Merge branch main into dev/zephyr_file_socket (#4531)

.github/scripts/run_qemu_arc.sh

@@ -0,0 +1,46 @@
+#!/bin/bash
+
+# THIS SCRIPT IS USED TO RUN QEMU ARC EMULATOR DIRECTLY ON CI ONLY.
+# USUALLY, you SHOULD NOT RUN IT ON YOUR LOCAL MACHINE.
+# INSTEAD, YOU SHOULD USE `west build -t run` COMMAND.
+
+# Two arguments. first is the path to the zephyr-sdk, second is the path to the zephyr elf.
+if [ "$#" -ne 2 ]; then
+    echo "Usage: $0 <path_to_zephyr_sdk> <path_to_zephyr_elf>"
+    exit 1
+fi
+
+ZEPHYR_SDK_PATH=$1
+ZEPHYR_ELF_PATH=$2
+
+if [ ! -d "$ZEPHYR_SDK_PATH" ]; then
+    echo "Error: Zephyr SDK path does not exist: $ZEPHYR_SDK_PATH"
+    exit 1
+fi
+if [ ! -f "$ZEPHYR_ELF_PATH" ]; then
+    echo "Error: Zephyr ELF file does not exist: $ZEPHYR_ELF_PATH"
+    exit 1
+fi
+
+# this command is copied from the content of build.ninja which is generated by west build.
+# please do not modify it unless synchronizing with the build.ninja file.
+$ZEPHYR_SDK_PATH/sysroots/x86_64-pokysdk-linux/usr/bin/qemu-system-arc \
+  -cpu arcem -m 8M \
+  -nographic -no-reboot -monitor none \
+  -global cpu.firq=false \
+  -global cpu.num-irqlevels=15 \
+  -global cpu.num-irq=25 \
+  -global cpu.ext-irq=20 \
+  -global cpu.freq_hz=10000000 \
+  -global cpu.timer0=true \
+  -global cpu.timer1=true \
+  -global cpu.has-mpu=true \
+  -global cpu.mpu-numreg=16 \
+  -net none \
+  -pidfile qemu.pid \
+  -chardev stdio,id=con,mux=on \
+  -serial chardev:con \
+  -mon chardev=con,mode=readline \
+  -icount shift=6,align=off,sleep=off \
+  -rtc clock=vm \
+  -kernel $ZEPHYR_ELF_PATH

.github/workflows/build_iwasm_release.yml

@@ -23,7 +23,7 @@ on:
         type: string
         required: true
       upload_url:
-        description: a semantic version number. it is required when `release` is true.
+        description: upload binary assets to the URL of release
         type: string
         required: false
       ver_num:
@@ -97,9 +97,9 @@ jobs:
       matrix:
         include:
           - build_options: $DEFAULT_BUILD_OPTIONS
-            suffix: ''
+            suffix: ""
           - build_options: $GC_EH_BUILD_OPTIONS
-            suffix: '-gc-eh'
+            suffix: "-gc-eh"
     permissions:
       contents: write # for uploading release artifacts
 
@@ -126,6 +126,30 @@ jobs:
           cmake --build build --config Release --parallel 4
         working-directory: ${{ inputs.cwd }}
 
+      - name: smoke test on non-Windows
+        if: ${{ !startsWith(inputs.runner, 'windows') }}
+        shell: bash
+        run: |
+          if [[ ! -f build/iwasm ]]; then
+            echo "iwasm binary is not found in the expected location."
+            exit 1
+          fi
+
+          build/iwasm --version
+        working-directory: ${{ inputs.cwd }}
+
+      - name: smoke test on Windows
+        if: ${{ startsWith(inputs.runner, 'windows') }}
+        shell: bash
+        run: |
+          if [[ ! -f build/Release/iwasm ]]; then
+            echo "iwasm binary is not found in the expected location."
+            exit 1
+          fi
+
+          build/Release/iwasm --version
+        working-directory: ${{ inputs.cwd }}
+
       - name: Compress the binary on Windows
         if: inputs.runner == 'windows-latest'
         run: |

.github/workflows/build_wamr_wasi_extensions.yml

@@ -0,0 +1,57 @@
+# Copyright (C) 2019 Intel Corporation.  All rights reserved.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+name: build wamr_wasi_extensions release
+
+on:
+  workflow_call:
+    inputs:
+      upload_url:
+        description: upload binary assets to the URL of release
+        type: string
+        required: false
+      ver_num:
+        description: a semantic version number. it is required when `release` is true.
+        type: string
+        required: false
+
+permissions:
+  contents: read
+
+jobs:
+  build_wamr_wasi_extensions:
+    runs-on: ${{ matrix.os }}
+    permissions:
+      contents: write # for uploading release artifacts
+    strategy:
+      matrix:
+        os: [ubuntu-22.04]
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: install-wasi-sdk-wabt
+        uses: ./.github/actions/install-wasi-sdk-wabt
+        with:
+          os: ${{ matrix.os }}
+
+      - name: Build wamr-wasi-extensions
+        run: |
+          mkdir dist
+          ./build_libs.sh $(pwd)/dist/wamr-wasi-extensions
+        working-directory: wamr-wasi-extensions
+
+      - name: Compress the binary
+        run: |
+          zip -r wamr-wasi-extensions-${{ inputs.ver_num }}.zip wamr-wasi-extensions
+        working-directory: wamr-wasi-extensions/dist
+
+      - name: Upload release zip
+        uses: actions/upload-release-asset@v1
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        with:
+          upload_url: ${{ inputs.upload_url }}
+          asset_path: wamr-wasi-extensions/dist/wamr-wasi-extensions-${{ inputs.ver_num }}.zip
+          asset_name: wamr-wasi-extensions-${{ inputs.ver_num }}.zip
+          asset_content_type: application/zip

.github/workflows/build_wamrc.yml

@@ -23,7 +23,7 @@ on:
         type: string
         required: true
       upload_url:
-        description: a semantic version number. it is required when `release` is true.
+        description: upload binary assets to the URL of release
         type: string
         required: false
       ver_num:
@@ -62,6 +62,30 @@ jobs:
           cmake --build build --config Release --parallel 4
         working-directory: wamr-compiler
 
+      - name: smoke test on non-windows
+        if: ${{ !startsWith(inputs.runner, 'windows') }}
+        shell: bash
+        run: |
+          if [[ ! -f build/wamrc ]]; then
+            echo "wamrc binary is not found in the expected location."
+            exit 1
+          fi
+
+          build/wamrc --version
+        working-directory: wamr-compiler
+
+      - name: smoke test on Windows
+        if: ${{ startsWith(inputs.runner, 'windows') }}
+        shell: bash
+        run: |
+          if [[ ! -f build/Release/wamrc ]]; then
+            echo "wamrc binary is not found in the expected location."
+            exit 1
+          fi
+
+          build/Release/wamrc --version
+        working-directory: wamr-compiler
+
       - name: Compress the binary on Windows
         if: inputs.runner == 'windows-latest' && inputs.release
         run: |

.github/workflows/codeql.yml

@@ -53,7 +53,7 @@ jobs:
 
     # Initializes the CodeQL tools for scanning.
     - name: Initialize CodeQL
-      uses: github/codeql-action/init@v3.29.0
+      uses: github/codeql-action/init@v3.29.4
       with:
         languages: ${{ matrix.language }}
 
@@ -70,7 +70,7 @@ jobs:
     - run: |
         ./.github/scripts/codeql_buildscript.sh
     - name: Perform CodeQL Analysis
-      uses: github/codeql-action/analyze@v3.29.0
+      uses: github/codeql-action/analyze@v3.29.4
       with:
         category: "/language:${{matrix.language}}"
         upload: false
@@ -99,7 +99,7 @@ jobs:
         output: ${{ steps.step1.outputs.sarif-output }}/cpp.sarif
 
     - name: Upload CodeQL results to code scanning
-      uses: github/codeql-action/upload-sarif@v3.29.0
+      uses: github/codeql-action/upload-sarif@v3.29.4
       with:
         sarif_file: ${{ steps.step1.outputs.sarif-output }}
         category: "/language:${{matrix.language}}"

.github/workflows/compilation_on_android_ubuntu.yml

@@ -69,6 +69,7 @@ env:
   GC_TEST_OPTIONS: "-s spec -G -b -P"
   MEMORY64_TEST_OPTIONS: "-s spec -W -b -P"
   MULTI_MEMORY_TEST_OPTIONS: "-s spec -E -b -P"
+  EXTENDED_CONST_EXPR_TEST_OPTIONS: "-s spec -N -b -P"
 
 permissions:
   contents: read
@@ -164,6 +165,7 @@ jobs:
             "-DWAMR_BUILD_MEMORY64=1",
             "-DWAMR_BUILD_MULTI_MEMORY=1",
             "-DWAMR_BUILD_SHARED=1",
+            "-DWAMR_BUILD_EXTENDED_CONST_EXPR=1",
           ]
         os: [ubuntu-22.04]
         platform: [android, linux]
@@ -369,6 +371,47 @@ jobs:
           ctest
         working-directory: tests/unit
 
+  build_regression_tests:
+    needs:
+      [build_llvm_libraries_on_ubuntu_2204]
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-22.04]
+        include:
+          - os: ubuntu-22.04
+            llvm_cache_key: ${{ needs.build_llvm_libraries_on_ubuntu_2204.outputs.cache_key }}
+    
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: Get LLVM libraries
+        id: retrieve_llvm_libs
+        uses: actions/cache@v4
+        with:
+          path: |
+            ./core/deps/llvm/build/bin
+            ./core/deps/llvm/build/include
+            ./core/deps/llvm/build/lib
+            ./core/deps/llvm/build/libexec
+            ./core/deps/llvm/build/share
+          key: ${{ matrix.llvm_cache_key }}
+
+      - name: Quit if cache miss
+        if: (steps.retrieve_llvm_libs.outputs.cache-hit != 'true')
+        run: echo "::error::can not get prebuilt llvm libraries" && exit 1
+      
+      - name: Build wamrc and iwasm
+        run: |
+          ./build_wamr.sh
+        working-directory: tests/regression/ba-issues
+      
+      - name: Run regression tests
+        run: |
+          python run.py
+        working-directory: tests/regression/ba-issues
+
   build_samples_wasm_c_api:
     needs:
       [
@@ -609,6 +652,7 @@ jobs:
             $GC_TEST_OPTIONS,
             $MEMORY64_TEST_OPTIONS,
             $MULTI_MEMORY_TEST_OPTIONS,
+            $EXTENDED_CONST_EXPR_TEST_OPTIONS,
           ]
         include:
           - os: ubuntu-22.04

.github/workflows/compilation_on_macos.yml

@@ -142,6 +142,7 @@ jobs:
             "-DWAMR_BUILD_SIMD=1",
             "-DWAMR_BUILD_TAIL_CALL=1",
             "-DWAMR_DISABLE_HW_BOUND_CHECK=1",
+            "-DWAMR_BUILD_EXTENDED_CONST_EXPR=1",
           ]
         os: [macos-13]
         platform: [darwin]

.github/workflows/compilation_on_sgx.yml

@@ -100,6 +100,7 @@ jobs:
             "-DWAMR_BUILD_MULTI_MODULE=1",
             "-DWAMR_BUILD_PERF_PROFILING=1",
             "-DWAMR_BUILD_REF_TYPES=1",
+            "-DWAMR_BUILD_EXTENDED_CONST_EXPR=1",
             # doesn't support
             "-DWAMR_BUILD_SIMD=0",
             "-DWAMR_BUILD_TAIL_CALL=1",

.github/workflows/compilation_on_zephyr.yml

@@ -0,0 +1,129 @@
+# Copyright (C) 2019 Intel Corporation.  All rights reserved.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+name: compilation on zephyr
+
+on:
+  # will be triggered on PR events
+  pull_request:
+    types:
+      - opened
+      - synchronize
+    paths:
+      - ".github/**"
+      - "build-scripts/**"
+      - "core/**"
+      - "!core/deps/**"
+      - "product-mini/platforms/common/**"
+      - "product-mini/platforms/zephyr/**"
+      - "samples/**"
+      - "!samples/workload/**"
+      - "tests/wamr-test-suites/**"
+      - "wamr-compiler/**"
+  # will be triggered on push events
+  push:
+    branches:
+      - main
+      - "dev/**"
+    paths:
+      - ".github/**"
+      - "build-scripts/**"
+      - "core/**"
+      - "!core/deps/**"
+      - "product-mini/platforms/common/**"
+      - "product-mini/platforms/zephyr/**"
+      - "samples/**"
+      - "!samples/workload/**"
+      - "tests/wamr-test-suites/**"
+      - "wamr-compiler/**"
+  # allow to be triggered manually
+  workflow_dispatch:
+
+# Cancel any in-flight jobs for the same PR/branch so there's only one active
+# at a time
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+env:
+  # FOR SETUP
+  ZEPHYR_SDK_VERSION: "0.16.9"
+  ZEPHYR_VERSION: "v3.7.0"
+  # FOR BUILD
+
+permissions:
+  contents: read
+
+jobs:
+  smoke_test:
+    runs-on: ubuntu-22.04
+    container:
+      # For Zephyr 3.7 LTS, use the v0.26-branch or the latest v0.26.x release Docker image.
+      # ci require a larger runner to avoid "no space left on device"
+      image: ghcr.io/zephyrproject-rtos/ci-base:v0.26-branch
+      options: --user root
+
+    steps:
+      # https://docs.zephyrproject.org/latest/develop/application/index.html#zephyr-workspace-application
+      # zephyrproject/       --> CI ROOT
+      # ├─── .west/
+      # │    └─── config
+      # ├─── bootloader/
+      # ├─── zephyr/         --> Zephyr source code
+      # ├─── zephyr-sdk/
+      # ├─── modules/
+      # │    |─── wasm-micro-runtime   --> WAMR source code
+      # ├─── tools/
+      # ├─── vendor/
+      # └─── application/    --> DUMMY. keep west_lite.yml here
+
+      - name: Checkout code
+        uses: actions/checkout@v3
+        with:
+          path: modules/wasm-micro-runtime
+
+      - name: Prepare Zephyr environment
+        shell: bash
+        run: |
+          mkdir -p application
+          cp modules/wasm-micro-runtime/product-mini/platforms/zephyr/simple/west_lite.yml application/west_lite.yml
+
+      - name: Setup Zephyr project
+        uses: zephyrproject-rtos/action-zephyr-setup@v1
+        with:
+          app-path: application
+          manifest-file-name: west_lite.yml
+          sdk-version: ${{ env.ZEPHYR_SDK_VERSION }}
+          toolchains: arc-zephyr-elf:arc64-zephyr-elf
+
+      - name: Build a sample application(simple)
+        shell: bash
+        run: |
+          pushd product-mini/platforms/zephyr/simple
+          west build . -b qemu_arc/qemu_arc_hs  -p always -- -DWAMR_BUILD_TARGET=ARC
+          popd
+
+          # west build -t run will fork several processes, which will cause the job to hang.
+          # run in the background and kill it after 5 seconds
+          .github/scripts/run_qemu_arc.sh \
+            ../../zephyr-sdk \
+            product-mini/platforms/zephyr/simple/build/zephyr/zephyr.elf &
+          sleep 5
+          pkill qemu-system-arc
+        working-directory: modules/wasm-micro-runtime
+
+      - name: Build a sample application(user-mode)
+        shell: bash
+        run: |
+          pushd product-mini/platforms/zephyr/user-mode
+          west build . -b qemu_arc/qemu_arc_hs -p always -- -DWAMR_BUILD_TARGET=ARC
+          popd
+
+          # west build -t run will fork several processes, which will cause the job to hang.
+          # run in the background and kill it after 5 seconds
+          .github/scripts/run_qemu_arc.sh \
+            ../../zephyr-sdk \
+            product-mini/platforms/zephyr/user-mode/build/zephyr/zephyr.elf &
+          sleep 5
+          pkill qemu-system-arc
+        working-directory: modules/wasm-micro-runtime

