LiteBox kernel platform and machine runner by sangho2 · Pull Request #491 · microsoft/litebox

sangho2 · 2025-11-13T22:18:03Z

This PR covers an initial effort to have a minimal LiteBox kernel platform and a test runner to use the kernel platform on top of a virtual machine (i.e., QEMU, KVM) to run a hello-world OP-TEE TA.

wdcui · 2025-11-20T01:09:17Z

@jstarks @CvvT I think we should merge this minimal kernel platform into LiteBox. Can you please help take a look at this PR? Thanks!

wdcui · 2025-11-20T01:17:58Z

@sangho2 would it make sense to move the changes to the lvbs platform to a separate PR?

CvvT · 2025-11-20T01:54:28Z

@jstarks @CvvT I think we should merge this minimal kernel platform into LiteBox. Can you please help take a look at this PR? Thanks!

Why we need to merge it into LiteBox? I thought it should be a separate crate for common kernel functionality like litebox_common_linux.

wdcui · 2025-11-20T02:38:47Z

@jstarks @CvvT I think we should merge this minimal kernel platform into LiteBox. Can you please help take a look at this PR? Thanks!

Why we need to merge it into LiteBox? I thought it should be a separate crate for common kernel functionality like litebox_common_linux.

Sorry. I meant to merge it into the LiteBox repo. :-)

sangho2 · 2025-11-20T03:19:49Z

@sangho2 would it make sense to move the changes to the lvbs platform to a separate PR?

The are simple bug fixes but I agree that a separate PR makes more sense.

wdcui

Thank you for creating this minimal kernel platform, Sangho. I left some minor comments. Please wait for comments from John and Weiteng.

litebox_platform_kernel/src/arch/x86/interrupts.rs

litebox_platform_kernel/src/arch/x86/ioport.rs

litebox_platform_kernel/src/lib.rs

litebox_platform_kernel/src/per_cpu_variables.rs

litebox_platform_kernel/src/ptr.rs

litebox_shim_optee/Cargo.toml

litebox_shim_optee/src/loader/elf.rs

github-actions · 2025-11-20T23:10:43Z

🤖 SemverChecks 🤖 No breaking API changes detected

Note: this does not mean API is unchanged, or even that there are no breaking changes; simply, none of the detections triggered.

CvvT

~~A lot of the code seem to be duplicated from other platforms, which make this PR difficult to review. I can quickly draft a PR to move some common code into a separate crate.~~ Will try to clean up the code later.

litebox_platform_kernel/src/syscall_entry.rs

CvvT · 2025-11-21T23:38:22Z

litebox_platform_kernel/src/syscall_entry.rs

+#[unsafe(naked)]
+unsafe extern "C" fn syscall_entry_wrapper() {
+    naked_asm!(
+        "swapgs",


We may also need to swap stack?

Yes, definitely.

syscall is pretty tricky to get right in the general case (e.g., what happens if you get an NMI before you swap GS? Does your NMI handler look at GS?). A simpler approach, which we'll take in Coconut initially, is to not enable the syscall instruction, so that a syscall causes a #UD exception. Then, in the #UD handler, you can read two bytes at RIP, and if it's a syscall instruction, you can emulate it. This is slower than syscall but is easier to reason about.

Now we are using the kernel stack. Disabling syscall sounds interesting.

litebox_platform_kernel/src/lib.rs

jstarks · 2025-11-24T22:39:07Z

litebox_platform_kernel/src/syscall_entry.rs

+        "pop r11",
+        "pop rbp",
+        "swapgs",
+        "sysretq",


Similar to the syscall side of things, I suggest we start with iret here. I like the way Linux handles this--they treat sysret as an optimization for switching to user mode when rcx == rip and r11 == rflags. It doesn't matter if they entered from an interrupt or exception or syscall, just that the context matches what would happen if they entered via a syscall.

now we are using iretq instead of sysretq.

jstarks · 2025-11-24T22:40:39Z

litebox_platform_kernel/src/ptr.rs

+/// Represent a user space pointer to a read-only object
+#[repr(C)]
+#[derive(Clone)]
+pub struct UserConstPtr<T> {


Can we use the existing common_providers implementation right away instead of cloning another copy of this?

sangho2 · 2025-12-02T22:41:03Z

Draft for now. I'll enhance it while transferring portions of it to the LVBS platform.

sangho2 · 2025-12-18T03:26:29Z

Overall Progress/Plan:

I've updated this PR to catch up some recent changes related to run_thread.
I'll port these changes to the LVBS platform.
Once the LVBS platform is ready (OP-TEE and syscall supports), I'll work on this PR again to clean it up (and hopefully support multi cores and linux shim).

