Fix ExecutionManager::GetCodeMethodDesc for interpreter

[janvorli]

The ExecutionManager::GetCodeMethodDesc expects that the address passed to it is an address of actual JIT/AOT or interpreter generated code. However, it is sometimes passed an address of the interpreter precode, e.g. in the case when called by the
MethodTable::GetMethodDescForSlotAddress. The VTable slots need to point to the interpreter stub so that these methods can be called by virtual calls from native code.

This issue occurs in several methods on the .NET runtime startup path when running interpreted.

This change adds a check for the interpreter precode to the ExecutionManager::GetCodeMethodDesc and extracts the MethodDesc from the interpreter data in case it is passed the precode address.

[dotnet-policy-service]

Tagging subscribers to this area: @BrzVlad, @janvorli, @kg
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull Request Overview

This PR enhances the runtime’s handling of interpreter precodes by enabling ExecutionManager::GetCodeMethodDesc and Precode::GetMethodDesc to resolve a MethodDesc when given an interpreter precode entry point. Key changes include:

• Added GetMethodDesc in InterpreterPrecode and wiring it into Precode::GetMethodDesc
• Introduced Precode::AsInterpreterPrecode helper for casting to InterpreterPrecode
• Updated ExecutionManager::GetCodeMethodDesc to detect interpreter precodes and return the associated MethodDesc

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/coreclr/vm/precode.h	Declare InterpreterPrecode::GetMethodDesc, PTR_InterpreterPrecode, and AsInterpreterPrecode
src/coreclr/vm/precode.cpp	Include interpreter header, implement InterpreterPrecode::GetMethodDesc, update Precode::GetMethodDesc
src/coreclr/vm/codeman.cpp	Add interpreter-precode check in ExecutionManager::GetCodeMethodDesc

Comments suppressed due to low confidence (2)

src/coreclr/vm/precode.h:339

• [nitpick] Consider adding a brief XML doc comment above GetMethodDesc() to explain that it returns the raw MethodDesc pointer stored in the interpreter precode data.

    TADDR GetMethodDesc();

src/coreclr/vm/codeman.cpp:5094

• It would be helpful to add a unit or integration test that drives ExecutionManager::GetCodeMethodDesc with an interpreter precode address to ensure this new branch correctly returns the expected MethodDesc.

    PTR_Precode pPrecode = Precode::GetPrecodeFromEntryPoint(currentPC);

[jkotas]

However, it is sometimes passed an address of the interpreter precode, e.g. in the case when called by the
MethodTable::GetMethodDescForSlotAddress

Would it be more appropriate for the fix to be in MethodTable::GetMethodDescForSlotAddress?

[jkotas]

ExecutionManager::GetCodeMethodDesc is meant to be used on arbitrary IP in the middle of a method. I think it is fragile to depend on the asm code match in Precode::GetPrecodeFromEntryPoint to be able to reject code from arbitrary IPs.

[janvorli]

ExecutionManager::GetCodeMethodDesc is meant to be used on arbitrary IP in the middle of a method.

Hmm, ok, I didn't know that. In that case, I'll move it to the MethodTable::GetMethodDescForSlotAddress. I have thought I've seen other caller ExecutionManager::GetCodeMethodDesc that could have the same problem, but checking it now, it was in the debugger code, so it is the known issue we have discussed some time ago and that I have a fix for locally.

[janvorli]

@jkotas, ok, the funny part is that with the changes I've added in precode.cpp and precode.h to get MethodDesc for InterpreterPrecode, neither the MethodTable::GetMethodDescForSlotAddress nor ExecutionManager::GetCodeMethodDesc needs to be modified. The MethodTable::GetMethodDescForSlotAddress actually gets the MethodDesc from the precode as the last step if other ones don't work.

[janvorli]

/ba-g the android build seems to be timing out on multiple CI runs