.github/workflows/nightly_run.yml

@@ -37,6 +37,7 @@ env:
   MULTI_TIER_JIT_BUILD_OPTIONS: "-DWAMR_BUILD_AOT=1 -DWAMR_BUILD_FAST_INTERP=0 -DWAMR_BUILD_INTERP=1 -DWAMR_BUILD_FAST_JIT=1 -DWAMR_BUILD_JIT=1 -DWAMR_BUILD_LAZY_JIT=1"
   # For Spec Test
   DEFAULT_TEST_OPTIONS: "-s spec -b -P"
+  EXTENDED_CONST_EXPR_TEST_OPTIONS: "-s spec -b -P -N"
   MULTI_MODULES_TEST_OPTIONS: "-s spec -b -P -M"
   SIMD_TEST_OPTIONS: "-s spec -b -P -S"
   THREADS_TEST_OPTIONS: "-s spec -b -P -p"
@@ -128,6 +129,7 @@ jobs:
             "-DWAMR_BUILD_MEMORY64=1",
             "-DWAMR_BUILD_MULTI_MEMORY=1",
             "-DWAMR_BUILD_SHARED=1",
+            "-DWAMR_BUILD_EXTENDED_CONST_EXPR=1",
           ]
         os: [ubuntu-22.04]
         platform: [android, linux]
@@ -588,6 +590,7 @@ jobs:
             $DEFAULT_TEST_OPTIONS,
             $MULTI_MODULES_TEST_OPTIONS,
             $SIMD_TEST_OPTIONS,
+            $EXTENDED_CONST_EXPR_TEST_OPTIONS,
             $THREADS_TEST_OPTIONS,
             $WASI_TEST_OPTIONS,
           ]

.github/workflows/release_process.yml

@@ -239,3 +239,12 @@ jobs:
       arch: universal
       upload_url: ${{ needs.create_release.outputs.upload_url }}
       ver_num: ${{ needs.create_tag.outputs.new_ver}}
+
+  release_wamr_wasi_extensions:
+    permissions:
+      contents: write # upload release artifact
+    needs: [create_tag, create_release]
+    uses: ./.github/workflows/build_wamr_wasi_extensions.yml
+    with:
+      upload_url: ${{ needs.create_release.outputs.upload_url }}
+      ver_num: ${{ needs.create_tag.outputs.new_ver }}

.github/workflows/supply_chain.yml

@@ -60,6 +60,6 @@ jobs:
 
       # Upload the results to GitHub's code scanning dashboard.
       - name: "Upload to code-scanning"
-        uses: github/codeql-action/upload-sarif@2847b7f7ab9f48fc49eca90a53fff6007285f399 
+        uses: github/codeql-action/upload-sarif@701df0e49d84a24bd8f0d01f80c0dbf69ab07674 
         with:
           sarif_file: results.sarif

.github/workflows/wamr_wasi_extensions.yml

@@ -0,0 +1,57 @@
+# Copyright (C) 2019 Intel Corporation.  All rights reserved.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+name: wamr_wasi_extensions
+
+on:
+  pull_request:
+    types:
+      - opened
+      - synchronize
+    paths:
+      - ".github/workflows/wamr_wasi_extensions.yml"
+      - "wamr_wasi_extensios/**"
+      - "core/iwasm/libraries/wasi-nn/include/**"
+      - "core/iwasm/libraries/lib-socket/**"
+  # allow to be triggered manually
+  workflow_dispatch:
+
+# Cancel any in-flight jobs for the same PR/branch so there's only one active
+# at a time
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  build_wamr_wasi_extensions:
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-22.04, macos-13, macos-14]
+    steps:
+      - name: checkout
+        uses: actions/checkout@v4
+
+      - name: install-wasi-sdk-wabt
+        uses: ./.github/actions/install-wasi-sdk-wabt
+        with:
+          os: ${{ matrix.os }}
+
+      - name: Build wamr-wasi-extensions
+        run: |
+          mkdir dist
+          ./build_libs.sh $(pwd)/dist/wamr-wasi-extensions
+        working-directory: wamr-wasi-extensions
+
+      - name: Build wamr-wasi-extensions samples
+        run: |
+          ./build_samples.sh $(pwd)/dist/wamr-wasi-extensions
+        working-directory: wamr-wasi-extensions
+
+      - name: Upload artifacts
+        if: matrix.os == 'macos-14'
+        uses: actions/upload-artifact@v4
+        with:
+          name: wamr-wasi-extensions
+          path: wamr-wasi-extensions/dist
+          retention-days: 10

CMakeLists.txt

@@ -99,9 +99,9 @@ if (NOT DEFINED WAMR_BUILD_LIB_WASI_THREADS)
   set (WAMR_BUILD_LIB_WASI_THREADS 0)
 endif ()
 
-if (NOT DEFINED WAMR_ENABLE_COPY_CALLSTACK)
+if (NOT DEFINED WAMR_BUILD_COPY_CALL_STACK)
   # Disable copy callstack by default
-  set (WAMR_ENABLE_COPY_CALLSTACK 0)
+  set (WAMR_BUILD_COPY_CALL_STACK 0)
 endif()
 
 if (NOT DEFINED WAMR_BUILD_MINI_LOADER)

README.md

@@ -48,6 +48,7 @@ WebAssembly Micro Runtime (WAMR) is a lightweight standalone WebAssembly (Wasm)
 - [Reference Types](https://github.com/WebAssembly/reference-types), ref to [document](doc/ref_types.md) and [sample](samples/ref-types)
 - [Bulk memory operations](https://github.com/WebAssembly/bulk-memory-operations), [Shared memory](https://github.com/WebAssembly/threads/blob/main/proposals/threads/Overview.md#shared-linear-memory), [Memory64](https://github.com/WebAssembly/memory64)
 - [Tail-call](https://github.com/WebAssembly/tail-call), [Garbage Collection](https://github.com/WebAssembly/gc), [Exception Handling](https://github.com/WebAssembly/exception-handling)
+- [Extended Constant Expressions](https://github.com/WebAssembly/extended-const)
 
 ### Supported architectures and platforms
 The WAMR VMcore supports the following architectures:

RELEASE_NOTES.md

@@ -1,3 +1,157 @@
+## WAMR-2.4.1
+
+### Breaking Changes
+
+- wasi_socket_ext.c: fix error reporting (#4476)
+- lib-socket: make getaddrinfo return EAI\_ values (#4498)
+- bump AOT_CURRENT_VERSION for extended-const (#4511)
+
+### New Features
+
+### Bug Fixes
+
+- modify macro related to simde when WASM_OP_SELECT_128 (#4461)
+- posix os_socket_accept: stop assuming socklen_t is unsigned int (#4488)
+- wasi_socket_ext.c: fix the number of getaddrinfo results (#4466)
+- Fix typos (#4472)
+- fix regression running_config.json (#4477)
+- posix os_socket_addr_resolve: relax compatibility check (#4469)
+- Add validation for recursive type count in loader (#4440)
+
+### Enhancements
+
+- Add CI on Zephyr (#4336)
+- introduce wasm_runtime_instantiate_ex2 (#4444)
+- Add a CLI option to specify shared heap size on Windows platform (#4503)
+- wasi-nn: add a comment on load_by_name_with_config (#4492)
+- nn-cli: add an option to use load_by_name (#4490)
+- wamr-wasi-extensions: document (#4493)
+- doc/socket_api.md: some historical notes (#4494)
+- lib-socket: implement getsockopt(SOL_SOCKET,SO_TYPE) (#4458)
+
+### Others
+
+- build(deps): Bump github/codeql-action from 3.29.2 to 3.29.3 (#4507)
+
+## WAMR-2.4.0
+
+### Breaking Changes
+
+- Refactor copy callstack feature (#4401)
+- Enable WAMR_BUILD_WASI_EPHEMERAL_NN by default (#4381)
+- Enable aot memory64 sw bounds checks by default (#4350)
+
+### New Features
+
+- Support extended constant expressions (#4432)
+- Shared heap enhancements for Interpreter and AOT (#4400)
+
+### Bug Fixes
+
+- posix os_socket_addr_resolve: return the consistent max_info_size (#4467)
+- fix a wamrc debug mode compile issue (#4470)
+- wasi-nn: do not pretend to support legacy abi in openvino and llamacpp (#4468)
+- appease a few compiler warnings (-Wstrict-prototypes) (#4465)
+- enable aux stack frame for aot compiler fuzz test (#4462)
+- improve logic of `heap_type` validation when `ref.null` (#4372)
+- wasi_nn_llamacpp.c: explicitly reject unimplemented input index (#4446)
+- wasi: avoid user-triggerable 0-sized allocations (#4452)
+- Fix socket shutdown (#12) (#4449)
+- wasi_nn_llamacpp.c: validate input tensor type/dimensions (#4442)
+- wasi_nn_llamacpp.c: reject invalid graph and execution context (#4422)
+- wasi_nn_openvino.c: avoid self-assignment warning (#4434)
+- Fix potential integer overflow issues (#4429)
+- Improve run.py of regression (#4417)
+- wasi-nn: reduce code duplication a bit (#4433)
+- Refactor AOTObjectData definition to use a forward declaration (#4428)
+- CI: revert SGX retry attempts (#4421)
+- loader: fix a potential overflow issue (#4427)
+- wasi_nn_openvino.c: fix a debug build (#4416)
+- Fix few shadow warnings (#4409)
+- wasi_nn_llamacpp.c: remove an unused variable (#4415)
+- wasi_nn_llamacpp.c: fix buffer overruns in set_input (#4420)
+- wasi-nn: make the host use the wasi_ephemeral_nn version of tensor_data (#4411)
+- Collective fix (#4413)
+- fix bug in bh_vector when extending (#4414)
+- wasi_nn_llamacpp.c: make this compilable (#4403)
+- Fix handling of non-nullable global_type during global import (#4408)
+- loader: add type index checking (#4402)
+- wasi_nn_tensorflowlite.cpp: fix get_output return size (#4390)
+- wasi-nn: fix context lifetime issues (#4396)
+- CI: fix the description of upload_url (#4407)
+- wamr-wasi-extensions/socket: disable reference-types (#4392)
+- wasi_nn_openvino.c: implement multiple models per instance (#4380)
+- Improve spec test execution by adding retry logic for transient errors (#4393)
+- wasi-nn: add minimum serialization on WASINNContext (#4387)
+- deprecate legacy WAMR-specific "wasi_nn" module (#4382)
+- wasi-nn: fix tensor_data abi for wasi_ephemeral_nn (#4379)
+- core/iwasm/libraries/wasi-nn/test: use the correct version of keras (#4383)
+- Fix several issues related to night-run CI and test scripts. (#4385)
+- wasi_nn_tensorflowlite.cpp: reject non-fp32 input earlier (#4388)
+- core/iwasm/libraries/wasi-nn/test/build.sh: add a tip for intel mac (#4389)
+- wasi-nn: don't try to deinit uninitialized backend (#4375)
+- wasi-nn: apply the shared library hack to darwin as well (#4374)
+- add nn-cli example (#4373)
+- wasi_nn_openvino.c: remove pre/postprocessing and layout assumptions (#4361)
+- send an empty/error reply from server (#4362)
+- wasi_nn_openvino.c: add a missing buffer overflow check in get_output (#4353)
+- wasi_ephemeral_nn.h: prefix identfiers to avoid too generic names (#4358)
+- wamr-wasi-extensions: add lib-socket things (#4360)
+- wasi_nn_openvino.c: remove broken xml check (#4365)
+- add validation for array type in load_init_expr(GC only) (#4370)
+- wasi-nn: fix backend leak on multiple loads (#4366)
+- Collective fix for typos and minor bugs (#4369)
+- Modify AOT static PGO to conform to llvm-18 and add a CI job to test static PGO on the coremark benchmark (#4345)
+- Update WABT downloads URL (#4357)
+- clean up incompatible running mode checks in test script and ci (#4342)
+- Follow #4268 to deprecate wamr_ide-related components (#4341)
+- Update type validation in load_table_import() and load_table() (#4296)
+- wasi_nn_openvino.c: remove the tensor layout adjustment logic (#4308)
+- add heap-type check for GC when ref.null (#4300)
+- wasi_nn_types.h: remove a seemingly stale comment (#4348)
+- wasi_socket_ext.c: avoid tls to make this library-friendly (#4338)
+- wasi-nn: do not assign wasi_nn_ctx->backend multiple times (#4329)
+- wasi_nn.h: make this compatible with wasi_ephemeral_nn (#4330)
+- remove temporary wasi-libc build steps from CI workflows (#4343)
+- wasi-nn: fix the size of tensor->type (#4333)
+- wasi-nn: move some host-only things out of wasi_nn_types.h (#4334)
+- Collective fix: fix some typos (#4337)
+- Update binary compression steps to follow symlinks for actual files (#4321)
+- Add wamrc compilation into Windows CI workflow (#4327)
+- wasi-nn: remove unused wasi_nn_dump_tensor_dimension prototype (#4325)
+- wasi_nn.h: add import_name attribute (#4328)
+- wasi-nn: protect the backend lookup table with a lock (#4319)
+- handle nullable heap reference types in import section (#4302)
+- wasi_nn_openvino.c: make this buildable (#4305)
+- wasi-nn: fix shared library filenames for macOS (#4306)
+- fix wasi-nn abi definitions (#4307)
+- wasi-nn: remove "backends" argument from detect_and_load_backend() (#4309)
+- wasi_nn_openvino.c: fix a few printf formats (#4310)
+- Bump uvwasi to latest commit #392e1f1 (#4312)
+
+### Enhancements
+
+- Add readme for extended const (#4471)
+- Add security issue runbook (#4450)
+- docs: fix cmake variable typo (#4441)
+- CI: add wamr_wasi_extensions to the release assets (#4425)
+- CI: build wamr-wasi-extensions (#4394)
+- improve installation steps for wasi-sdk and wabt on Windows (#4359)
+- wamr-wasi-extensions: add a cmake package to provide our wasi extension (#4344)
+- Update Dockerfile for Zephyr SDK and Zephyr-project versioning (#4335)
+- add load_by_name in wasi-nn (#4298)
+
+### Others
+
+- build(deps): Bump ossf/scorecard-action from 2.4.1 to 2.4.2 (#4315)
+- build(deps): Bump github/codeql-action from 3.29.1 to 3.29.2 (#4459)
+- build(deps): Bump github/codeql-action from 3.29.0 to 3.29.1 (#4436)
+- build(deps): Bump github/codeql-action from 3.28.19 to 3.29.0 (#4371)
+- build(deps): Bump github/codeql-action from 3.28.18 to 3.28.19 (#4346)
+- build(deps): Bump requests from 2.32.3 to 2.32.4 in /build-scripts (#4349)
+
+---
+
 ## WAMR-2.3.1
 
 ### Breaking Changes

build-scripts/config_common.cmake

@@ -211,6 +211,10 @@ if (NOT DEFINED WAMR_BUILD_TAIL_CALL)
   set (WAMR_BUILD_TAIL_CALL 0)
 endif ()
 
