Add platform specific servtd_info configuration for the hash tool. by agokarn · Pull Request #37 · microsoft/MigTD

agokarn · 2025-09-30T22:11:58Z

This is related to Issue - #7

Problem statement -
servtd_info.json is used by the hash tool to pre- calculate MigTD TDINFO and TDINFO_HASH. MigTD TDINFO fields such as attributes and xfam are determined by HW capability and VMM configuration (`xfam' reflect the Intel CPU extended features supported by the HW and enabled by VMM). To properly pre-calculate Mig TDINFO for a specific platform, a platform specific servtd_info configuration is needed.

Depending on the targeted platform's capability and VMM configuration, the same MigTD binary running on different platforms can produce different attributes and xfam value in MigTD TDINFO, requiring VMM to use platform specific TDINFO_HASH value when invoke TDX LM prebind API.

The platform specific servtd_info configuration might need be integrated in MigTD release pipeline and a platform specific TDINFO_HASH metadata might need to be packaged with the MigTD binary for VMM to use in prebind API.

Current solution-
Xfam value determination and security implications are still under consideration. However, the Makefile now has a way to have different configurations for different platforms and generate different hashes that can be used by the build pipeline.

Fix formatting. modified: Makefile new file: config/servtd_info.txt new file: config/servtd_info_EMR.json new file: config/servtd_info_GNR.json

Copilot

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

Copilot

Pull Request Overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

Copilot · 2025-10-01T17:32:56Z

config/servtd_info_GNR.json

+    "attributes":  "0000000000000000",
+    "xfam":  "e71e960000000000",
+    "mrconfigid":  "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",
+    "mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", 


Remove trailing whitespace after the comma on line 5.

Suggested change

"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",

"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",

Copilot · 2025-10-01T17:32:56Z

config/servtd_info_EMR.json

+    "attributes":  "0000000000000000",
+    "xfam":  "e718060000000000",
+    "mrconfigid":  "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",
+    "mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000", 


Remove trailing whitespace after the comma on line 5.

Suggested change

"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",

"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",

Copilot · 2025-10-01T17:32:56Z

config/servtd_info.txt

+    19-20     | AMX     | ✓   | ✓   | AMX state (TILECFG, TILEDATA)
+    23        | CET     | -   | ✓   | Control-flow Enforcement
+
+Info generated by VSCode extension for "Claude Sonnet 4"


[nitpick] This line appears to be metadata about how the documentation was generated rather than actual documentation content. Consider removing it or moving it to a comment format.

Suggested change

Info generated by VSCode extension for "Claude Sonnet 4"

Copilot

Pull Request Overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

Copilot · 2025-10-01T17:36:16Z

config/servtd_info.txt

@@ -0,0 +1,22 @@
+Info generated by VSCode extension for "Claude Sonnet 4"


This comment appears to be generated metadata rather than proper documentation. Consider replacing with actual project documentation or removing this line.

Suggested change

Info generated by VSCode extension for "Claude Sonnet 4"

Copilot · 2025-10-01T17:36:16Z

config/servtd_info.txt

+XFAM Values for Intel CPU Generations
+
+EMR (Emerald Rapids) CPU - Expected XFAM
+    Recommended XFAM: 0x00000006071e8000 (same as current)


The comment '(same as current)' is unclear without context about what 'current' refers to. Consider removing or clarifying what the current reference is.

Suggested change

Recommended XFAM: 0x00000006071e8000 (same as current)

Recommended XFAM: 0x00000006071e8000

davidmatson · 2025-10-01T18:22:33Z

Our current design is based on a stable hash across all machines. Should we meet to discuss? I'm not sure the platform can handle divergent hashes very well.

agokarn · 2025-10-01T18:27:41Z

Our current design is based on a stable hash across all machines. Should we meet to discuss? I'm not sure the platform can handle divergent hashes very well.

Sure. Would the Thursday call work? Or do you want me to schedule an additional time slot?

davidmatson · 2025-10-01T18:29:52Z

Our current design is based on a stable hash across all machines. Should we meet to discuss? I'm not sure the platform can handle divergent hashes very well.

Sure. Would the Thursday call work? Or do you want me to schedule an additional time slot?

I don't currently see a Thursday appointment on my calendar, but anytime that afternoon should work.

agokarn · 2025-10-01T18:52:00Z

Our current design is based on a stable hash across all machines. Should we meet to discuss? I'm not sure the platform can handle divergent hashes very well.

Sure. Would the Thursday call work? Or do you want me to schedule an additional time slot?

I don't currently see a Thursday appointment on my calendar, but anytime that afternoon should work.

The series had expired. Extended it into October

davidmatson · 2025-10-01T18:53:52Z

Our current design is based on a stable hash across all machines. Should we meet to discuss? I'm not sure the platform can handle divergent hashes very well.

