnacki MichalTarnacki force-pushed the issue822 branch 3 times, most recently from adcbeb6 to 5b3dcba Compare May 14, 2026 06:11
@MichalTarnacki @web-flow
feat(migtd): support tdinfo_init in MigtdMigrationInformation
9fb5097 
Per GHCI 1.5, the VMM must put migpolicy.policy_key and
migpolicy.policy_svn in TDINFO.MROWNER and TDINFO.MROWNERCONFIG
respectively, and may provide initMigtdData in the StartMigration
request for subsequent migrations.

This change implements:

a) Allow VMM to pass initMigtdData via the StartMigration request.
   MigtdMigrationInformation now has has_init_data field at offset 9
   and the parser accepts variable-length payload with init data.

b) MigTD startup self-check: verify that own TDINFO.MROWNER equals
   SHA384(policy signing public key) and TDINFO.MROWNERCONFIG equals
   own policy_svn. Reject with InvalidPolicyError on mismatch.

c) Migration flow verification: when VMM provides initMigtdData,
   verify that initMigtdData.MROWNER matches own policy signer key
   hash and initMigtdData.MROWNERCONFIG <= own policy_svn before
   proceeding with key exchange.

d) SPDM requester for migration uses VMM-provided initMigtdData
   (when available) instead of always generating local TDINFO,
   enabling correct attestation for non-initial migrations.

Closes: intel#822

Signed-off-by: Michal Tarnacki <michal.tarnacki@intel.com>
Co-authored-by: GitHub Copilot <noreply@github.com>
sgrams
sgrams reviewed May 14, 2026
View reviewed changes
Comment thread src/migtd/src/migration/init_data.rs
@@ -0,0 +1,120 @@
// Copyright (c) 2025 Intel Corporation
Copy link
Copy Markdown
Contributor

@sgrams sgrams May 14, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

2026

@sgrams sgrams merged commit e8b7c5b into intel:main May 14, 2026
55 checks passed
haitaohuang added a commit to haitaohuang/MigTD that referenced this pull request May 16, 2026
@haitaohuang
fix: port PR intel#825 to Microsoft fork's cfg structure
cbd953c 
The cherry-picked Intel PR intel#825 doesn't compile on the Microsoft fork due
to two differences in the fork's module layout:

1. main.rs uses 'panic_with_guest_crash_reg_report' but the Microsoft
   fork does not import it at the file scope (unlike intel/main which
   has 'use migtd::driver::vmcall_raw::panic_with_guest_crash_reg_report'
   at the top). Add the import inside the new cfg block where it is
   referenced.

2. mig_policy::verify_init_migtd_data_policy_binding references
   crate::migration::rebinding::InitData, but on this fork the
   'rebinding' module is gated on
   all(feature = "main", feature = "policy_v2", feature = "vmcall-raw").
   Gate the function on all(feature = "main", feature = "vmcall-raw")
   (policy_v2 is implied by the surrounding 'mod v2' block) so the lib
   still compiles in spdm_attestation + virtio-vsock + policy_v2 builds
   where rebinding is absent.

Verified with cargo check on x86_64-unknown-none for feature combos:
  - main,stack-guard,vmcall-raw,policy_v2
  - main,stack-guard,vmcall-raw
  - main,stack-guard,virtio-vsock
  - main,stack-guard,spdm_attestation,virtio-vsock,policy_v2
  - main,stack-guard,spdm_attestation,vmcall-raw,policy_v2
  - main,stack-guard,test_disable_ra_and_accept_all,vmcall-raw,policy_v2

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Signed-off-by: Haitao Huang <haitaohuang@microsoft.com>
haitaohuang added a commit to haitaohuang/MigTD that referenced this pull request May 16, 2026
@haitaohuang
Skip own-TDINFO check under AzCVMEmu
a62cf9b 
Intel PR intel#825 added `mig_policy::verify_own_tdinfo()` (GHCI 1.5) which
checks at startup that the running MigTD's TDINFO matches its policy:
  td_info.mrowner       == SHA384(policy signer public key)
  td_info.mrownerconfig == policy_svn as u32 LE || [0u8; 44]

These fields are normally provisioned by the host VMM. AzCVMEmu uses a
hardcoded mock TDREPORT whose MROWNER/MROWNERCONFIG are zero, so the
check would always fail under emulation and block all policy_v2 +
mock-report tests.