+if (NOT DEFINED WAMR_BUILD_EXTENDED_CONST_EXPR)
+  set (WAMR_BUILD_EXTENDED_CONST_EXPR 0)
+endif ()
+
 ########################################
 # Compilation options to marco
 ########################################
@@ -334,15 +338,10 @@ if (WAMR_BUILD_SHARED_HEAP EQUAL 1)
   add_definitions (-DWASM_ENABLE_SHARED_HEAP=1)
   message ("     Shared heap enabled")
 endif()
-
+if (WAMR_BUILD_COPY_CALL_STACK EQUAL 1)
-if (WAMR_ENABLE_COPY_CALLSTACK EQUAL 1)
+  add_definitions (-DWASM_ENABLE_COPY_CALL_STACK=1)
-  add_definitions (-DWAMR_ENABLE_COPY_CALLSTACK=1)
   message("     Copy callstack enabled")
-else ()
-  add_definitions (-DWAMR_ENABLE_COPY_CALLSTACK=0)
-  message("     Copy callstack disabled")
 endif()
-
 if (WAMR_BUILD_MEMORY64 EQUAL 1)
   # if native is 32-bit or cross-compiled to 32-bit
   if (NOT WAMR_BUILD_TARGET MATCHES ".*64.*")
@@ -539,6 +538,9 @@ if (WAMR_BUILD_WASI_NN EQUAL 1)
   if (DEFINED WAMR_BUILD_WASI_NN_EXTERNAL_DELEGATE_PATH)
       add_definitions (-DWASM_WASI_NN_EXTERNAL_DELEGATE_PATH="${WAMR_BUILD_WASI_NN_EXTERNAL_DELEGATE_PATH}")
   endif ()
+  if (NOT DEFINED WAMR_BUILD_WASI_EPHEMERAL_NN)
+      set(WAMR_BUILD_WASI_EPHEMERAL_NN 1)
+  endif()
   if (WAMR_BUILD_WASI_EPHEMERAL_NN EQUAL 1)
       message ("     WASI-NN: use 'wasi_ephemeral_nn' instead of 'wasi-nn'")
       add_definitions (-DWASM_ENABLE_WASI_EPHEMERAL_NN=1)
@@ -675,7 +677,13 @@ if (WAMR_BUILD_INSTRUCTION_METERING EQUAL 1)
   message ("     Instruction metering enabled")
   add_definitions (-DWASM_ENABLE_INSTRUCTION_METERING=1)
 endif ()
-
+if (WAMR_BUILD_EXTENDED_CONST_EXPR EQUAL 1)
+  message ("     Extended constant expression enabled")
+  add_definitions(-DWASM_ENABLE_EXTENDED_CONST_EXPR=1)
+else()
+  message ("     Extended constant expression disabled")
+  add_definitions(-DWASM_ENABLE_EXTENDED_CONST_EXPR=0)
+endif ()
 ########################################
 # Show Phase4 Wasm proposals status.
 ########################################
@@ -689,6 +697,7 @@ message (
 "       \"WebAssembly C and C++ API\"\n"
 "     Configurable. 0 is OFF. 1 is ON:\n"
 "       \"Bulk Memory Operation\" via WAMR_BUILD_BULK_MEMORY: ${WAMR_BUILD_BULK_MEMORY}\n"
+"       \"Extended Constant Expressions\" via WAMR_BUILD_EXTENDED_CONST_EXPR: ${WAMR_BUILD_EXTENDED_CONST_EXPR}\n"
 "       \"Fixed-width SIMD\" via WAMR_BUILD_SIMD: ${WAMR_BUILD_SIMD}\n"
 "       \"Garbage collection\" via WAMR_BUILD_GC: ${WAMR_BUILD_GC}\n"
 "       \"Legacy Exception handling\" via WAMR_BUILD_EXCE_HANDLING: ${WAMR_BUILD_EXCE_HANDLING}\n"
@@ -703,7 +712,6 @@ message (
 "       \"Branch Hinting\"\n"
 "       \"Custom Annotation Syntax in the Text Format\"\n"
 "       \"Exception handling\"\n"
-"       \"Extended Constant Expressions\"\n"
 "       \"Import/Export of Mutable Globals\"\n"
 "       \"JS String Builtins\"\n"
 "       \"Relaxed SIMD\"\n"

build-scripts/runtime_lib.cmake

@@ -106,6 +106,7 @@ endif ()
 
 if (WAMR_BUILD_WASI_NN EQUAL 1)
     include (${IWASM_DIR}/libraries/wasi-nn/cmake/wasi_nn.cmake)
+    set (WAMR_BUILD_MODULE_INST_CONTEXT 1)
 endif ()
 
 if (WAMR_BUILD_LIB_PTHREAD EQUAL 1)

build-scripts/version.cmake

@@ -7,7 +7,7 @@ if(NOT WAMR_ROOT_DIR)
 endif()
 
 set(WAMR_VERSION_MAJOR 2)
-set(WAMR_VERSION_MINOR 3)
+set(WAMR_VERSION_MINOR 4)
 set(WAMR_VERSION_PATCH 1)
 
 message("-- WAMR version: ${WAMR_VERSION_MAJOR}.${WAMR_VERSION_MINOR}.${WAMR_VERSION_PATCH}")

ci/coding_guidelines_check.py

@@ -4,7 +4,6 @@
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 #
 import argparse
-import re
 from pathlib import Path
 import re
 import shlex
@@ -39,7 +38,7 @@ INVALID_FILE_NAME_SEGMENT = r"([a-zA-Z0-9]+\-[a-zA-Z0-9]+)"
 
 def locate_command(command: str) -> bool:
     if not shutil.which(command):
-        print(f"Command '{command}'' not found")
+        print(f"Command '{command}' not found")
         return False
 
     return True

core/config.h

@@ -84,7 +84,7 @@
 #endif
 
 #define AOT_MAGIC_NUMBER 0x746f6100
-#define AOT_CURRENT_VERSION 4
+#define AOT_CURRENT_VERSION 5
 
 #ifndef WASM_ENABLE_JIT
 #define WASM_ENABLE_JIT 0
@@ -193,8 +193,8 @@
 #error "Heap aux stack allocation must be enabled for WASI threads"
 #endif
 
-#ifndef WAMR_ENABLE_COPY_CALLSTACK
+#ifndef WASM_ENABLE_COPY_CALL_STACK
-#define WAMR_ENABLE_COPY_CALLSTACK 0
+#define WASM_ENABLE_COPY_CALL_STACK 0
 #endif
 
 #ifndef WASM_ENABLE_BASE_LIB
@@ -720,4 +720,8 @@ unless used elsewhere */
 #define WASM_ENABLE_INSTRUCTION_METERING 0
 #endif
 
+#ifndef WASM_ENABLE_EXTENDED_CONST_EXPR
+#define WASM_ENABLE_EXTENDED_CONST_EXPR 0
+#endif
+
 #endif /* end of _CONFIG_H_ */

core/iwasm/aot/aot_loader.c

@@ -968,6 +968,35 @@ fail:
     return false;
 }
 
+#if WASM_ENABLE_GC != 0 || WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+static void
+destroy_init_expr(InitializerExpression *expr)
+{
+#if WASM_ENABLE_GC != 0
+    if (expr->init_expr_type == INIT_EXPR_TYPE_STRUCT_NEW
+        || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW
+        || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW_FIXED) {
+        wasm_runtime_free(expr->u.unary.v.data);
+    }
+#endif
+
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    // free left expr and right expr for binary oprand
+    if (!is_expr_binary_op(expr->init_expr_type)) {
+        return;
+    }
+    if (expr->u.binary.l_expr) {
+        destroy_init_expr_recursive(expr->u.binary.l_expr);
+    }
+    if (expr->u.binary.r_expr) {
+        destroy_init_expr_recursive(expr->u.binary.r_expr);
+    }
+    expr->u.binary.l_expr = expr->u.binary.r_expr = NULL;
+#endif
+}
+#endif /* end of WASM_ENABLE_GC != 0 || WASM_ENABLE_EXTENDED_CONST_EXPR != 0 \
+        */
+
 static void
 destroy_import_memories(AOTImportMemory *import_memories)
 {
@@ -993,6 +1022,10 @@ destroy_mem_init_data_list(AOTModule *module, AOTMemInitData **data_list,
             /* If the module owns the binary data, free the bytes buffer */
             if (module->is_binary_freeable && data_list[i]->bytes)
                 wasm_runtime_free(data_list[i]->bytes);
+
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            destroy_init_expr(&data_list[i]->offset);
+#endif
             /* Free the data segment structure itself */
             wasm_runtime_free(data_list[i]);
         }
@@ -1043,11 +1076,11 @@ load_mem_init_data_list(const uint8 **p_buf, const uint8 *buf_end,
         uint32 byte_count;
         uint32 is_passive;
         uint32 memory_index;
-        InitializerExpression init_value;
+        InitializerExpression offset_expr;
 
         read_uint32(buf, buf_end, is_passive);
         read_uint32(buf, buf_end, memory_index);
-        if (!load_init_expr(&buf, buf_end, module, &init_value, error_buf,
+        if (!load_init_expr(&buf, buf_end, module, &offset_expr, error_buf,
                             error_buf_size)) {
             return false;
         }
@@ -1062,8 +1095,7 @@ load_mem_init_data_list(const uint8 **p_buf, const uint8 *buf_end,
         data_list[i]->is_passive = (bool)is_passive;
         data_list[i]->memory_index = memory_index;
 #endif
-        data_list[i]->offset.init_expr_type = init_value.init_expr_type;
+        data_list[i]->offset = offset_expr;
-        data_list[i]->offset.u = init_value.u;
         data_list[i]->byte_count = byte_count;
         data_list[i]->bytes = NULL;
         /* If the module owns the binary data, clone the bytes buffer */
@@ -1148,18 +1180,6 @@ fail:
     return false;
 }
 
-#if WASM_ENABLE_GC != 0
-static void
-destroy_init_expr(InitializerExpression *expr)
-{
-    if (expr->init_expr_type == INIT_EXPR_TYPE_STRUCT_NEW
-        || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW
-        || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW_FIXED) {
-        wasm_runtime_free(expr->u.data);
-    }
-}
-#endif /* end of WASM_ENABLE_GC != 0 */
-
 static void
 destroy_import_tables(AOTImportTable *import_tables)
 {
@@ -1183,6 +1203,9 @@ destroy_table_init_data_list(AOTTableInitData **data_list, uint32 count)
             for (j = 0; j < data_list[i]->value_count; j++) {
                 destroy_init_expr(&data_list[i]->init_values[j]);
             }
+#endif
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            destroy_init_expr(&data_list[i]->offset);
 #endif
             wasm_runtime_free(data_list[i]);
         }
@@ -1208,34 +1231,34 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             break;
         case INIT_EXPR_TYPE_I32_CONST:
         case INIT_EXPR_TYPE_F32_CONST:
-            read_uint32(buf, buf_end, expr->u.i32);
+            read_uint32(buf, buf_end, expr->u.unary.v.i32);
             break;
         case INIT_EXPR_TYPE_I64_CONST:
         case INIT_EXPR_TYPE_F64_CONST:
-            read_uint64(buf, buf_end, expr->u.i64);
+            read_uint64(buf, buf_end, expr->u.unary.v.i64);
             break;
         case INIT_EXPR_TYPE_V128_CONST:
-            i64x2 = (uint64 *)expr->u.v128.i64x2;
+            i64x2 = (uint64 *)expr->u.unary.v.v128.i64x2;
             CHECK_BUF(buf, buf_end, sizeof(uint64) * 2);
             wasm_runtime_read_v128(buf, &i64x2[0], &i64x2[1]);
             buf += sizeof(uint64) * 2;
             break;
         case INIT_EXPR_TYPE_GET_GLOBAL:
-            read_uint32(buf, buf_end, expr->u.global_index);
+            read_uint32(buf, buf_end, expr->u.unary.v.global_index);
             break;
         /* INIT_EXPR_TYPE_FUNCREF_CONST can be used when
            both reference types and GC are disabled */
         case INIT_EXPR_TYPE_FUNCREF_CONST:
-            read_uint32(buf, buf_end, expr->u.ref_index);
+            read_uint32(buf, buf_end, expr->u.unary.v.ref_index);
             break;
 #if WASM_ENABLE_GC != 0 || WASM_ENABLE_REF_TYPES != 0
         case INIT_EXPR_TYPE_REFNULL_CONST:
-            read_uint32(buf, buf_end, expr->u.ref_index);
+            read_uint32(buf, buf_end, expr->u.unary.v.ref_index);
             break;
 #endif /* end of WASM_ENABLE_GC != 0 || WASM_ENABLE_REF_TYPES != 0 */
 #if WASM_ENABLE_GC != 0
         case INIT_EXPR_TYPE_I31_NEW:
-            read_uint32(buf, buf_end, expr->u.i32);
+            read_uint32(buf, buf_end, expr->u.unary.v.i32);
             break;
         case INIT_EXPR_TYPE_STRUCT_NEW:
         {
@@ -1256,7 +1279,7 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             free_if_fail = true;
             init_values->count = field_count;
             init_values->type_idx = type_idx;
-            expr->u.data = init_values;
+            expr->u.unary.v.data = init_values;
 
             if (type_idx >= module->type_count) {
                 set_error_buf(error_buf, error_buf_size,
@@ -1294,7 +1317,7 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             break;
         }
         case INIT_EXPR_TYPE_STRUCT_NEW_DEFAULT:
-            read_uint32(buf, buf_end, expr->u.type_index);
+            read_uint32(buf, buf_end, expr->u.unary.v.type_index);
             break;
         case INIT_EXPR_TYPE_ARRAY_NEW:
         case INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT:
@@ -1317,8 +1340,8 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             }
 
             if (init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT) {
-                expr->u.array_new_default.type_index = type_idx;
+                expr->u.unary.v.array_new_default.type_index = type_idx;
-                expr->u.array_new_default.length = length;
+                expr->u.unary.v.array_new_default.length = length;
             }
             else {
                 uint32 i, elem_size, elem_data_count;
@@ -1329,7 +1352,7 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
                     return false;
                 }
                 free_if_fail = true;
-                expr->u.data = init_values;
+                expr->u.unary.v.data = init_values;
 
                 init_values->type_idx = type_idx;
                 init_values->length = length;
@@ -1357,6 +1380,34 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             break;
         }
 #endif /* end of WASM_ENABLE_GC != 0 */
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        case INIT_EXPR_TYPE_I32_ADD:
+        case INIT_EXPR_TYPE_I32_SUB:
+        case INIT_EXPR_TYPE_I32_MUL:
+        case INIT_EXPR_TYPE_I64_ADD:
+        case INIT_EXPR_TYPE_I64_SUB:
+        case INIT_EXPR_TYPE_I64_MUL:
+        {
+            expr->u.binary.l_expr = expr->u.binary.r_expr = NULL;
+            if (!(expr->u.binary.l_expr =
+                      loader_malloc(sizeof(InitializerExpression), error_buf,
+                                    error_buf_size))) {
+                goto fail;
+            }
+            if (!load_init_expr(&buf, buf_end, module, expr->u.binary.l_expr,
+                                error_buf, error_buf_size))
+                goto fail;
+            if (!(expr->u.binary.r_expr =
+                      loader_malloc(sizeof(InitializerExpression), error_buf,
+                                    error_buf_size))) {
+                goto fail;
+            }
+            if (!load_init_expr(&buf, buf_end, module, expr->u.binary.r_expr,
+                                error_buf, error_buf_size))
+                goto fail;
+            break;
+        }
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */
         default:
             set_error_buf(error_buf, error_buf_size, "invalid init expr type.");
             return false;
@@ -1369,10 +1420,13 @@ load_init_expr(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
 fail:
 #if WASM_ENABLE_GC != 0
     if (free_if_fail) {
-        wasm_runtime_free(expr->u.data);
+        wasm_runtime_free(expr->u.unary.v.data);
     }
 #else
     (void)free_if_fail;
+#endif
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    destroy_init_expr(expr);
 #endif
     return false;
 }
@@ -1535,14 +1589,16 @@ load_table_init_data_list(const uint8 **p_buf, const uint8 *buf_end,
     /* Create each table data segment */
     for (i = 0; i < module->table_init_data_count; i++) {
         uint32 mode, elem_type;
-        uint32 table_index, init_expr_type, value_count;
+        uint32 table_index, value_count;
-        uint64 init_expr_value, size1;
+        uint64 size1;
+        InitializerExpression offset_expr;
 
         read_uint32(buf, buf_end, mode);
         read_uint32(buf, buf_end, elem_type);
         read_uint32(buf, buf_end, table_index);
-        read_uint32(buf, buf_end, init_expr_type);
+        if (!load_init_expr(&buf, buf_end, module, &offset_expr, error_buf,
-        read_uint64(buf, buf_end, init_expr_value);
+                            error_buf_size))
+            return false;
 #if WASM_ENABLE_GC != 0
         if (wasm_is_type_multi_byte_type(elem_type)) {
             uint16 ref_type, nullable;
@@ -1588,8 +1644,7 @@ load_table_init_data_list(const uint8 **p_buf, const uint8 *buf_end,
             }
         }
 #endif
-        data_list[i]->offset.init_expr_type = (uint8)init_expr_type;
+        data_list[i]->offset = offset_expr;
-        data_list[i]->offset.u.i64 = (int64)init_expr_value;
         data_list[i]->value_count = value_count;
         for (j = 0; j < data_list[i]->value_count; j++) {
             if (!load_init_expr(&buf, buf_end, module,
@@ -1730,6 +1785,13 @@ load_types(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
             (void)u8;
 
             read_uint32(buf, buf_end, j);
+#if WASM_ENABLE_AOT_VALIDATOR != 0
+            /* an equivalence type should be before the type it refers to */
+            if (j > i) {
+                set_error_buf(error_buf, error_buf_size, "invalid type index");
+                goto fail;
+            }
+#endif
             if (module->types[j]->ref_count == UINT16_MAX) {
                 set_error_buf(error_buf, error_buf_size,
                               "wasm type's ref count too large");
@@ -1744,7 +1806,12 @@ load_types(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
         read_uint32(buf, buf_end, parent_type_idx);
         read_uint16(buf, buf_end, rec_count);
         read_uint16(buf, buf_end, rec_idx);
-
+#if WASM_ENABLE_AOT_VALIDATOR != 0
+        if (rec_idx > i) {
+            set_error_buf(error_buf, error_buf_size, "invalid rec_idx");
+            goto fail;
+        }
+#endif
         if (type_flag == WASM_TYPE_FUNC) {
             AOTFuncType *func_type;
 