Sure. Would the Thursday call work? Or do you want me to schedule an additional time slot?

I don't currently see a Thursday appointment on my calendar, but anytime that afternoon should work.

The series had expired. Extended it into October

thanks - sure, that time should work fine

bodzhang · 2025-10-02T16:15:01Z

config/servtd_info.txt

@@ -0,0 +1,22 @@
+Info generated by VSCode extension for "Claude Sonnet 4"
+Reference: https://lists.nongnu.org/archive/html/qemu-devel/2023-05/msg03255.html#:~:text=Emerald%20Rapids%20(EMR)%20is%20the,1%20reply%20via%20email%20to


According to https://cdrdv2.intel.com/v1/dl/getContent/733575, xfam is defined as below:

xfam is set by VMM when the MigTD is launched, controlling what extended CPU instruction sets can be used by MigTD. VMM needs to make sure the enabled extended CPU instruction sets are sufficient for MigTD to function. For example, if MigTD implementation requires AMX to function, but VMM did not enable it, MigTD operation will fail. To predict the expected MigTD TDINFO and pre-calculate MIGTD TDINFO_HASH, the tool needs to know what VMM will do. If VMM update to a platform changes the xfam configuration, and the pre-calculated hash value is not updated, the hash will be wrong. How should ensure the pre-calculated hash update is synchronized with xfam configuration update in VMM?

xfam is set by VMM when the MigTD is launched, controlling what extended CPU instruction sets can be used by MigTD. VMM needs to make sure the enabled extended CPU instruction sets are sufficient for MigTD to function. For example, if MigTD implementation requires AMX to function, but VMM did not enable it, MigTD operation will fail. To predict the expected MigTD TDINFO and pre-calculate MIGTD TDINFO_HASH, the tool needs to know what VMM will do. If VMM update to a platform changes the xfam configuration, and the pre-calculated hash value is not updated, the hash will be wrong. How should ensure the pre-calculated hash update is synchronized with xfam configuration update in VMM?

I think we have three options here:

Use mulitple xfams, and keep xfam as part of the MigTD hash.

Use a mulitple xfams, but set ignore_xfam to exclude it from the MigTD hash.

Use a single xfam, matching the TD's documented requirement here in source control.

(Note that both options 2 and 3 require platform changes.)

I tend to think option 3 might be the simplest - thoughts?

Discussed further with platform folks, and we're planning on adding support for specifying the desired xfam in the IGVM file, so I think we could do option 3. Thoughts?

agokarn requested a review from Copilot September 30, 2025 22:11

Add platform specific servtd_info configuration for the hash tool.

a299437

Fix formatting. modified: Makefile new file: config/servtd_info.txt new file: config/servtd_info_EMR.json new file: config/servtd_info_GNR.json

agokarn force-pushed the user/arthig/platform-specific-migtd-and-hash branch from 6f83c50 to a299437 Compare September 30, 2025 22:14

Copilot AI reviewed Sep 30, 2025

View reviewed changes

agokarn requested a review from Copilot October 1, 2025 17:32

Copilot AI reviewed Oct 1, 2025

View reviewed changes

Address PR comments.

457aae2

agokarn requested a review from Copilot October 1, 2025 17:35

Copilot AI reviewed Oct 1, 2025

View reviewed changes

agokarn requested review from CyanDevs, a-trikalinou, bhargavshah1988, bodzhang, daperoni, davidmatson, haitaohuang and ksayid October 1, 2025 17:44

bodzhang reviewed Oct 2, 2025

View reviewed changes

santiudhejej-debug approved these changes Jan 29, 2026

View reviewed changes

	"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",
	"mrowner": "000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000",

		@@ -0,0 +1,22 @@
		Info generated by VSCode extension for "Claude Sonnet 4"

	Recommended XFAM: 0x00000006071e8000 (same as current)
	Recommended XFAM: 0x00000006071e8000

		@@ -0,0 +1,22 @@
		Info generated by VSCode extension for "Claude Sonnet 4"
		Reference: https://lists.nongnu.org/archive/html/qemu-devel/2023-05/msg03255.html#:~:text=Emerald%20Rapids%20(EMR)%20is%20the,1%20reply%20via%20email%20to

Conversation

agokarn commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Oct 1, 2025

Choose a reason for hiding this comment

Uh oh!

davidmatson commented Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

agokarn commented Oct 1, 2025

Uh oh!

davidmatson commented Oct 1, 2025

Uh oh!

agokarn commented Oct 1, 2025

Uh oh!

davidmatson commented Oct 1, 2025

Uh oh!

bodzhang Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

davidmatson Oct 2, 2025

Choose a reason for hiding this comment

Uh oh!

davidmatson Oct 7, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

agokarn commented Sep 30, 2025 •

edited

Loading

davidmatson commented Oct 1, 2025 •

edited

Loading