[wasm][coreclr] Get further in the runtime initialization #116383

radekdoulik · 2025-06-06T16:18:40Z

Add corewasmhost binary to test runtime initialization

Fix linking of the host binary with coreclr and interpreter

Allow calling interpreter with arguments and handle return value

Use fake precode stubs to pass the precode type and method desc

Make alternate signal stack calls no-op on wasm

Implement CallDescrWorkerInternal on wasm

Use flat PE image loading on wasm

Do not use unsupported syscalls

Add corewasmhost binary to test runtime initialization Fix linking of the host binary with coreclr and interpreter Allow calling interpreter with arguments and handle return value Use fake precode stubs to pass the precode type and method desc Make alternate signal stack calls no-op on wasm Implement CallDescrWorkerInternal on wasm Fix PE image loading without RTR on wasm Do not use unsupported syscalls

src/coreclr/vm/corhost.cpp

dotnet-policy-service · 2025-06-08T14:22:36Z

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

Co-authored-by: Adeel Mujahid <3840695+am11@users.noreply.github.com>

src/coreclr/vm/wasm/calldescrworkerwasm.cpp

src/coreclr/vm/wasm/helpers.cpp

The jit helpers are used later unconditionally

…wip-3-squashed

Fix build

src/native/minipal/random.c

src/native/libs/System.Globalization.Native/entrypoints.c

src/coreclr/vm/corhost.cpp

src/coreclr/clrdefinitions.cmake

src/coreclr/vm/wasm/helpers.cpp

src/coreclr/minipal/Unix/doublemapping.cpp

And do not build coreclr

Disable larger part of GCToOSInterface::Initialize than before

Make InitializeFlushProcessWriteBuffers no-op on wasm

Co-authored-by: Jan Kotas <jkotas@microsoft.com>

Return false from ExecutableAllocator::IsDoubleMappingEnabled instead of making the double mapping init no-op on wasm

Co-authored-by: Jan Kotas <jkotas@microsoft.com>

Fix few asserts

Define PEIMAGE_FLAT_LAYOUT_ONLY in inc\switches.h

Replace __wasm__ where appropriate

Do not use __builtin_return_address on wasm

Add InvokeCalliStub to helpers

Update comments and remove unwanted fields

Add comment to the wasm specific call description data fields

Add portability assert in DispatchCallSimple to light up if we get called through this path

…wip-3-squashed

src/coreclr/dlls/mscoree/coreclr/CMakeLists.txt

Use FEATURE_JIT instead of not WASM check

jkotas · 2025-06-25T22:17:28Z

src/coreclr/CMakeLists.txt

@@ -39,6 +39,11 @@ if (DEFINED CLR_CMAKE_ICU_DIR)
  include_directories(${CLR_CMAKE_ICU_DIR}/include)
 endif(DEFINED CLR_CMAKE_ICU_DIR)

+if (CLR_CMAKE_TARGET_ARCH_WASM)
+    add_compile_options(-fwasm-exceptions)
+    add_link_options(-fwasm-exceptions -sEXIT_RUNTIME=1)


Suggested change

add_link_options(-fwasm-exceptions -sEXIT_RUNTIME=1)

This is duplicated in src/coreclr/hosts/corewasmrun/CMakeLists.txt. Do we need it here?

jkotas · 2025-06-25T22:19:54Z

src/coreclr/dlls/mscoree/coreclr/CMakeLists.txt

-set_property(TARGET coreclr APPEND_STRING PROPERTY LINK_DEPENDS ${EXPORTS_FILE})
+    set_property(TARGET coreclr APPEND_STRING PROPERTY LINK_FLAGS ${EXPORTS_LINKER_OPTION})
+    set_property(TARGET coreclr APPEND_STRING PROPERTY LINK_DEPENDS ${EXPORTS_FILE})
+endif(NOT CLR_CMAKE_HOST_ARCH_WASM)


Suggested change

endif(NOT CLR_CMAKE_HOST_ARCH_WASM)

set(LIB_CORDBEE cordbee_wks)

set(LIB_INTEROP interop)

set(LIB_CDAC_CONTRACT_DESCRIPTOR cdac_contract_descriptor)

endif(NOT CLR_CMAKE_HOST_ARCH_WASM)

jkotas · 2025-06-25T22:20:27Z

src/coreclr/dlls/mscoree/coreclr/CMakeLists.txt

+
+if (NOT CLR_CMAKE_TARGET_ARCH_WASM)
+    set(LIB_CORDBEE cordbee_wks)
+    set(LIB_INTEROP interop)
+    set(LIB_CDAC_CONTRACT_DESCRIPTOR cdac_contract_descriptor)
+endif (NOT CLR_CMAKE_TARGET_ARCH_WASM)


Suggested change

if (NOT CLR_CMAKE_TARGET_ARCH_WASM)

set(LIB_CORDBEE cordbee_wks)

set(LIB_INTEROP interop)

set(LIB_CDAC_CONTRACT_DESCRIPTOR cdac_contract_descriptor)

endif (NOT CLR_CMAKE_TARGET_ARCH_WASM)