@@ -1993,6 +2060,13 @@ load_types(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
                 AOTType *cur_type = module->types[j];
                 parent_type_idx = cur_type->parent_type_idx;
                 if (parent_type_idx != (uint32)-1) { /* has parent */
+#if WASM_ENABLE_AOT_VALIDATOR != 0
+                    if (parent_type_idx >= module->type_count) {
+                        set_error_buf(error_buf, error_buf_size,
+                                      "invalid parent type index");
+                        goto fail;
+                    }
+#endif
                     AOTType *parent_type = module->types[parent_type_idx];
 
                     module->types[j]->parent_type = parent_type;
@@ -2016,6 +2090,13 @@ load_types(const uint8 **p_buf, const uint8 *buf_end, AOTModule *module,
                 AOTType *cur_type = module->types[j];
                 parent_type_idx = cur_type->parent_type_idx;
                 if (parent_type_idx != (uint32)-1) { /* has parent */
+#if WASM_ENABLE_AOT_VALIDATOR != 0
+                    if (parent_type_idx >= module->type_count) {
+                        set_error_buf(error_buf, error_buf_size,
+                                      "invalid parent type index");
+                        goto fail;
+                    }
+#endif
                     AOTType *parent_type = module->types[parent_type_idx];
                     /* subtyping has been checked during compilation */
                     bh_assert(wasm_type_is_subtype_of(
@@ -4340,7 +4421,7 @@ aot_compatible_version(uint32 version)
      * refer to "AoT-compiled module compatibility among WAMR versions" in
      * ./doc/biuld_wasm_app.md
      */
-    return version == 4 || version == 3;
+    return version == AOT_CURRENT_VERSION;
 }
 
 static bool
@@ -4480,7 +4561,7 @@ aot_unload(AOTModule *module)
         destroy_import_globals(module->import_globals);
 
     if (module->globals) {
-#if WASM_ENABLE_GC != 0
+#if WASM_ENABLE_GC != 0 || WASM_ENABLE_EXTENDED_CONST_EXPR != 0
         uint32 i;
         for (i = 0; i < module->global_count; i++) {
             destroy_init_expr(&module->globals[i].init_expr);
@@ -4636,7 +4717,7 @@ aot_unload(AOTModule *module)
 }
 
 uint32
-aot_get_plt_table_size()
+aot_get_plt_table_size(void)
 {
     return get_plt_table_size();
 }

core/iwasm/aot/aot_reloc.h

@@ -185,6 +185,13 @@ typedef struct {
 #define REG_STRINGREF_SYM()
 #endif
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+#define REG_SHARED_HEAP_SYM()                 \
+    REG_SYM(wasm_runtime_check_and_update_last_used_shared_heap),
+#else
+#define REG_SHARED_HEAP_SYM()
+#endif
+
 #define REG_COMMON_SYMBOLS                \
     REG_SYM(aot_set_exception_with_id),   \
     REG_SYM(aot_invoke_native),           \
@@ -218,6 +225,7 @@ typedef struct {
     REG_LLVM_PGO_SYM()                    \
     REG_GC_SYM()                          \
     REG_STRINGREF_SYM()                   \
+    REG_SHARED_HEAP_SYM()                 \
 
 #define CHECK_RELOC_OFFSET(data_size) do {              \
     if (!check_reloc_offset(target_section_size,        \

core/iwasm/aot/aot_runtime.c

@@ -60,6 +60,16 @@ bh_static_assert(offsetof(AOTModuleInstanceExtra, stack_sizes) == 0);
 bh_static_assert(offsetof(AOTModuleInstanceExtra, shared_heap_base_addr_adj)
                  == 8);
 bh_static_assert(offsetof(AOTModuleInstanceExtra, shared_heap_start_off) == 16);
+bh_static_assert(offsetof(AOTModuleInstanceExtra, shared_heap_end_off) == 24);
+bh_static_assert(offsetof(AOTModuleInstanceExtra, shared_heap) == 32);
+
+bh_static_assert(offsetof(WASMSharedHeap, next) == 0);
+bh_static_assert(offsetof(WASMSharedHeap, chain_next) == 8);
+bh_static_assert(offsetof(WASMSharedHeap, heap_handle) == 16);
+bh_static_assert(offsetof(WASMSharedHeap, base_addr) == 24);
+bh_static_assert(offsetof(WASMSharedHeap, size) == 32);
+bh_static_assert(offsetof(WASMSharedHeap, start_off_mem64) == 40);
+bh_static_assert(offsetof(WASMSharedHeap, start_off_mem32) == 48);
 
 bh_static_assert(sizeof(CApiFuncImport) == sizeof(uintptr_t) * 3);
 
@@ -279,18 +289,21 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
     switch (flag) {
         case INIT_EXPR_TYPE_GET_GLOBAL:
         {
-            if (!check_global_init_expr(module, init_expr->u.global_index,
+            if (!check_global_init_expr(module,
+                                        init_expr->u.unary.v.global_index,
                                         error_buf, error_buf_size)) {
                 return false;
             }
-            if (init_expr->u.global_index < module->import_global_count) {
+            if (init_expr->u.unary.v.global_index
+                < module->import_global_count) {
                 PUT_REF_TO_ADDR(
-                    addr, module->import_globals[init_expr->u.global_index]
+                    addr,
+                    module->import_globals[init_expr->u.unary.v.global_index]
                         .global_data_linked.gc_obj);
             }
             else {
-                uint32 global_idx =
+                uint32 global_idx = init_expr->u.unary.v.global_index
-                    init_expr->u.global_index - module->import_global_count;
+                                    - module->import_global_count;
                 return assign_table_init_value(
                     module_inst, module, &module->globals[global_idx].init_expr,
                     addr, error_buf, error_buf_size);
@@ -306,7 +319,7 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
         case INIT_EXPR_TYPE_FUNCREF_CONST:
         {
             WASMFuncObjectRef func_obj = NULL;
-            uint32 func_idx = init_expr->u.u32;
+            uint32 func_idx = init_expr->u.unary.v.u32;
 
             if (func_idx != UINT32_MAX) {
                 if (!(func_obj =
@@ -321,7 +334,8 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
         }
         case INIT_EXPR_TYPE_I31_NEW:
         {
-            WASMI31ObjectRef i31_obj = wasm_i31_obj_new(init_expr->u.i32);
+            WASMI31ObjectRef i31_obj =
+                wasm_i31_obj_new(init_expr->u.unary.v.i32);
             PUT_REF_TO_ADDR(addr, i31_obj);
             break;
         }
@@ -335,11 +349,12 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
             uint32 type_idx;
 
             if (flag == INIT_EXPR_TYPE_STRUCT_NEW) {
-                init_values = (WASMStructNewInitValues *)init_expr->u.data;
+                init_values =
+                    (WASMStructNewInitValues *)init_expr->u.unary.v.data;
                 type_idx = init_values->type_idx;
             }
             else {
-                type_idx = init_expr->u.type_index;
+                type_idx = init_expr->u.unary.v.type_index;
             }
 
             struct_type = (WASMStructType *)module->types[type_idx];
@@ -388,12 +403,13 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
             uint32 type_idx, len;
 
             if (flag == INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT) {
-                type_idx = init_expr->u.array_new_default.type_index;
+                type_idx = init_expr->u.unary.v.array_new_default.type_index;
-                len = init_expr->u.array_new_default.length;
+                len = init_expr->u.unary.v.array_new_default.length;
                 arr_init_val = &empty_val;
             }
             else {
-                init_values = (WASMArrayNewInitValues *)init_expr->u.data;
+                init_values =
+                    (WASMArrayNewInitValues *)init_expr->u.unary.v.data;
                 type_idx = init_values->type_idx;
                 len = init_values->length;
 
@@ -444,6 +460,90 @@ assign_table_init_value(AOTModuleInstance *module_inst, AOTModule *module,
 }
 #endif /* end of WASM_ENABLE_GC != 0 */
 
+static bool
+get_init_value_recursive(AOTModuleInstance *module_inst, AOTModule *module,
+                         InitializerExpression *expr, WASMValue *value,
+                         char *error_buf, uint32 error_buf_size)
+{
+    uint8 flag = expr->init_expr_type;
+    switch (flag) {
+        case INIT_EXPR_TYPE_GET_GLOBAL:
+        {
+            if (!check_global_init_expr(module, expr->u.unary.v.global_index,
+                                        error_buf, error_buf_size)) {
+                return false;
+            }
+#if WASM_ENABLE_GC == 0
+            *value = module->import_globals[expr->u.unary.v.global_index]
+                         .global_data_linked;
+#else
+            if (expr->u.unary.v.global_index < module->import_global_count) {
+                *value = module->import_globals[expr->u.unary.v.global_index]
+                             .global_data_linked;
+            }
+            else {
+                *value = module
+                             ->globals[expr->u.unary.v.global_index
+                                       - module->import_global_count]
+                             .init_expr.u.unary.v;
+            }
+#endif
+            break;
+        }
+        case INIT_EXPR_TYPE_I32_CONST:
+        case INIT_EXPR_TYPE_I64_CONST:
+        {
+            *value = expr->u.unary.v;
+            break;
+        }
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        case INIT_EXPR_TYPE_I32_ADD:
+        case INIT_EXPR_TYPE_I32_SUB:
+        case INIT_EXPR_TYPE_I32_MUL:
+        case INIT_EXPR_TYPE_I64_ADD:
+        case INIT_EXPR_TYPE_I64_SUB:
+        case INIT_EXPR_TYPE_I64_MUL:
+        {
+            WASMValue l_value, r_value;
+            if (!get_init_value_recursive(module_inst, module,
+                                          expr->u.binary.l_expr, &l_value,
+                                          error_buf, error_buf_size)) {
+                return false;
+            }
+            if (!get_init_value_recursive(module_inst, module,
+                                          expr->u.binary.r_expr, &r_value,
+                                          error_buf, error_buf_size)) {
+                return false;
+            }
+
+            if (flag == INIT_EXPR_TYPE_I32_ADD) {
+                value->i32 = l_value.i32 + r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I32_SUB) {
+                value->i32 = l_value.i32 - r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I32_MUL) {
+                value->i32 = l_value.i32 * r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_ADD) {
+                value->i64 = l_value.i64 + r_value.i64;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_SUB) {
+                value->i64 = l_value.i64 - r_value.i64;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_MUL) {
+                value->i64 = l_value.i64 * r_value.i64;
+            }
+            break;
+        }
+#endif
+        default:
+            return false;
+    }
+
+    return true;
+}
+
 static bool
 global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
                    char *error_buf, uint32 error_buf_size)
@@ -472,30 +572,24 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
         flag = init_expr->init_expr_type;
         switch (flag) {
             case INIT_EXPR_TYPE_GET_GLOBAL:
+            case INIT_EXPR_TYPE_I32_CONST:
+            case INIT_EXPR_TYPE_I64_CONST:
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            case INIT_EXPR_TYPE_I32_ADD:
+            case INIT_EXPR_TYPE_I32_SUB:
+            case INIT_EXPR_TYPE_I32_MUL:
+            case INIT_EXPR_TYPE_I64_ADD:
+            case INIT_EXPR_TYPE_I64_SUB:
+            case INIT_EXPR_TYPE_I64_MUL:
+#endif
             {
-                if (!check_global_init_expr(module, init_expr->u.global_index,
+                WASMValue value;
-                                            error_buf, error_buf_size)) {
+                if (!get_init_value_recursive(module_inst, module, init_expr,
+                                              &value, error_buf,
+                                              error_buf_size)) {
                     return false;
                 }
-#if WASM_ENABLE_GC == 0
+                init_global_data(p, global->type.val_type, &value);
-                init_global_data(
-                    p, global->type.val_type,
-                    &module->import_globals[init_expr->u.global_index]
-                         .global_data_linked);
-#else
-                if (init_expr->u.global_index < module->import_global_count) {
-                    init_global_data(
-                        p, global->type.val_type,
-                        &module->import_globals[init_expr->u.global_index]
-                             .global_data_linked);
-                }
-                else {
-                    uint32 global_idx =
-                        init_expr->u.global_index - module->import_global_count;
-                    init_global_data(p, global->type.val_type,
-                                     &module->globals[global_idx].init_expr.u);
-                }
-#endif
                 break;
             }
 #if WASM_ENABLE_GC == 0 && WASM_ENABLE_REF_TYPES != 0
@@ -516,7 +610,7 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
             case INIT_EXPR_TYPE_FUNCREF_CONST:
             {
                 WASMFuncObjectRef func_obj = NULL;
-                uint32 func_idx = init_expr->u.u32;
+                uint32 func_idx = init_expr->u.unary.v.ref_index;
 
                 if (func_idx != UINT32_MAX) {
                     if (!(func_obj =
@@ -531,7 +625,8 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
             }
             case INIT_EXPR_TYPE_I31_NEW:
             {
-                WASMI31ObjectRef i31_obj = wasm_i31_obj_new(init_expr->u.i32);
+                WASMI31ObjectRef i31_obj =
+                    wasm_i31_obj_new(init_expr->u.unary.v.i32);
                 PUT_REF_TO_ADDR(p, i31_obj);
                 break;
             }
@@ -545,11 +640,12 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
                 uint32 type_idx;
 
                 if (flag == INIT_EXPR_TYPE_STRUCT_NEW) {
-                    init_values = (WASMStructNewInitValues *)init_expr->u.data;
+                    init_values =
+                        (WASMStructNewInitValues *)init_expr->u.unary.v.data;
                     type_idx = init_values->type_idx;
                 }
                 else {
-                    type_idx = init_expr->u.type_index;
+                    type_idx = init_expr->u.unary.v.type_index;
                 }
 
                 struct_type = (WASMStructType *)module->types[type_idx];
@@ -599,12 +695,14 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
                 uint32 type_idx, len;
 
                 if (flag == INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT) {
-                    type_idx = init_expr->u.array_new_default.type_index;
+                    type_idx =
-                    len = init_expr->u.array_new_default.length;
+                        init_expr->u.unary.v.array_new_default.type_index;
+                    len = init_expr->u.unary.v.array_new_default.length;
                     arr_init_val = &empty_val;
                 }
                 else {
-                    init_values = (WASMArrayNewInitValues *)init_expr->u.data;
+                    init_values =
+                        (WASMArrayNewInitValues *)init_expr->u.unary.v.data;
                     type_idx = init_values->type_idx;
                     len = init_values->length;
 
@@ -650,7 +748,8 @@ global_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
 #endif /* end of WASM_ENABLE_GC != 0 */
             default:
             {
-                init_global_data(p, global->type.val_type, &init_expr->u);
+                init_global_data(p, global->type.val_type,
+                                 &init_expr->u.unary.v);
                 break;
             }
         }
@@ -671,6 +770,7 @@ tables_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
     uint64 total_size;
     AOTTableInitData *table_seg;
     AOTTableInstance *tbl_inst = first_tbl_inst;
+    uint8 offset_flag;
 
     total_size = (uint64)sizeof(AOTTableInstance *) * module_inst->table_count;
     if (total_size > 0
@@ -743,28 +843,25 @@ tables_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
         tbl_inst = module_inst->tables[table_seg->table_index];
         bh_assert(tbl_inst);
 