sangho2 changed the title ~~PoC: LiteBox hypervisor platform and runner~~ PoC: LiteBox machine platform and runner Nov 17, 2025

sangho2 changed the title ~~PoC: LiteBox machine platform and runner~~ LiteBox machine platform and runner Nov 17, 2025

sangho2 changed the title ~~LiteBox machine platform and runner~~ WIP: LiteBox machine platform and runner Nov 17, 2025

sangho2 force-pushed the sanghle/hypervisor/bootloader branch from 7d79944 to 65341b9 Compare November 19, 2025 00:00

sangho2 changed the title ~~WIP: LiteBox machine platform and runner~~ WIP: LiteBox kernel platform and machine runner Nov 19, 2025

sangho2 force-pushed the sanghle/hypervisor/bootloader branch from a2ff438 to b77ec38 Compare November 19, 2025 18:37

sangho2 changed the title ~~WIP: LiteBox kernel platform and machine runner~~ LiteBox kernel platform and machine runner Nov 19, 2025

sangho2 marked this pull request as ready for review November 19, 2025 19:17

sangho2 marked this pull request as draft November 19, 2025 19:22

sangho2 force-pushed the sanghle/hypervisor/bootloader branch from 34085b4 to fb312be Compare November 19, 2025 19:40

sangho2 marked this pull request as ready for review November 19, 2025 19:57

wdcui requested review from CvvT and jstarks November 20, 2025 01:07

wdcui reviewed Nov 20, 2025

View reviewed changes

sangho2 force-pushed the sanghle/hypervisor/bootloader branch 2 times, most recently from a3977b4 to 62f5896 Compare November 20, 2025 05:51

sangho2 and others added 9 commits November 20, 2025 22:59

LiteBox (virtual/bare-metal) machine platform and runner

840e987

simply states save/restore

8586400

rename

bd692f0

fix shim tls handling and improve tests

70efe14

rebase

bdd2ad0

revert unneeded stack prepopulation

9b14d51

rebase

2c6ba6f

minor clean up

71c2457

support swapgs

42beb5e

Sangho Lee added 9 commits November 20, 2025 22:59

revert lvbs changes

8d7c652

ratchet

2a09dc2

crng placeholder

fe2f0f0

use own kernel stack. specify qemu_exit

75cc4fd

revise swapgs stuffs

2a3245f

addressed some comments

c6668d3

simplify Cargo.toml

9f745e1

addressed some comments

fa09547

rebase

eb4cb38

sangho2 force-pushed the sanghle/hypervisor/bootloader branch from 927a573 to eb4cb38 Compare November 20, 2025 23:01

CvvT reviewed Nov 21, 2025

View reviewed changes

CvvT reviewed Nov 22, 2025

View reviewed changes

jstarks reviewed Nov 24, 2025

View reviewed changes

sangho2 marked this pull request as draft December 2, 2025 22:35

jstarks mentioned this pull request Dec 16, 2025

Open sourcing tasks #559

Closed

2 tasks

revising syscall handling

f5870e0

sangho2 added the must-not-merge:prototype An experimental/proof-of-concept PR that must not be merged. label Dec 17, 2025

Sangho Lee added 2 commits December 17, 2025 19:50

pt_regs and iretq

6d17ab5

modernize run_thread

e99da23

sangho2 added must-not-merge:undergoing-restructuring Known deeper set of changes are happening on this PR before it is mergeable again and removed must-not-merge:prototype An experimental/proof-of-concept PR that must not be merged. labels Dec 18, 2025

Conversation

sangho2 commented Nov 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

wdcui commented Nov 20, 2025

Uh oh!

wdcui commented Nov 20, 2025

Uh oh!

CvvT commented Nov 20, 2025

Uh oh!

wdcui commented Nov 20, 2025

Uh oh!

sangho2 commented Nov 20, 2025

Uh oh!

wdcui left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

github-actions bot commented Nov 20, 2025

Uh oh!

CvvT left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

CvvT Nov 21, 2025

Choose a reason for hiding this comment

Uh oh!

jstarks Nov 24, 2025

Choose a reason for hiding this comment

Uh oh!

sangho2 Dec 18, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

jstarks Nov 24, 2025

Choose a reason for hiding this comment

Uh oh!

sangho2 Dec 18, 2025

Choose a reason for hiding this comment

Uh oh!

jstarks Nov 24, 2025

Choose a reason for hiding this comment

Uh oh!

sangho2 commented Dec 2, 2025

Uh oh!

sangho2 commented Dec 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

sangho2 commented Nov 13, 2025 •

edited

Loading

CvvT left a comment •

edited

Loading

sangho2 commented Dec 18, 2025 •

edited

Loading