Move coreclr EH second pass to native code

[janvorli]

There were some GC holes discovered caused by the fact that GC can be triggered during 2nd pass of EH in-between calls to finally handlers and catch handler. After considering options, moving the 2nd pass to native code seems the most reasonable solution.

[janvorli]

The current state is that it almost works. A catch handler is correctly invoked, but then something crashes.

[Copilot]

Pull Request Overview

This PR moves the CoreCLR exception handling second pass from managed to native code to fix GC holes that can occur when GC is triggered between finally handlers and catch handlers during the second pass of exception handling.

• Splits existing RhThrowEx/RhThrowHwEx methods into separate first-pass handler finding methods and second-pass execution
• Moves the second pass execution logic (DispatchExPass2) from managed to native code
• Removes QCALL exports for CallCatchFunclet and CallFinallyFunclet as they're now called directly from native code

Reviewed Changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
src/coreclr/vm/qcallentrypoints.cpp	Removes QCALL exports for CallCatchFunclet and CallFinallyFunclet
src/coreclr/vm/metasig.h	Updates method signatures to include handlingFrameSP and pCatchHandler out parameters
src/coreclr/vm/exinfo.h	Adds ContainsCodeOffset helper method to RhEHClause
src/coreclr/vm/exceptionhandlingqcalls.h	Removes QCALL declarations for catch/finally funclets
src/coreclr/vm/exceptionhandling.h	Adds declaration for new native DispatchExPass2 function
src/coreclr/vm/exceptionhandling.cpp	Major refactoring: converts QCALL functions to native, adds new DispatchExPass2 implementation
src/coreclr/vm/excep.cpp	Updates managed fault handling to use new method signatures
src/coreclr/vm/corelib.h	Updates method name mappings for exception handling entry points
src/coreclr/nativeaot/Runtime.Base/src/System/Runtime/ExceptionHandling.cs	Splits exception handling into separate first/second pass methods

[janvorli]

Also, please note that the collided unwind detection is not working yet, as I cannot rely on the pinvokes from the managed EH code anymore. I am working on a fix.
With the contract and EH clauses enumeration fixes based on David's feedback, many EH tests now pass until I hit one with collided unwind. I am reusing the interpreter test ran without interpreter for now.

[janvorli]

And one more thing not done yet is rethrow.

[dotnet-policy-service]

Tagging subscribers to 'arch-wasm': @lewing, @pavelsavara
See info in area-owners.md if you want to be subscribed.

[janvorli]

I am considering reducing the number of changes in the exceptionhandling.cs by a slightly different approach - keeping only the methods we had and adding an extra argument "firstPassOnly" to them. Then if that argument is set to true, in DispatchEx store the handlingFrameSP/PC, _pReversePInvokePropagationCallback/Context and the pCatchHandler into the ExInfo and return;
The native side would then take these from the ExInfo. On NativeAOT, this argument would be ignored.

The current state is that, with some additional changes that I am going to commit after I fix the offsets in StackFrameIterator for some targets, it passes all diagnostics tests and coreclr tests (except ForeignThreadExceptionTest where it ends up messing GC mode when calling managed callback from native code catch)

I also need to add proper contracts here and there.

[janvorli]

/azp run runtime-coreclr outerloop

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[janvorli]

/azp run runtime-coreclr gcstress0x3-gcstress0xc

[azure-pipelines]

Azure Pipelines successfully started running 1 pipeline(s).

[janvorli]

@jkotas, @davidwrighton the current state of this PR is final from my point of view. Can you please review it?

[davidwrighton]

I believe this is correct and will fix the gc issue. I wish this invariant could be better documented/tested.

[janvorli]

/backport to release/10.0

[github-actions]

Started backporting to release/10.0: https://github.com/dotnet/runtime/actions/runs/18106203717

[github-actions]

@janvorli backporting to "release/10.0" failed, the patch most likely resulted in conflicts:

$ git am --3way --empty=keep --ignore-whitespace --keep-non-patch changes.patch

Applying: [WIP] Move coreclr EH second pass to native code
Applying: Several fixes
Applying: Fix arm64 build and remove now useless stuff from stackwalk
Applying: Fix build break
Applying: Fix offsets, exception interception and MUSL build break
Applying: Alternative change with minimalistic managed code differences
error: sha1 information is lacking or useless (src/coreclr/System.Private.CoreLib/src/System/Runtime/ExceptionServices/AsmOffsets.cs).
error: could not build fake ancestor
hint: Use 'git am --show-current-patch=diff' to see the failed patch
hint: When you have resolved this problem, run "git am --continue".
hint: If you prefer to skip this patch, run "git am --skip" instead.
hint: To restore the original branch and stop patching, run "git am --abort".
hint: Disable this message with "git config set advice.mergeConflict false"
Patch failed at 0006 Alternative change with minimalistic managed code differences
Error: The process '/usr/bin/git' failed with exit code 128

Please backport manually!