+        offset_flag = table_seg->offset.init_expr_type;
+
 #if WASM_ENABLE_REF_TYPES != 0
-        bh_assert(
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-            table_seg->offset.init_expr_type
+                  || offset_flag == INIT_EXPR_TYPE_FUNCREF_CONST
-                == (tbl_inst->is_table64 ? INIT_EXPR_TYPE_I64_CONST
+                  || offset_flag == INIT_EXPR_TYPE_REFNULL_CONST
-                                         : INIT_EXPR_TYPE_I32_CONST)
+                  || (tbl_inst->is_table64 ? is_valid_i64_offset(offset_flag)
-            || table_seg->offset.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL
+                                           : is_valid_i32_offset(offset_flag)));
-            || table_seg->offset.init_expr_type == INIT_EXPR_TYPE_FUNCREF_CONST
-            || table_seg->offset.init_expr_type
-                   == INIT_EXPR_TYPE_REFNULL_CONST);
 #else
-        bh_assert(table_seg->offset.init_expr_type
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                      == (tbl_inst->is_table64 ? INIT_EXPR_TYPE_I64_CONST
+                  || (tbl_inst->is_table64 ? is_valid_i64_offset(offset_flag)
-                                               : INIT_EXPR_TYPE_I32_CONST)
+                                           : is_valid_i32_offset(offset_flag)));
-                  || table_seg->offset.init_expr_type
-                         == INIT_EXPR_TYPE_GET_GLOBAL);
 #endif
 
         /* Resolve table data base offset */
         /* TODO: The table64 current implementation assumes table max size
          * UINT32_MAX, so the offset conversion here is safe */
-        if (table_seg->offset.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL) {
+        if (offset_flag == INIT_EXPR_TYPE_GET_GLOBAL) {
-            global_index = table_seg->offset.u.global_index;
+            global_index = table_seg->offset.u.unary.v.global_index;
 
             if (!check_global_init_expr(module, global_index, error_buf,
                                         error_buf_size)) {
@@ -782,8 +879,15 @@ tables_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
             base_offset =
                 *(uint32 *)(module_inst->global_data + global_data_offset);
         }
-        else
+        else {
-            base_offset = (uint32)table_seg->offset.u.i32;
+            WASMValue offset_value;
+            if (!get_init_value_recursive(module_inst, module,
+                                          &table_seg->offset, &offset_value,
+                                          error_buf, error_buf_size)) {
+                return false;
+            }
+            base_offset = (uint32)offset_value.i32;
+        }
 
         /* Copy table data */
         /* base_offset only since length might negative */
@@ -818,7 +922,7 @@ tables_instantiate(AOTModuleInstance *module_inst, AOTModule *module,
 #if WASM_ENABLE_GC == 0
         for (j = 0; j < length; j++) {
             tbl_inst->elems[base_offset + j] =
-                table_seg->init_values[j].u.ref_index;
+                table_seg->init_values[j].u.unary.v.ref_index;
         }
 #endif
     }
@@ -1118,6 +1222,7 @@ memories_instantiate(AOTModuleInstance *module_inst, AOTModuleInstance *parent,
     AOTMemInitData *data_seg;
     uint64 total_size;
     mem_offset_t base_offset;
+    uint8 offset_flag;
 
     module_inst->memory_count = memory_count;
     total_size = sizeof(AOTMemoryInstance *) * (uint64)memory_count;
@@ -1156,15 +1261,15 @@ memories_instantiate(AOTModuleInstance *module_inst, AOTModuleInstance *parent,
                initialized */
             continue;
 
-        bh_assert(data_seg->offset.init_expr_type
+        offset_flag = data_seg->offset.init_expr_type;
-                      == (memory_inst->is_memory64 ? INIT_EXPR_TYPE_I64_CONST
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                                                   : INIT_EXPR_TYPE_I32_CONST)
+                  || (memory_inst->is_memory64
-                  || data_seg->offset.init_expr_type
+                          ? is_valid_i64_offset(offset_flag)
-                         == INIT_EXPR_TYPE_GET_GLOBAL);
+                          : is_valid_i32_offset(offset_flag)));
 
         /* Resolve memory data base offset */
-        if (data_seg->offset.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL) {
+        if (offset_flag == INIT_EXPR_TYPE_GET_GLOBAL) {
-            global_index = data_seg->offset.u.global_index;
+            global_index = data_seg->offset.u.unary.v.global_index;
 
             if (!check_global_init_expr(module, global_index, error_buf,
                                         error_buf_size)) {
@@ -1192,14 +1297,20 @@ memories_instantiate(AOTModuleInstance *module_inst, AOTModuleInstance *parent,
             }
         }
         else {
+            WASMValue offset_value;
+            if (!get_init_value_recursive(module_inst, module,
+                                          &data_seg->offset, &offset_value,
+                                          error_buf, error_buf_size)) {
+                return false;
+            }
 #if WASM_ENABLE_MEMORY64 != 0
             if (memory_inst->is_memory64) {
-                base_offset = data_seg->offset.u.i64;
+                base_offset = offset_value.i64;
             }
             else
 #endif
             {
-                base_offset = data_seg->offset.u.u32;
+                base_offset = offset_value.u32;
             }
         }
 
@@ -1989,6 +2100,8 @@ aot_instantiate(AOTModule *module, AOTModuleInstance *parent,
 #else
     extra->shared_heap_start_off.u32[0] = UINT32_MAX;
 #endif
+    /* After shared heap chain, will early stop if shared heap is NULL */
+    extra->shared_heap = NULL;
 
 #if WASM_ENABLE_PERF_PROFILING != 0
     total_size = sizeof(AOTFuncPerfProfInfo)
@@ -2043,6 +2156,7 @@ aot_instantiate(AOTModule *module, AOTModuleInstance *parent,
         uint8 tbl_elem_type;
         uint32 tbl_init_size, tbl_max_size, j;
         WASMRefType *tbl_elem_ref_type;
+        WASMValue offset_value;
 
         bh_assert(table_init_data);
 
@@ -2074,69 +2188,73 @@ aot_instantiate(AOTModule *module, AOTModuleInstance *parent,
         if (!wasm_elem_is_active(table_init_data->mode)) {
             continue;
         }
-
+        uint8 offset_flag = table_init_data->offset.init_expr_type;
-        bh_assert(table_init_data->offset.init_expr_type
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                      == INIT_EXPR_TYPE_I32_CONST
+                  || offset_flag == INIT_EXPR_TYPE_FUNCREF_CONST
-                  || table_init_data->offset.init_expr_type
+                  || offset_flag == INIT_EXPR_TYPE_REFNULL_CONST
-                         == INIT_EXPR_TYPE_GET_GLOBAL
+                  || offset_flag == INIT_EXPR_TYPE_I32_CONST
-                  || table_init_data->offset.init_expr_type
+                  || offset_flag == INIT_EXPR_TYPE_I32_ADD
-                         == INIT_EXPR_TYPE_FUNCREF_CONST
+                  || offset_flag == INIT_EXPR_TYPE_I32_SUB
-                  || table_init_data->offset.init_expr_type
+                  || offset_flag == INIT_EXPR_TYPE_I32_MUL);
-                         == INIT_EXPR_TYPE_REFNULL_CONST);
 
         /* init vec(funcidx) or vec(expr) */
-        if (table_init_data->offset.init_expr_type
+        if (offset_flag == INIT_EXPR_TYPE_GET_GLOBAL) {
-            == INIT_EXPR_TYPE_GET_GLOBAL) {
             uint32 data_offset;
-            if (!check_global_init_expr(module,
+            if (!check_global_init_expr(
-                                        table_init_data->offset.u.global_index,
+                    module, table_init_data->offset.u.unary.v.global_index,
                     error_buf, error_buf_size)) {
                 goto fail;
             }
 
-            if (table_init_data->offset.u.global_index
+            if (table_init_data->offset.u.unary.v.global_index
                 < module->import_global_count) {
-                data_offset =
+                data_offset = module
-                    module
+                                  ->import_globals[table_init_data->offset.u
-                        ->import_globals[table_init_data->offset.u.global_index]
+                                                       .unary.v.global_index]
                                   .data_offset;
             }
             else {
                 data_offset =
                     module
-                        ->globals[table_init_data->offset.u.global_index
+                        ->globals[table_init_data->offset.u.unary.v.global_index
                                   - module->import_global_count]
                         .data_offset;
             }
-
+            offset_value.i32 =
-            table_init_data->offset.u.i32 =
                 *(uint32 *)(module_inst->global_data + data_offset);
         }
+        else {
+            if (!get_init_value_recursive(
+                    module_inst, module, &table_init_data->offset,
+                    &offset_value, error_buf, error_buf_size)) {
+                goto fail;
+            }
+        }
 
         /* check offset since length might negative */
-        if ((uint32)table_init_data->offset.u.i32 > table->cur_size) {
+        if ((uint32)offset_value.i32 > table->cur_size) {
-            LOG_DEBUG("base_offset(%d) > table->cur_size(%d)",
+            LOG_DEBUG("base_offset(%d) > table->cur_size(%d)", offset_value.i32,
-                      table_init_data->offset.u.i32, table->cur_size);
+                      table->cur_size);
             set_error_buf(error_buf, error_buf_size,
                           "out of bounds table access");
             goto fail;
         }
 
-        if ((uint32)table_init_data->offset.u.i32 + table_init_data->value_count
+        if ((uint32)offset_value.i32 + table_init_data->value_count
             > table->cur_size) {
             LOG_DEBUG("base_offset(%d) + length(%d) > table->cur_size(%d)",
-                      table_init_data->offset.u.i32,
+                      offset_value.i32, table_init_data->value_count,
-                      table_init_data->value_count, table->cur_size);
+                      table->cur_size);
             set_error_buf(error_buf, error_buf_size,
                           "out of bounds table access");
             goto fail;
         }
 
         for (j = 0; j < module->table_init_data_list[i]->value_count; j++) {
-            if (!assign_table_init_value(
+            if (!assign_table_init_value(module_inst, module,
-                    module_inst, module, &table_init_data->init_values[j],
+                                         &table_init_data->init_values[j],
-                    table_data + table_init_data->offset.u.i32 + j, error_buf,
+                                         table_data + offset_value.i32 + j,
-                    error_buf_size)) {
+                                         error_buf, error_buf_size)) {
                 goto fail;
             }
         }
@@ -3639,7 +3757,7 @@ aot_get_module_inst_mem_consumption(const AOTModuleInstance *module_inst,
     for (i = 0; i < module_inst->memory_count; i++) {
         AOTMemoryInstance *mem_inst = module_inst->memories[i];
         mem_conspn->memories_size +=
-            mem_inst->num_bytes_per_page * mem_inst->cur_page_count;
+            (uint64)mem_inst->num_bytes_per_page * mem_inst->cur_page_count;
         mem_conspn->app_heap_size =
             mem_inst->heap_data_end - mem_inst->heap_data;
         /* size of app heap structure */
@@ -3729,10 +3847,10 @@ aot_table_init(AOTModuleInstance *module_inst, uint32 tbl_idx,
     for (i = 0; i < length; i++) {
 #if WASM_ENABLE_GC != 0
         /* UINT32_MAX indicates that it is a null ref */
-        if (init_values[i].u.ref_index != UINT32_MAX) {
+        if (init_values[i].u.unary.v.ref_index != UINT32_MAX) {
-            if (!(func_obj = aot_create_func_obj(module_inst,
+            if (!(func_obj = aot_create_func_obj(
-                                                 init_values[i].u.ref_index,
+                      module_inst, init_values[i].u.unary.v.ref_index, true,
-                                                 true, NULL, 0))) {
+                      NULL, 0))) {
                 aot_set_exception_with_id(module_inst, EXCE_NULL_FUNC_OBJ);
                 return;
             }
@@ -3742,7 +3860,7 @@ aot_table_init(AOTModuleInstance *module_inst, uint32 tbl_idx,
             table_elems[i] = NULL_REF;
         }
 #else
-        table_elems[i] = init_values[i].u.ref_index;
+        table_elems[i] = init_values[i].u.unary.v.ref_index;
 #endif
     }
 }
@@ -4137,9 +4255,9 @@ aot_frame_update_profile_info(WASMExecEnv *exec_env, bool alloc_frame)
 }
 #endif /* end of WASM_ENABLE_AOT_STACK_FRAME != 0 */
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 uint32
-aot_copy_callstack_tiny_frame(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+aot_copy_callstack_tiny_frame(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                               const uint32 length, const uint32 skip_n,
                               char *error_buf, uint32 error_buf_size)
 {
@@ -4193,7 +4311,7 @@ aot_copy_callstack_tiny_frame(WASMExecEnv *exec_env, wasm_frame_t *buffer,
 }
 
 uint32
-aot_copy_callstack_standard_frame(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+aot_copy_callstack_standard_frame(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                                   const uint32 length, const uint32 skip_n,
                                   char *error_buf, uint32_t error_buf_size)
 {
@@ -4243,7 +4361,7 @@ aot_copy_callstack_standard_frame(WASMExecEnv *exec_env, wasm_frame_t *buffer,
 }
 
 uint32
-aot_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+aot_copy_callstack(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                    const uint32 length, const uint32 skip_n, char *error_buf,
                    uint32_t error_buf_size)
 {
@@ -4265,7 +4383,7 @@ aot_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
                                              error_buf, error_buf_size);
     }
 }
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 #if WASM_ENABLE_DUMP_CALL_STACK != 0
 bool

core/iwasm/aot/aot_runtime.h

@@ -125,6 +125,8 @@ typedef struct AOTModuleInstanceExtra {
      */
     DefPointer(uint8 *, shared_heap_base_addr_adj);
     MemBound shared_heap_start_off;
+    MemBound shared_heap_end_off;
+    DefPointer(WASMSharedHeap *, shared_heap);
 
     WASMModuleInstanceExtraCommon common;
 
@@ -142,9 +144,6 @@ typedef struct AOTModuleInstanceExtra {
     WASMModuleInstanceCommon **import_func_module_insts;
 #endif
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap;
-#endif
 } AOTModuleInstanceExtra;
 
 #if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64)
@@ -787,12 +786,12 @@ aot_frame_update_profile_info(WASMExecEnv *exec_env, bool alloc_frame);
 bool
 aot_create_call_stack(struct WASMExecEnv *exec_env);
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 uint32
-aot_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+aot_copy_callstack(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                    const uint32 length, const uint32 skip_n, char *error_buf,
                    uint32_t error_buf_size);
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 /**
  * @brief Dump wasm call stack or get the size

core/iwasm/common/arch/fneh.txt

@@ -0,0 +1,3 @@
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_aarch64.s

@@ -79,3 +79,6 @@ return:
         add     sp, sp, #0x30          /* restore sp */
         ret
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_aarch64_simd.s

@@ -77,3 +77,6 @@ return:
         add     sp, sp, #0x30          /* restore sp */
         ret
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_arc.s

@@ -67,3 +67,6 @@ call_func:
     j_s     [blink]             /* ret */
     nop_s
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_arm.s

@@ -73,3 +73,6 @@ return:
         add     sp, sp, #4
         ldmfd   sp!, {r4, r5, r6, r7, lr}
         bx      lr
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_arm_vfp.s

@@ -84,3 +84,6 @@ return:
         ldmfd   sp!, {r4, r5, r6, r7, lr}
         bx      lr
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_em64.s

@@ -62,3 +62,6 @@ push_args_end:
     leave
     ret
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_em64_simd.s

@@ -62,3 +62,6 @@ push_args_end:
     leave
     ret
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_ia32.s

@@ -35,3 +35,6 @@ skip_push_args:
     leave
     ret
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_mips.s

@@ -72,3 +72,6 @@ done:
 
     j $31
     .end invokeNative
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_thumb.s

@@ -89,3 +89,6 @@ return:
         pop     {r4, r5, r6, r7}
         mov     lr, r3
         bx      lr
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_thumb_vfp.s