Folded this into the earlier if?

jkotas · 2025-06-25T23:01:36Z

src/coreclr/gc/unix/gcenv.unix.cpp

@@ -576,7 +577,7 @@ static void* VirtualReserveInner(size_t size, size_t alignment, uint32_t flags,
        }

        pRetVal = pAlignedRetVal;
-#ifdef MADV_DONTDUMP
+#if defined(MADV_DONTDUMP) && !defined(TARGET_BROWSER)


Do you believe that the rest of this implementation is going to work on wasm?

runtimelab branch reimplemented it from scratch for wasm: https://github.com/dotnet/runtimelab/blob/feature/NativeAOT-LLVM/src/coreclr/gc/wasm/gcenv.wasm.cpp . dotnet/runtimelab#1510 describes the motivation. Should we adopt this implementation instead of trying to ifdef what's here today?

jkotas · 2025-06-25T23:03:08Z

src/coreclr/gc/unix/gcenv.unix.cpp

@@ -624,9 +625,13 @@ bool GCToOSInterface::VirtualRelease(void* address, size_t size)
 //  true if it has succeeded, false if it has failed
 static bool VirtualCommitInner(void* address, size_t size, uint16_t node, bool newMemory)
 {
+#ifndef TARGET_BROWSER


jkotas · 2025-06-25T23:26:50Z

src/coreclr/vm/threads.cpp

@@ -6047,11 +6047,13 @@ BOOL Thread::SetStackLimits(SetStackLimitScope scope)
    {
        m_CacheStackBase  = GetStackUpperBound();
        m_CacheStackLimit = GetStackLowerBound();
+#ifndef TARGET_WASM // stack can start at address 0 on wasm/emscripten and usually does


Is this still needed now that we are using emscripten_stack_get_base?

jkotas · 2025-06-25T23:32:27Z

src/coreclr/vm/callhelpers.cpp

@@ -515,6 +515,9 @@ void MethodDescCallSite::CallTargetWorker(const ARG_SLOT *pArguments, ARG_SLOT *
    CallDescrData callDescrData;

    callDescrData.pSrc = pTransitionBlock + sizeof(TransitionBlock);
+#ifdef TARGET_WASM
+    callDescrData.pTransitionBlock = (TransitionBlock*)pTransitionBlock;


I think I can live with what's in this PR, but we should avoid growing it organically. We want to have discussion about how the calling convention should actually work before doing more work on it here.

this PR vs. extra PR

This PR has a bunch of good changes and a bunch of changes that need more work. If you would like to, I can go over it with you over it and separate the good non-controversial changes and get those in, and then work on the rest separately.

jkotas · 2025-06-25T23:33:10Z

src/coreclr/vm/interpexec.cpp


 // for numeric_limits
 #include <limits>

+FCDECL1(float, JIT_ULng2Flt, uint64_t val);


Is this incorrectly resolved merge conflict?

jkotas · 2025-06-25T23:33:36Z

src/coreclr/vm/metasig.h

@@ -201,6 +201,7 @@ DEFINE_METASIG_T(SM(Exception_Obj_RefIntPtr_RetVoidPtr, C(EXCEPTION) j r(I), P(v
 #endif // FEATURE_OBJCMARSHAL
 DEFINE_METASIG(SM(Int_RetVoid, i, v))
 DEFINE_METASIG(SM(Int_RetObj, i, j))
+DEFINE_METASIG(SM(Int_RetInt, i, i))


Incorrectly resolved merge conflict?

jkotas · 2025-06-25T23:38:00Z

src/coreclr/pal/src/thread/process.cpp

@@ -2664,18 +2664,22 @@ FlushProcessWriteBuffers()
        int status = pthread_mutex_lock(&flushProcessWriteBuffersMutex);
        FATAL_ASSERT(status == 0, "Failed to lock the flushProcessWriteBuffersMutex lock");

+#ifndef TARGET_BROWSER


This is same code as what's in src/coreclr/gc/unix/gcenv.unix.cpp. It should be ifdefed same way.

(Ideally, we would move this to minipal to avoid the duplication - but that would be larger change.)

jkotas · 2025-06-26T00:30:58Z

src/coreclr/vm/peassembly.cpp

@@ -260,7 +266,11 @@ TADDR PEAssembly::GetIL(RVA il)
    CONTRACT_END;

    PEImageLayout *image = NULL;
+#ifdef PEIMAGE_FLAT_LAYOUT_ONLY


It may be better to return the flat layout from GetLoadedLayout. I suspect that we would need to ifdef a lot of calls to GetLoadedLayout otherwise.

github-actions bot added the area-VM-coreclr label Jun 6, 2025

dotnet-policy-service bot assigned radekdoulik Jun 6, 2025

radekdoulik force-pushed the wasm-coreclr-wip-3-squashed branch from c49f135 to b32ad25 Compare June 6, 2025 16:23

radekdoulik added this to the Future milestone Jun 6, 2025

am11 reviewed Jun 8, 2025

View reviewed changes

src/coreclr/vm/corhost.cpp Outdated Show resolved Hide resolved

am11 added arch-wasm WebAssembly architecture os-browser Browser variant of arch-wasm labels Jun 8, 2025