Skip the call site in main.rs when the `AzCVMEmu` feature is active:

    #[cfg(all(feature = "vmcall-raw", not(feature = "AzCVMEmu")))]

`verify_own_tdinfo()` itself remains compiled under `vmcall-raw`, so
real (non-emulator) builds still enforce the GHCI 1.5 binding.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
haitaohuang added a commit to haitaohuang/MigTD that referenced this pull request May 22, 2026
@haitaohuang
Skip own-TDINFO check under AzCVMEmu
e23e849 
Intel PR intel#825 added `mig_policy::verify_own_tdinfo()` (GHCI 1.5) which
checks at startup that the running MigTD's TDINFO matches its policy:
  td_info.mrowner       == SHA384(policy signer public key)
  td_info.mrownerconfig == policy_svn as u32 LE || [0u8; 44]

These fields are normally provisioned by the host VMM. AzCVMEmu uses a
hardcoded mock TDREPORT whose MROWNER/MROWNERCONFIG are zero, so the
check would always fail under emulation and block all policy_v2 +
mock-report tests.

Skip the call site in main.rs when the `AzCVMEmu` feature is active:

    #[cfg(all(feature = "vmcall-raw", not(feature = "AzCVMEmu")))]

`verify_own_tdinfo()` itself remains compiled under `vmcall-raw`, so
real (non-emulator) builds still enforce the GHCI 1.5 binding.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
haitaohuang added a commit to haitaohuang/MigTD that referenced this pull request May 23, 2026
@haitaohuang
Skip own-TDINFO check under AzCVMEmu
bfa5361 
Intel PR intel#825 added `mig_policy::verify_own_tdinfo()` (GHCI 1.5) which
checks at startup that the running MigTD's TDINFO matches its policy:
  td_info.mrowner       == SHA384(policy signer public key)
  td_info.mrownerconfig == policy_svn as u32 LE || [0u8; 44]

These fields are normally provisioned by the host VMM. AzCVMEmu uses a
hardcoded mock TDREPORT whose MROWNER/MROWNERCONFIG are zero, so the
check would always fail under emulation and block all policy_v2 +
mock-report tests.

Skip the call site in main.rs when the `AzCVMEmu` feature is active:

    #[cfg(all(feature = "vmcall-raw", not(feature = "AzCVMEmu")))]

`verify_own_tdinfo()` itself remains compiled under `vmcall-raw`, so
real (non-emulator) builds still enforce the GHCI 1.5 binding.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
haitaohuang added a commit to haitaohuang/MigTD that referenced this pull request May 23, 2026
@haitaohuang
Skip own-TDINFO check under AzCVMEmu
0a1bd0f 
Intel PR intel#825 added `mig_policy::verify_own_tdinfo()` (GHCI 1.5) which
checks at startup that the running MigTD's TDINFO matches its policy:
  td_info.mrowner       == SHA384(policy signer public key)
  td_info.mrownerconfig == policy_svn as u32 LE || [0u8; 44]

These fields are normally provisioned by the host VMM. AzCVMEmu uses a
hardcoded mock TDREPORT whose MROWNER/MROWNERCONFIG are zero, so the
check would always fail under emulation and block all policy_v2 +
mock-report tests.

Skip the call site in main.rs when the `AzCVMEmu` feature is active:

    #[cfg(all(feature = "vmcall-raw", not(feature = "AzCVMEmu")))]

`verify_own_tdinfo()` itself remains compiled under `vmcall-raw`, so
real (non-emulator) builds still enforce the GHCI 1.5 binding.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@sgrams sgrams sgrams approved these changes

@jyao1 jyao1 Awaiting requested review from jyao1 jyao1 is a code owner

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

tdinfo_init needed in MigtdMigrationInformation

3 participants

@MichalTarnacki @haitaohuang @sgrams