@@ -98,3 +98,6 @@ return:
         mov     lr, r3
         bx      lr
 
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/arch/invokeNative_xtensa.s

@@ -72,3 +72,6 @@ call_func:
 
 return:
         retw.n
+#if defined(__linux__) && defined(__ELF__)
+.section .note.GNU-stack,"",%progbits
+#endif

core/iwasm/common/wasm_loader_common.c

@@ -225,3 +225,18 @@ read_leb(uint8 **p_buf, const uint8 *buf_end, uint32 maxbits, bool sign,
             return false;
     }
 }
+
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+void
+destroy_init_expr_recursive(InitializerExpression *expr)
+{
+    if (expr == NULL) {
+        return;
+    }
+    if (is_expr_binary_op(expr->init_expr_type)) {
+        destroy_init_expr_recursive(expr->u.binary.l_expr);
+        destroy_init_expr_recursive(expr->u.binary.r_expr);
+    }
+    wasm_runtime_free(expr);
+}
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */

core/iwasm/common/wasm_loader_common.h

@@ -50,6 +50,11 @@ void
 wasm_loader_set_error_buf(char *error_buf, uint32 error_buf_size,
                           const char *string, bool is_aot);
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+void
+destroy_init_expr_recursive(InitializerExpression *expr);
+#endif
+
 #ifdef __cplusplus
 }
 #endif

core/iwasm/common/wasm_memory.c

@@ -143,7 +143,7 @@ is_bounds_checks_enabled(WASMModuleInstanceCommon *module_inst)
 
 #if WASM_ENABLE_SHARED_HEAP != 0
 static void *
-wasm_mmap_linear_memory(uint64_t map_size, uint64 commit_size);
+wasm_mmap_linear_memory(uint64 map_size, uint64 commit_size);
 static void
 wasm_munmap_linear_memory(void *mapped_mem, uint64 commit_size,
                           uint64 map_size);
@@ -177,19 +177,35 @@ wasm_runtime_create_shared_heap(SharedHeapInitArgs *init_args)
         goto fail1;
     }
 
-    if (!(heap->heap_handle =
+    size = align_uint(size, os_getpagesize());
-              runtime_malloc(mem_allocator_get_heap_struct_size()))) {
+    if (size > APP_HEAP_SIZE_MAX || size < APP_HEAP_SIZE_MIN) {
+        LOG_WARNING("Invalid size of shared heap");
         goto fail2;
     }
 
-    size = align_uint(size, os_getpagesize());
     heap->size = size;
     heap->start_off_mem64 = UINT64_MAX - heap->size + 1;
     heap->start_off_mem32 = UINT32_MAX - heap->size + 1;
+    heap->attached_count = 0;
 
-    if (size > APP_HEAP_SIZE_MAX || size < APP_HEAP_SIZE_MIN) {
+    if (init_args->pre_allocated_addr != NULL) {
-        LOG_WARNING("Invalid size of shared heap");
+        /* Create shared heap from a pre allocated buffer, its size need to
-        goto fail3;
+         * align with system page */
+        if (size != init_args->size) {
+            LOG_WARNING("Pre allocated size need to be aligned with system "
+                        "page size to create shared heap");
+            goto fail2;
+        }
+
+        heap->heap_handle = NULL;
+        heap->base_addr = init_args->pre_allocated_addr;
+        LOG_VERBOSE("Create preallocated shared heap %p with size %u",
+                    heap->base_addr, size);
+    }
+    else {
+        if (!(heap->heap_handle =
+                  runtime_malloc(mem_allocator_get_heap_struct_size()))) {
+            goto fail2;
         }
 
 #ifndef OS_ENABLE_HW_BOUND_CHECK
@@ -211,6 +227,9 @@ wasm_runtime_create_shared_heap(SharedHeapInitArgs *init_args)
             LOG_WARNING("init share heap failed");
             goto fail4;
         }
+        LOG_VERBOSE("Create pool shared heap %p with size %u", heap->base_addr,
+                    size);
+    }
 
     os_mutex_lock(&shared_heap_list_lock);
     if (shared_heap_list == NULL) {
@@ -233,6 +252,219 @@ fail1:
     return NULL;
 }
 
+WASMSharedHeap *
+wasm_runtime_chain_shared_heaps(WASMSharedHeap *head, WASMSharedHeap *body)
+{
+    WASMSharedHeap *cur;
+    bool heap_handle_exist = false;
+
+    if (!head || !body) {
+        LOG_WARNING("Invalid shared heap to chain.");
+        return NULL;
+    }
+    heap_handle_exist = head->heap_handle != NULL;
+
+    os_mutex_lock(&shared_heap_list_lock);
+    if (head->attached_count != 0 || body->attached_count != 0) {
+        LOG_WARNING("To create shared heap chain, all shared heap need to be "
+                    "detached first.");
+        os_mutex_unlock(&shared_heap_list_lock);
+        return NULL;
+    }
+    for (cur = shared_heap_list; cur; cur = cur->next) {
+        if (cur->chain_next == body || cur->chain_next == head) {
+            LOG_WARNING(
+                "To create shared heap chain, both the 'head' and 'body' "
+                "shared heap can't already be the 'body' in another a chain");
+            os_mutex_unlock(&shared_heap_list_lock);
+            return NULL;
+        }
+        if (cur == head && cur->chain_next) {
+            LOG_WARNING(
+                "To create shared heap chain, the 'head' shared heap can't "
+                "already be the 'head' in another a chain");
+            os_mutex_unlock(&shared_heap_list_lock);
+            return NULL;
+        }
+    }
+    for (cur = body; cur; cur = cur->chain_next) {
+        if (cur->heap_handle && heap_handle_exist) {
+            LOG_WARNING(
+                "To create shared heap chain, only one of shared heap can "
+                "dynamically shared_heap_malloc and shared_heap_free, the rest "
+                "can only be pre-allocated shared heap");
+            os_mutex_unlock(&shared_heap_list_lock);
+            return NULL;
+        }
+        if (cur->heap_handle)
+            heap_handle_exist = true;
+    }
+
+    head->start_off_mem64 = body->start_off_mem64 - head->size;
+    head->start_off_mem32 = body->start_off_mem32 - head->size;
+    head->chain_next = body;
+    os_mutex_unlock(&shared_heap_list_lock);
+    return head;
+}
+
+WASMSharedHeap *
+wasm_runtime_unchain_shared_heaps(WASMSharedHeap *head, bool entire_chain)
+{
+    WASMSharedHeap *cur, *tmp;
+
+    if (!head || !head->chain_next) {
+        LOG_WARNING("Invalid shared heap chain to disconnect the head from.");
+        return NULL;
+    }
+
+    os_mutex_lock(&shared_heap_list_lock);
+    if (head->attached_count != 0) {
+        LOG_WARNING("To disconnect the shared heap head from the shared heap "
+                    "chain, the shared heap chain needs to be detached first.");
+        os_mutex_unlock(&shared_heap_list_lock);
+        return NULL;
+    }
+
+    cur = head;
+    while (cur && cur->chain_next) {
+        cur->start_off_mem64 = UINT64_MAX - cur->size + 1;
+        cur->start_off_mem32 = UINT32_MAX - cur->size + 1;
+        tmp = cur;
+        cur = cur->chain_next;
+        tmp->chain_next = NULL;
+        if (!entire_chain)
+            break;
+    }
+    os_mutex_unlock(&shared_heap_list_lock);
+    return cur;
+}
+
+static uint8 *
+get_last_used_shared_heap_base_addr_adj(WASMModuleInstanceCommon *module_inst)
+{
+#if WASM_ENABLE_INTERP != 0
+    if (module_inst->module_type == Wasm_Module_Bytecode) {
+        WASMModuleInstanceExtra *e =
+            (WASMModuleInstanceExtra *)((WASMModuleInstance *)module_inst)->e;
+        return e->shared_heap_base_addr_adj;
+    }
+#endif /* end of WASM_ENABLE_INTERP != 0 */
+#if WASM_ENABLE_AOT != 0
+    if (module_inst->module_type == Wasm_Module_AoT) {
+        AOTModuleInstanceExtra *e =
+            (AOTModuleInstanceExtra *)((AOTModuleInstance *)module_inst)->e;
+        return e->shared_heap_base_addr_adj;
+    }
+#endif /* end of WASM_ENABLE_AOT != 0 */
+    return 0;
+}
+
+static uintptr_t
+get_last_used_shared_heap_start_offset(WASMModuleInstanceCommon *module_inst)
+{
+#if WASM_ENABLE_INTERP != 0
+    if (module_inst->module_type == Wasm_Module_Bytecode) {
+        WASMModuleInstanceExtra *e =
+            (WASMModuleInstanceExtra *)((WASMModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        return e->shared_heap_start_off.u64;
+#else
+        return e->shared_heap_start_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_INTERP != 0 */
+#if WASM_ENABLE_AOT != 0
+    if (module_inst->module_type == Wasm_Module_AoT) {
+        AOTModuleInstanceExtra *e =
+            (AOTModuleInstanceExtra *)((AOTModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        return e->shared_heap_start_off.u64;
+#else
+        return e->shared_heap_start_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_AOT != 0 */
+    return 0;
+}
+
+static uintptr_t
+get_last_used_shared_heap_end_offset(WASMModuleInstanceCommon *module_inst)
+{
+#if WASM_ENABLE_INTERP != 0
+    if (module_inst->module_type == Wasm_Module_Bytecode) {
+        WASMModuleInstanceExtra *e =
+            (WASMModuleInstanceExtra *)((WASMModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        return e->shared_heap_end_off.u64;
+#else
+        return e->shared_heap_end_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_INTERP != 0 */
+#if WASM_ENABLE_AOT != 0
+    if (module_inst->module_type == Wasm_Module_AoT) {
+        AOTModuleInstanceExtra *e =
+            (AOTModuleInstanceExtra *)((AOTModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        return e->shared_heap_end_off.u64;
+#else
+        return e->shared_heap_end_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_AOT != 0 */
+    return 0;
+}
+
+static void
+update_last_used_shared_heap(WASMModuleInstanceCommon *module_inst,
+                             WASMSharedHeap *shared_heap, bool is_memory64)
+{
+#if WASM_ENABLE_INTERP != 0
+    if (module_inst->module_type == Wasm_Module_Bytecode) {
+        WASMModuleInstanceExtra *e =
+            (WASMModuleInstanceExtra *)((WASMModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        if (is_memory64)
+            e->shared_heap_start_off.u64 = shared_heap->start_off_mem64;
+        else
+            e->shared_heap_start_off.u64 = shared_heap->start_off_mem32;
+        e->shared_heap_end_off.u64 =
+            e->shared_heap_start_off.u64 - 1 + shared_heap->size;
+        e->shared_heap_base_addr_adj =
+            shared_heap->base_addr - e->shared_heap_start_off.u64;
+#else
+        e->shared_heap_start_off.u32[0] = (uint32)shared_heap->start_off_mem32;
+        e->shared_heap_end_off.u32[0] =
+            e->shared_heap_start_off.u32[0] - 1 + shared_heap->size;
+        e->shared_heap_base_addr_adj =
+            shared_heap->base_addr - e->shared_heap_start_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_INTERP != 0 */
+#if WASM_ENABLE_AOT != 0
+    if (module_inst->module_type == Wasm_Module_AoT) {
+        AOTModuleInstanceExtra *e =
+            (AOTModuleInstanceExtra *)((AOTModuleInstance *)module_inst)->e;
+#if UINTPTR_MAX == UINT64_MAX
+        if (is_memory64)
+            e->shared_heap_start_off.u64 = shared_heap->start_off_mem64;
+        else
+            e->shared_heap_start_off.u64 = shared_heap->start_off_mem32;
+        e->shared_heap_end_off.u64 =
+            e->shared_heap_start_off.u64 - 1 + shared_heap->size;
+        e->shared_heap_base_addr_adj =
+            shared_heap->base_addr - e->shared_heap_start_off.u64;
+#else
+        e->shared_heap_start_off.u32[0] = (uint32)shared_heap->start_off_mem32;
+        e->shared_heap_end_off.u32[0] =
+            e->shared_heap_start_off.u32[0] - 1 + shared_heap->size;
+        e->shared_heap_base_addr_adj =
+            shared_heap->base_addr - e->shared_heap_start_off.u32[0];
+#endif
+    }
+#endif /* end of WASM_ENABLE_AOT != 0 */
+}
+
 bool
 wasm_runtime_attach_shared_heap_internal(WASMModuleInstanceCommon *module_inst,
                                          WASMSharedHeap *shared_heap)
@@ -263,20 +495,6 @@ wasm_runtime_attach_shared_heap_internal(WASMModuleInstanceCommon *module_inst,
             return false;
         }
         e->shared_heap = shared_heap;
-#if WASM_ENABLE_JIT != 0
-#if UINTPTR_MAX == UINT64_MAX
-        if (memory->is_memory64)
-            e->shared_heap_start_off.u64 = shared_heap->start_off_mem64;
-        else
-            e->shared_heap_start_off.u64 = shared_heap->start_off_mem32;
-        e->shared_heap_base_addr_adj =
-            shared_heap->base_addr - e->shared_heap_start_off.u64;
-#else
-        e->shared_heap_start_off.u32[0] = (uint32)shared_heap->start_off_mem32;
-        e->shared_heap_base_addr_adj =
-            shared_heap->base_addr - e->shared_heap_start_off.u32[0];
-#endif
-#endif /* end of WASM_ENABLE_JIT != 0 */
     }
 #endif /* end of WASM_ENABLE_INTERP != 0 */
 #if WASM_ENABLE_AOT != 0
@@ -288,21 +506,15 @@ wasm_runtime_attach_shared_heap_internal(WASMModuleInstanceCommon *module_inst,
             return false;
         }
         e->shared_heap = shared_heap;
-#if UINTPTR_MAX == UINT64_MAX
-        if (memory->is_memory64)
-            e->shared_heap_start_off.u64 = shared_heap->start_off_mem64;
-        else
-            e->shared_heap_start_off.u64 = shared_heap->start_off_mem32;
-        e->shared_heap_base_addr_adj =
-            shared_heap->base_addr - e->shared_heap_start_off.u64;
-#else
-        e->shared_heap_start_off.u32[0] = (uint32)shared_heap->start_off_mem32;
-        e->shared_heap_base_addr_adj =
-            shared_heap->base_addr - e->shared_heap_start_off.u32[0];
-#endif
     }
 #endif /* end of WASM_ENABLE_AOT != 0 */
+    update_last_used_shared_heap(module_inst, shared_heap, memory->is_memory64);
 
+    os_mutex_lock(&shared_heap_list_lock);
+    shared_heap->attached_count++;
+    os_mutex_unlock(&shared_heap_list_lock);
+    LOG_VERBOSE("Shared heap %p is attached to module instance %p", shared_heap,
+                module_inst);
     return true;
 }
 
@@ -320,34 +532,51 @@ wasm_runtime_attach_shared_heap(WASMModuleInstanceCommon *module_inst,
 void
 wasm_runtime_detach_shared_heap_internal(WASMModuleInstanceCommon *module_inst)
 {
+    /* Reset shared_heap_end_off = UINT64/32_MAX - 1 to handling a corner case,
+      app_offset >= shared_heap_start && app_offset <= shared_heap_end-bytes+1
+      when bytes=1 and both e->shared_heap_start_off and e->shared_heap_end_off
+      is 0xffffffff */
 #if WASM_ENABLE_INTERP != 0
     if (module_inst->module_type == Wasm_Module_Bytecode) {
         WASMModuleInstanceExtra *e =
             (WASMModuleInstanceExtra *)((WASMModuleInstance *)module_inst)->e;
+        if (e->shared_heap != NULL) {
+            os_mutex_lock(&shared_heap_list_lock);
+            e->shared_heap->attached_count--;
+            os_mutex_unlock(&shared_heap_list_lock);
+        }
         e->shared_heap = NULL;
-#if WASM_ENABLE_JIT != 0
 #if UINTPTR_MAX == UINT64_MAX
         e->shared_heap_start_off.u64 = UINT64_MAX;
+        e->shared_heap_end_off.u64 = UINT64_MAX - 1;
 #else
         e->shared_heap_start_off.u32[0] = UINT32_MAX;
+        e->shared_heap_end_off.u32[0] = UINT32_MAX - 1;
 #endif
         e->shared_heap_base_addr_adj = NULL;
-#endif
     }
 #endif /* end of WASM_ENABLE_INTERP != 0 */
 #if WASM_ENABLE_AOT != 0
     if (module_inst->module_type == Wasm_Module_AoT) {
         AOTModuleInstanceExtra *e =
             (AOTModuleInstanceExtra *)((AOTModuleInstance *)module_inst)->e;
+        if (e->shared_heap != NULL) {
+            os_mutex_lock(&shared_heap_list_lock);
+            e->shared_heap->attached_count--;
+            os_mutex_unlock(&shared_heap_list_lock);
+        }
         e->shared_heap = NULL;
 #if UINTPTR_MAX == UINT64_MAX
         e->shared_heap_start_off.u64 = UINT64_MAX;
+        e->shared_heap_end_off.u64 = UINT64_MAX - 1;
 #else
         e->shared_heap_start_off.u32[0] = UINT32_MAX;
+        e->shared_heap_end_off.u32[0] = UINT32_MAX - 1;
 #endif
         e->shared_heap_base_addr_adj = NULL;
     }
 #endif /* end of WASM_ENABLE_AOT != 0 */
+    LOG_VERBOSE("Shared heap is detached from module instance %p", module_inst);
 }
 
 void
@@ -385,71 +614,93 @@ wasm_runtime_get_shared_heap(WASMModuleInstanceCommon *module_inst_comm)
     return get_shared_heap(module_inst_comm);
 }
 