Fix condition

790712a

Co-authored-by: Adeel Mujahid <3840695+am11@users.noreply.github.com>

am11 reviewed Jun 9, 2025

View reviewed changes

src/coreclr/vm/wasm/calldescrworkerwasm.cpp Show resolved Hide resolved

am11 reviewed Jun 9, 2025

View reviewed changes

src/coreclr/vm/wasm/helpers.cpp Show resolved Hide resolved

radekdoulik added 11 commits June 9, 2025 22:06

Fix build

0036748

The jit helpers are used later unconditionally

Update after merge

788fecf

Feedback

0119876

Merge remote-tracking branch 'remotes/origin/main' into wasm-coreclr-…

9605e64

…wip-3-squashed

Update after merge

43c83fd

Fix release build offsets

70b06c5

Build Debug runtime configuration on CI, so that interpreter is enabled

51f0214

Fix build

c667e6c

Enable interpreter in all configuration on wasm

e45f726

Fix build

Merge branch 'main' into wasm-coreclr-wip-3-squashed

3470ff4

Build in Release again as we are missing debug packages

91f336c

build-analysis bot mentioned this pull request Jun 10, 2025

System.Net.Quic.Tests.MsQuicTests.WriteTests failed with "System.Net.Quic.QuicException : The connection timed out from inactivity." #105177

Open

jkotas reviewed Jun 10, 2025

View reviewed changes

src/native/minipal/random.c Outdated Show resolved Hide resolved

src/native/libs/System.Globalization.Native/entrypoints.c Show resolved Hide resolved

src/coreclr/vm/corhost.cpp Outdated Show resolved Hide resolved

janvorli reviewed Jun 10, 2025

View reviewed changes

src/coreclr/clrdefinitions.cmake Outdated Show resolved Hide resolved

jkotas reviewed Jun 18, 2025

View reviewed changes

src/coreclr/vm/wasm/helpers.cpp Outdated Show resolved Hide resolved

jkotas reviewed Jun 18, 2025

View reviewed changes

src/coreclr/minipal/Unix/doublemapping.cpp Outdated Show resolved Hide resolved

radekdoulik and others added 17 commits June 20, 2025 07:29

Use coreclr_static library for wasm

ee63503

And do not build coreclr

Feedback

ebc8f45

Disable larger part of GCToOSInterface::Initialize than before

Feedback

18e25e4

Make InitializeFlushProcessWriteBuffers no-op on wasm

Write barriers helper updated

6ca105c

Co-authored-by: Jan Kotas <jkotas@microsoft.com>

Write barriers helper updated

542d732

Co-authored-by: Jan Kotas <jkotas@microsoft.com>

Feedback

f907a4d

Return false from ExecutableAllocator::IsDoubleMappingEnabled instead of making the double mapping init no-op on wasm

Make InitJITWriteBarrierHelpers no-op on wasm

f5c3500

Co-authored-by: Jan Kotas <jkotas@microsoft.com>

Feedback

2bdc376

Fix few asserts

Feedback

c3ca64a

Define PEIMAGE_FLAT_LAYOUT_ONLY in inc\switches.h

Feedback

efd8722

Replace __wasm__ where appropriate

Feedback

fb6f747

Do not use __builtin_return_address on wasm

Feedback

13b945a

Add InvokeCalliStub to helpers

Feedback

da2b405

Update comments and remove unwanted fields

Feedback

7c912c5

Add comment to the wasm specific call description data fields

Feedback

bd16e77

Add portability assert in DispatchCallSimple to light up if we get called through this path

Merge remote-tracking branch 'remotes/origin/main' into wasm-coreclr-…

035332b

…wip-3-squashed

Fix build

defe297

build-analysis bot mentioned this pull request Jun 24, 2025

[iOS/tvOS] System.Runtime.Tests crash with signal 4 #116815

Open

radekdoulik added 2 commits June 25, 2025 14:51

Fix non-wasm build

ef5b327

Merge branch 'main' into wasm-coreclr-wip-3-squashed

a7ef899

jkotas reviewed Jun 25, 2025

View reviewed changes

src/coreclr/dlls/mscoree/coreclr/CMakeLists.txt Outdated Show resolved Hide resolved

Feedback

71643ec

Use FEATURE_JIT instead of not WASM check

jkotas reviewed Jun 25, 2025

View reviewed changes

jkotas reviewed Jun 26, 2025

View reviewed changes

[wasm][coreclr] Get further in the runtime initialization #116383

Are you sure you want to change the base?

[wasm][coreclr] Get further in the runtime initialization #116383

Conversation

radekdoulik commented Jun 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

dotnet-policy-service bot commented Jun 8, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

radekdoulik commented Jun 6, 2025 •

edited

Loading