-static bool
+bool
 is_app_addr_in_shared_heap(WASMModuleInstanceCommon *module_inst,
                            bool is_memory64, uint64 app_offset, uint32 bytes)
 {
-    WASMSharedHeap *heap = get_shared_heap(module_inst);
+    WASMSharedHeap *heap = get_shared_heap(module_inst), *cur;
+    uint64 shared_heap_start, shared_heap_end;
 
     if (!heap) {
-        return false;
+        goto fail;
     }
 
     if (bytes == 0) {
         bytes = 1;
     }
 
-    if (!is_memory64) {
+    shared_heap_start =
-        if (app_offset >= heap->start_off_mem32
+        (uint64)get_last_used_shared_heap_start_offset(module_inst);
-            && app_offset <= UINT32_MAX - bytes + 1) {
+    shared_heap_end = (uint64)get_last_used_shared_heap_end_offset(module_inst);
+    if (bytes - 1 <= shared_heap_end && app_offset >= shared_heap_start
+        && app_offset <= shared_heap_end - bytes + 1) {
         return true;
     }
+
+    /* Early stop for app start address not in the shared heap(chain) at all */
+    shared_heap_start =
+        is_memory64 ? heap->start_off_mem64 : heap->start_off_mem32;
+    shared_heap_end = is_memory64 ? UINT64_MAX : UINT32_MAX;
+    if (bytes - 1 > shared_heap_end || app_offset < shared_heap_start
+        || app_offset > shared_heap_end - bytes + 1) {
+        goto fail;
     }
-    else {
+
-        if (app_offset >= heap->start_off_mem64
+    /* Find the exact shared heap that app addr is in, and update last used
-            && app_offset <= UINT64_MAX - bytes + 1) {
+     * shared heap info in module inst extra */
+    for (cur = heap; cur; cur = cur->chain_next) {
+        shared_heap_start =
+            is_memory64 ? cur->start_off_mem64 : cur->start_off_mem32;
+        shared_heap_end = shared_heap_start - 1 + cur->size;
+        if (bytes - 1 <= shared_heap_end && app_offset >= shared_heap_start
+            && app_offset <= shared_heap_end - bytes + 1) {
+            update_last_used_shared_heap(module_inst, cur, is_memory64);
             return true;
         }
     }
 
+fail:
     return false;
 }
 
 static bool
 is_native_addr_in_shared_heap(WASMModuleInstanceCommon *module_inst,
-                              uint8 *addr, uint32 bytes)
+                              bool is_memory64, uint8 *addr, uint32 bytes)
 {
-    WASMSharedHeap *heap = get_shared_heap(module_inst);
+    WASMSharedHeap *cur, *heap = get_shared_heap(module_inst);
-    uintptr_t base_addr;
+    uintptr_t base_addr, addr_int, end_addr;
-    uintptr_t addr_int;
-    uintptr_t end_addr;
 
     if (!heap) {
-        return false;
+        goto fail;
     }
 
-    base_addr = (uintptr_t)heap->base_addr;
+    /* Iterate through shared heap chain to find whether native addr in one of
+     * shared heap */
+    for (cur = heap; cur != NULL; cur = cur->chain_next) {
+        base_addr = (uintptr_t)cur->base_addr;
         addr_int = (uintptr_t)addr;
-    if (addr_int < base_addr) {
+        if (addr_int < base_addr)
-        return false;
+            continue;
-    }
 
         end_addr = addr_int + bytes;
         /* Check for overflow */
-    if (end_addr <= addr_int) {
+        if (end_addr <= addr_int)
-        return false;
+            continue;
-    }
 
-    if (end_addr > base_addr + heap->size) {
+        if (end_addr > base_addr + cur->size)
-        return false;
+            continue;
-    }
 
+        update_last_used_shared_heap(module_inst, cur, is_memory64);
         return true;
+    }
+
+fail:
+    return false;
 }
 
 uint64
 wasm_runtime_shared_heap_malloc(WASMModuleInstanceCommon *module_inst,
-                                uint64_t size, void **p_native_addr)
+                                uint64 size, void **p_native_addr)
 {
     WASMMemoryInstance *memory =
         wasm_get_default_memory((WASMModuleInstance *)module_inst);
@@ -459,6 +710,14 @@ wasm_runtime_shared_heap_malloc(WASMModuleInstanceCommon *module_inst,
     if (!memory || !shared_heap)
         return 0;
 
+    while (shared_heap && !shared_heap->heap_handle) {
+        shared_heap = shared_heap->chain_next;
+    }
+    if (!shared_heap) {
+        LOG_WARNING("Can't allocate from pre allocated shared heap");
+        return 0;
+    }
+
     native_addr = mem_allocator_malloc(shared_heap->heap_handle, size);
     if (!native_addr)
         return 0;
@@ -467,11 +726,9 @@ wasm_runtime_shared_heap_malloc(WASMModuleInstanceCommon *module_inst,
         *p_native_addr = native_addr;
     }
 
-    if (memory->is_memory64)
+    return memory->is_memory64
-        return shared_heap->start_off_mem64
+               ? shared_heap->start_off_mem64
-               + ((uint8 *)native_addr - shared_heap->base_addr);
+               : shared_heap->start_off_mem32
-    else
-        return shared_heap->start_off_mem32
                      + ((uint8 *)native_addr - shared_heap->base_addr);
 }
 
@@ -487,6 +744,14 @@ wasm_runtime_shared_heap_free(WASMModuleInstanceCommon *module_inst, uint64 ptr)
         return;
     }
 
+    while (shared_heap && !shared_heap->heap_handle) {
+        shared_heap = shared_heap->chain_next;
+    }
+    if (!shared_heap) {
+        LOG_WARNING("The address to free is from pre allocated shared heap");
+        return;
+    }
+
     if (memory->is_memory64) {
         if (ptr < shared_heap->start_off_mem64) { /* ptr can not > UINT64_MAX */
             LOG_WARNING("The address to free isn't in shared heap");
@@ -564,6 +829,7 @@ destroy_shared_heaps()
     while (heap) {
         cur = heap;
         heap = heap->next;
+        if (cur->heap_handle) {
             mem_allocator_destroy(cur->heap_handle);
             wasm_runtime_free(cur->heap_handle);
 #ifndef OS_ENABLE_HW_BOUND_CHECK
@@ -572,6 +838,7 @@ destroy_shared_heaps()
             map_size = 8 * (uint64)BH_GB;
 #endif
             wasm_munmap_linear_memory(cur->base_addr, cur->size, map_size);
+        }
         wasm_runtime_free(cur);
     }
     os_mutex_destroy(&shared_heap_list_lock);
@@ -798,6 +1065,10 @@ wasm_runtime_validate_app_str_addr(WASMModuleInstanceCommon *module_inst_comm,
     WASMMemoryInstance *memory_inst;
     uint64 app_end_offset, max_linear_memory_size = MAX_LINEAR_MEMORY_SIZE;
     char *str, *str_end;
+#if WASM_ENABLE_SHARED_HEAP != 0
+    uintptr_t shared_heap_end_off;
+    char *shared_heap_base_addr_adj;
+#endif
 
     bh_assert(module_inst_comm->module_type == Wasm_Module_Bytecode
               || module_inst_comm->module_type == Wasm_Module_AoT);
@@ -814,12 +1085,12 @@ wasm_runtime_validate_app_str_addr(WASMModuleInstanceCommon *module_inst_comm,
 #if WASM_ENABLE_SHARED_HEAP != 0
     if (is_app_addr_in_shared_heap(module_inst_comm, memory_inst->is_memory64,
                                    app_str_offset, 1)) {
-        WASMSharedHeap *shared_heap = get_shared_heap(module_inst_comm);
+        shared_heap_end_off =
-        str = (char *)shared_heap->base_addr
+            get_last_used_shared_heap_end_offset(module_inst_comm);
-              + (memory_inst->is_memory64
+        shared_heap_base_addr_adj =
-                     ? (app_str_offset - shared_heap->start_off_mem64)
+            (char *)get_last_used_shared_heap_base_addr_adj(module_inst_comm);
-                     : (app_str_offset - shared_heap->start_off_mem32));
+        str = shared_heap_base_addr_adj + app_str_offset;
-        str_end = (char *)shared_heap->base_addr + shared_heap->size;
+        str_end = shared_heap_base_addr_adj + shared_heap_end_off + 1;
     }
     else
 #endif
@@ -884,7 +1155,8 @@ wasm_runtime_validate_native_addr(WASMModuleInstanceCommon *module_inst_comm,
     }
 
 #if WASM_ENABLE_SHARED_HEAP != 0
-    if (is_native_addr_in_shared_heap(module_inst_comm, native_ptr, size)) {
+    if (is_native_addr_in_shared_heap(
+            module_inst_comm, memory_inst->is_memory64, native_ptr, size)) {
         return true;
     }
 #endif
@@ -926,17 +1198,8 @@ wasm_runtime_addr_app_to_native(WASMModuleInstanceCommon *module_inst_comm,
 #if WASM_ENABLE_SHARED_HEAP != 0
     if (is_app_addr_in_shared_heap(module_inst_comm, memory_inst->is_memory64,
                                    app_offset, 1)) {
-        WASMSharedHeap *shared_heap = get_shared_heap(module_inst_comm);
+        return get_last_used_shared_heap_base_addr_adj(module_inst_comm)
-        uint64 shared_heap_start = 0;
+               + app_offset;
-
-        if (memory_inst && !memory_inst->is_memory64) {
-            shared_heap_start = shared_heap->start_off_mem32;
-        }
-        else if (memory_inst && memory_inst->is_memory64) {
-            shared_heap_start = shared_heap->start_off_mem64;
-        }
-
-        return shared_heap->base_addr + app_offset - shared_heap_start;
     }
 #endif
 
@@ -974,29 +1237,17 @@ wasm_runtime_addr_native_to_app(WASMModuleInstanceCommon *module_inst_comm,
 
     bounds_checks = is_bounds_checks_enabled(module_inst_comm);
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-    /* If shared heap is enabled, bounds check is always needed */
-    bounds_checks = true;
-#endif
-
     memory_inst = wasm_get_default_memory(module_inst);
     if (!memory_inst) {
         return 0;
     }
 
 #if WASM_ENABLE_SHARED_HEAP != 0
-    if (is_native_addr_in_shared_heap(module_inst_comm, addr, 1)) {
+    if (is_native_addr_in_shared_heap(module_inst_comm,
-        WASMSharedHeap *shared_heap = get_shared_heap(module_inst_comm);
+                                      memory_inst->is_memory64, addr, 1)) {
-        uint64 shared_heap_start = 0;
+        return (uint64)(uintptr_t)(addr
-
+                                   - get_last_used_shared_heap_base_addr_adj(
-        if (memory_inst && !memory_inst->is_memory64) {
+                                       module_inst_comm));
-            shared_heap_start = shared_heap->start_off_mem32;
-        }
-        else if (memory_inst && memory_inst->is_memory64) {
-            shared_heap_start = shared_heap->start_off_mem64;
-        }
-
-        return shared_heap_start + (addr - shared_heap->base_addr);
     }
 #endif
 
@@ -1098,8 +1349,8 @@ wasm_check_app_addr_and_convert(WASMModuleInstance *module_inst, bool is_str,
     uint8 *native_addr;
     bool bounds_checks;
 #if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap;
+    uint8 *shared_heap_base_addr_adj = NULL;
-    bool is_in_shared_heap = false;
+    uintptr_t shared_heap_end_off = 0;
 #endif
 
     bh_assert(app_buf_addr <= UINTPTR_MAX && app_buf_size <= UINTPTR_MAX);
@@ -1113,36 +1364,17 @@ wasm_check_app_addr_and_convert(WASMModuleInstance *module_inst, bool is_str,
     if (is_app_addr_in_shared_heap((WASMModuleInstanceCommon *)module_inst,
                                    memory_inst->is_memory64, app_buf_addr,
                                    app_buf_size)) {
-        shared_heap = get_shared_heap((WASMModuleInstanceCommon *)module_inst);
-        native_addr = shared_heap->base_addr
-                      + (memory_inst->is_memory64
-                             ? (app_buf_addr - shared_heap->start_off_mem64)
-                             : (app_buf_addr - shared_heap->start_off_mem32));
-        is_in_shared_heap = true;
-    }
-    else
-#endif
-    {
-        native_addr = memory_inst->memory_data + (uintptr_t)app_buf_addr;
-    }
-
-    bounds_checks =
-        is_bounds_checks_enabled((WASMModuleInstanceCommon *)module_inst);
-
-    if (!bounds_checks) {
-        if (app_buf_addr == 0) {
-            native_addr = NULL;
-        }
-        goto success;
-    }
-
-#if WASM_ENABLE_SHARED_HEAP != 0
-    if (is_in_shared_heap) {
         const char *str, *str_end;
+        shared_heap_base_addr_adj = get_last_used_shared_heap_base_addr_adj(
+            (WASMModuleInstanceCommon *)module_inst);
+        shared_heap_end_off = get_last_used_shared_heap_end_offset(
+            (WASMModuleInstanceCommon *)module_inst);
+        native_addr = shared_heap_base_addr_adj + (uintptr_t)app_buf_addr;
 
-        /* The whole string must be in the linear memory */
+        /* The whole string must be in the shared heap */
         str = (const char *)native_addr;
-        str_end = (const char *)shared_heap->base_addr + shared_heap->size;
+        str_end =
+            (const char *)shared_heap_base_addr_adj + shared_heap_end_off + 1;
         while (str < str_end && *str != '\0')
             str++;
         if (str == str_end) {
@@ -1154,6 +1386,17 @@ wasm_check_app_addr_and_convert(WASMModuleInstance *module_inst, bool is_str,
     }
 #endif
 
+    native_addr = memory_inst->memory_data + (uintptr_t)app_buf_addr;
+    bounds_checks =
+        is_bounds_checks_enabled((WASMModuleInstanceCommon *)module_inst);
+
+    if (!bounds_checks) {
+        if (app_buf_addr == 0) {
+            native_addr = NULL;
+        }
+        goto success;
+    }
+
     /* No need to check the app_offset and buf_size if memory access
        boundary check with hardware trap is enabled */
 #ifndef OS_ENABLE_HW_BOUND_CHECK

core/iwasm/common/wasm_memory.h

@@ -41,10 +41,60 @@ SET_LINEAR_MEMORY_SIZE(WASMMemoryInstance *memory, uint64 size)
 #define SET_LINEAR_MEMORY_SIZE(memory, size) memory->memory_data_size = size
 #endif
 
+#if WASM_ENABLE_INTERP != 0
 #if WASM_ENABLE_SHARED_HEAP != 0
+
+#if WASM_ENABLE_MULTI_MEMORY != 0
+/* Only enable shared heap for the default memory */
+#define is_default_memory (memidx == 0)
+#else
+#define is_default_memory true
+#endif
+
+#if UINTPTR_MAX == UINT64_MAX
+#define get_shared_heap_end_off() module->e->shared_heap_end_off.u64
+#else
+#define get_shared_heap_end_off() \
+    (uint64)(module->e->shared_heap_end_off.u32[0])
+#endif
+
+#if WASM_ENABLE_MEMORY64 != 0
+#define shared_heap_is_memory64 is_memory64
+#else
+#define shared_heap_is_memory64 false
+#endif
+
+#define app_addr_in_shared_heap(app_addr, bytes)                              \
+    (is_default_memory                                                        \
+     && is_app_addr_in_shared_heap((WASMModuleInstanceCommon *)module,        \
+                                   shared_heap_is_memory64, (uint64)app_addr, \
+                                   bytes))
+#define shared_heap_addr_app_to_native(app_addr, native_addr) \
+    native_addr = module->e->shared_heap_base_addr_adj + app_addr
+#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr) \
+    if (app_addr_in_shared_heap(app_addr, bytes))                \
+        shared_heap_addr_app_to_native(app_addr, native_addr);   \
+    else
+
+#else /* else of WASM_ENABLE_SHARED_HEAP != 0 */
+#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr)
+#endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
+#endif /* end of WASM_ENABLE_INTERP != 0 */
+
+#if WASM_ENABLE_SHARED_HEAP != 0
+bool
+is_app_addr_in_shared_heap(WASMModuleInstanceCommon *module_inst,
+                           bool is_memory64, uint64 app_offset, uint32 bytes);
+
 WASMSharedHeap *
 wasm_runtime_create_shared_heap(SharedHeapInitArgs *init_args);
 
+WASMSharedHeap *
+wasm_runtime_chain_shared_heaps(WASMSharedHeap *head, WASMSharedHeap *body);
+
+WASMSharedHeap *
+wasm_runtime_unchain_shared_heaps(WASMSharedHeap *head, bool entire_chain);
+
 bool
 wasm_runtime_attach_shared_heap(WASMModuleInstanceCommon *module_inst,
                                 WASMSharedHeap *shared_heap);
@@ -68,7 +118,7 @@ wasm_runtime_shared_heap_malloc(WASMModuleInstanceCommon *module_inst,
 void
 wasm_runtime_shared_heap_free(WASMModuleInstanceCommon *module_inst,
                               uint64 ptr);
-#endif
+#endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 
 bool
 wasm_runtime_memory_init(mem_alloc_type_t mem_alloc_type,

core/iwasm/common/wasm_runtime_common.c

@@ -1654,14 +1654,70 @@ wasm_runtime_instantiate(WASMModuleCommon *module, uint32 stack_size,
                                              error_buf_size);
 }
 
+static void
+instantiation_args_set_defaults(struct InstantiationArgs2 *args)
+{
+    memset(args, 0, sizeof(*args));
+}
+
 WASMModuleInstanceCommon *
 wasm_runtime_instantiate_ex(WASMModuleCommon *module,
                             const InstantiationArgs *args, char *error_buf,
                             uint32 error_buf_size)
+{
+    struct InstantiationArgs2 v2;
+    instantiation_args_set_defaults(&v2);
+    v2.v1 = *args;
+    return wasm_runtime_instantiate_ex2(module, &v2, error_buf, error_buf_size);
+}
+
+bool
+wasm_runtime_instantiation_args_create(struct InstantiationArgs2 **p)
+{
+    struct InstantiationArgs2 *args = wasm_runtime_malloc(sizeof(*args));
+    if (args == NULL) {
+        return false;
+    }
+    instantiation_args_set_defaults(args);
+    *p = args;
+    return true;
+}
+
+void
+wasm_runtime_instantiation_args_destroy(struct InstantiationArgs2 *p)
+{
+    wasm_runtime_free(p);
+}
+
+void
+wasm_runtime_instantiation_args_set_default_stack_size(
+    struct InstantiationArgs2 *p, uint32 v)
+{
+    p->v1.default_stack_size = v;
+}
+
+void
+wasm_runtime_instantiation_args_set_host_managed_heap_size(
+    struct InstantiationArgs2 *p, uint32 v)
+{
+    p->v1.host_managed_heap_size = v;
+}
+
+void
+wasm_runtime_instantiation_args_set_max_memory_pages(
+    struct InstantiationArgs2 *p, uint32 v)
+{
+    p->v1.max_memory_pages = v;
+}
+
+WASMModuleInstanceCommon *
+wasm_runtime_instantiate_ex2(WASMModuleCommon *module,
+                             const struct InstantiationArgs2 *args,
+                             char *error_buf, uint32 error_buf_size)
 {
     return wasm_runtime_instantiate_internal(
-        module, NULL, NULL, args->default_stack_size,
+        module, NULL, NULL, args->v1.default_stack_size,
-        args->host_managed_heap_size, args->max_memory_pages, error_buf,
+        args->v1.host_managed_heap_size, args->v1.max_memory_pages, error_buf,
         error_buf_size);
 }
 
@@ -1743,9 +1799,9 @@ wasm_runtime_destroy_exec_env(WASMExecEnv *exec_env)
     wasm_exec_env_destroy(exec_env);
 }
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 uint32
-wasm_copy_callstack(const wasm_exec_env_t exec_env, wasm_frame_t *buffer,
+wasm_copy_callstack(const wasm_exec_env_t exec_env, WASMCApiFrame *buffer,
                     const uint32 length, const uint32 skip_n, char *error_buf,
                     uint32_t error_buf_size)
 {
@@ -1780,7 +1836,7 @@ wasm_copy_callstack(const wasm_exec_env_t exec_env, wasm_frame_t *buffer,
     strncpy(error_buf, err_msg, error_buf_size);
     return 0;
 }
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 bool
 wasm_runtime_init_thread_env(void)
@@ -3754,7 +3810,15 @@ wasm_runtime_init_wasi(WASMModuleInstanceCommon *module_inst,
         address = strtok(cp, "/");
         mask = strtok(NULL, "/");
 
-        ret = addr_pool_insert(apool, address, (uint8)(mask ? atoi(mask) : 0));
+        if (!mask) {
+            snprintf(error_buf, error_buf_size,
+                     "Invalid address pool entry: %s, must be in the format of "
+                     "ADDRESS/MASK",
+                     addr_pool[i]);
+            goto fail;
+        }
+
+        ret = addr_pool_insert(apool, address, (uint8)atoi(mask));
         wasm_runtime_free(cp);
         if (!ret) {
             set_error_buf(error_buf, error_buf_size,
@@ -7883,3 +7947,37 @@ wasm_runtime_is_underlying_binary_freeable(WASMModuleCommon *const module)
 
     return true;
 }
+
+#if WASM_ENABLE_SHARED_HEAP != 0
+bool
+wasm_runtime_check_and_update_last_used_shared_heap(
+    WASMModuleInstanceCommon *module_inst, uintptr_t app_offset, size_t bytes,
+    uintptr_t *shared_heap_start_off_p, uintptr_t *shared_heap_end_off_p,
+    uint8 **shared_heap_base_addr_adj_p, bool is_memory64)
+{
+    WASMSharedHeap *heap = wasm_runtime_get_shared_heap(module_inst), *cur;
+    uint64 shared_heap_start, shared_heap_end;
+
+    if (bytes == 0) {
+        bytes = 1;
+    }
+
+    /* Find the exact shared heap that app addr is in, and update last used
+     * shared heap info in func context */
+    for (cur = heap; cur; cur = cur->chain_next) {
+        shared_heap_start =
+            is_memory64 ? cur->start_off_mem64 : cur->start_off_mem32;
+        shared_heap_end = shared_heap_start - 1 + cur->size;
+        if (bytes - 1 <= shared_heap_end && app_offset >= shared_heap_start
+            && app_offset <= shared_heap_end - bytes + 1) {
+            *shared_heap_start_off_p = (uintptr_t)shared_heap_start;
+            *shared_heap_end_off_p = (uintptr_t)shared_heap_end;
+            *shared_heap_base_addr_adj_p =
+                cur->base_addr - (uintptr_t)shared_heap_start;
+            return true;
+        }
+    }
+
+    return false;
+}
+#endif

core/iwasm/common/wasm_runtime_common.h

@@ -612,6 +612,10 @@ WASMExecEnv *
 wasm_runtime_get_exec_env_tls(void);
 #endif
 
+struct InstantiationArgs2 {
+    InstantiationArgs v1;
+};
+
 /* See wasm_export.h for description */
 WASM_RUNTIME_API_EXTERN bool
 wasm_runtime_init(void);
@@ -700,6 +704,40 @@ wasm_runtime_instantiate_ex(WASMModuleCommon *module,
                             const InstantiationArgs *args, char *error_buf,
                             uint32 error_buf_size);
 
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN
+bool
+wasm_runtime_instantiation_args_create(struct InstantiationArgs2 **p);
+
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN
+void
+wasm_runtime_instantiation_args_destroy(struct InstantiationArgs2 *p);
+
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN
+void
+wasm_runtime_instantiation_args_set_default_stack_size(
+    struct InstantiationArgs2 *p, uint32 v);
+
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN
+void
+wasm_runtime_instantiation_args_set_host_managed_heap_size(
+    struct InstantiationArgs2 *p, uint32 v);
+
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN
+void
+wasm_runtime_instantiation_args_set_max_memory_pages(
+    struct InstantiationArgs2 *p, uint32 v);
+
+/* See wasm_export.h for description */
+WASM_RUNTIME_API_EXTERN WASMModuleInstanceCommon *
+wasm_runtime_instantiate_ex2(WASMModuleCommon *module,
+                             const struct InstantiationArgs2 *args,
+                             char *error_buf, uint32 error_buf_size);
+
 /* See wasm_export.h for description */
 WASM_RUNTIME_API_EXTERN bool
 wasm_runtime_set_running_mode(wasm_module_inst_t module_inst,
@@ -758,12 +796,12 @@ wasm_runtime_create_exec_env(WASMModuleInstanceCommon *module_inst,
 WASM_RUNTIME_API_EXTERN void
 wasm_runtime_destroy_exec_env(WASMExecEnv *exec_env);
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 WASM_RUNTIME_API_EXTERN uint32_t
-wasm_copy_callstack(const wasm_exec_env_t exec_env, wasm_frame_t *buffer,
+wasm_copy_callstack(const wasm_exec_env_t exec_env, WASMCApiFrame *buffer,
                     const uint32 length, const uint32 skip_n, char *error_buf,
                     uint32 error_buf_size);
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 /* See wasm_export.h for description */
 WASM_RUNTIME_API_EXTERN WASMModuleInstanceCommon *
@@ -1336,6 +1374,14 @@ void
 wasm_runtime_set_linux_perf(bool flag);
 #endif
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+bool
+wasm_runtime_check_and_update_last_used_shared_heap(
+    WASMModuleInstanceCommon *module_inst, uintptr_t app_offset, size_t bytes,
+    uintptr_t *shared_heap_start_off_p, uintptr_t *shared_heap_end_off_p,
+    uint8 **shared_heap_base_addr_adj_p, bool is_memory64);
+#endif
+
 #ifdef __cplusplus
 }
 #endif

core/iwasm/compilation/aot.c

@@ -8,7 +8,7 @@
 static char aot_error[128];
 
 char *
-aot_get_last_error()
+aot_get_last_error(void)
 {
     return aot_error[0] == '\0' ? "" : aot_error;
 }

core/iwasm/compilation/aot_compiler.c

@@ -1316,13 +1316,6 @@ aot_compile_func(AOTCompContext *comp_ctx, uint32 func_index)
                     return false;
                 break;
 
-#if WASM_ENABLE_SIMD != 0
-            case WASM_OP_SELECT_128:
-                if (!aot_compile_op_select(comp_ctx, func_ctx, true))
-                    return false;
-                break;
-#endif
-
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
             case WASM_OP_SELECT_T:
             {

core/iwasm/compilation/aot_emit_aot_file.c

@@ -48,7 +48,7 @@ typedef struct AOTSymbolList {
 } AOTSymbolList;
 
 /* AOT object data */
-typedef struct AOTObjectData {
+struct AOTObjectData {
     AOTCompContext *comp_ctx;
 
     LLVMMemoryBufferRef mem_buf;
@@ -82,7 +82,7 @@ typedef struct AOTObjectData {
     const char *stack_sizes_section_name;
     uint32 stack_sizes_offset;
     uint32 *stack_sizes;
-} AOTObjectData;
+};
 
 #if 0
 static void dump_buf(uint8 *buf, uint32 size, char *title)
@@ -216,7 +216,7 @@ get_init_expr_size(const AOTCompContext *comp_ctx, const AOTCompData *comp_data,
 #if WASM_ENABLE_GC != 0
     WASMModule *module = comp_data->wasm_module;
 #endif
-
+    bh_assert(expr != NULL);
     /* + init value size */
     switch (expr->init_expr_type) {
         case INIT_EXPR_NONE:
@@ -248,7 +248,7 @@ get_init_expr_size(const AOTCompContext *comp_ctx, const AOTCompData *comp_data,
         {
             uint32 i;
             WASMStructNewInitValues *struct_new_init_values =
-                (WASMStructNewInitValues *)expr->u.data;
+                (WASMStructNewInitValues *)expr->u.unary.v.data;
 
             /* type_index + field_count + fields */
             size += sizeof(uint32) + sizeof(uint32);
@@ -285,7 +285,7 @@ get_init_expr_size(const AOTCompContext *comp_ctx, const AOTCompData *comp_data,
         case INIT_EXPR_TYPE_ARRAY_NEW_FIXED:
         {
             WASMArrayNewInitValues *array_new_init_values =
-                (WASMArrayNewInitValues *)expr->u.data;
+                (WASMArrayNewInitValues *)expr->u.unary.v.data;
             WASMArrayType *array_type = NULL;
             uint32 value_count;
 
@@ -302,12 +302,27 @@ get_init_expr_size(const AOTCompContext *comp_ctx, const AOTCompData *comp_data,
 
             /* array_elem_type + type_index + len + elems */
             size += sizeof(uint32) * 3
-                    + wasm_value_type_size_internal(array_type->elem_type,
+                    + (uint64)wasm_value_type_size_internal(
-                                                    comp_ctx->pointer_size)
+                          array_type->elem_type, comp_ctx->pointer_size)
                           * value_count;
             break;
         }
 #endif /* end of WASM_ENABLE_GC != 0 */
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        case INIT_EXPR_TYPE_I32_ADD:
+        case INIT_EXPR_TYPE_I32_SUB:
+        case INIT_EXPR_TYPE_I32_MUL:
+        case INIT_EXPR_TYPE_I64_ADD:
+        case INIT_EXPR_TYPE_I64_SUB:
+        case INIT_EXPR_TYPE_I64_MUL:
+        {
+            size +=
+                get_init_expr_size(comp_ctx, comp_data, expr->u.binary.l_expr);
+            size +=
+                get_init_expr_size(comp_ctx, comp_data, expr->u.binary.r_expr);
+            break;
+        }
+#endif
         default:
             bh_assert(0);
     }
@@ -324,15 +339,16 @@ get_table_init_data_size(AOTCompContext *comp_ctx,
     /*
      * mode (4 bytes), elem_type (4 bytes)
      *
-     * table_index(4 bytes) + init expr type (4 bytes) + init expr value (8
+     * table_index(4 bytes)
-     * bytes)
      */
-    size = (uint32)(sizeof(uint32) * 2 + sizeof(uint32) + sizeof(uint32)
+    size = (uint32)(sizeof(uint32) * 2 + sizeof(uint32))
-                    + sizeof(uint64))
            /* Size of WasmRefType - inner padding (ref type + nullable +
               heap_type) */
            + 8;
 
+    size += get_init_expr_size(comp_ctx, comp_ctx->comp_data,
+                               &table_init_data->offset);
+
     /* + value count/func index count (4 bytes) + init_values */
     size += sizeof(uint32);
     for (i = 0; i < table_init_data->value_count; i++) {
@@ -1811,6 +1827,10 @@ static bool
 aot_emit_init_expr(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
                    AOTCompContext *comp_ctx, InitializerExpression *expr)
 {
+    if (expr == NULL) {
+        aot_set_last_error("invalid init expr.");
+        return false;
+    }
     uint32 offset = *p_offset;
 #if WASM_ENABLE_GC != 0
     WASMModule *module = comp_ctx->comp_data->wasm_module;
@@ -1824,31 +1844,31 @@ aot_emit_init_expr(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
             break;
         case INIT_EXPR_TYPE_I32_CONST:
         case INIT_EXPR_TYPE_F32_CONST:
-            EMIT_U32(expr->u.i32);
+            EMIT_U32(expr->u.unary.v.i32);
             break;
         case INIT_EXPR_TYPE_I64_CONST:
         case INIT_EXPR_TYPE_F64_CONST:
-            EMIT_U64(expr->u.i64);
+            EMIT_U64(expr->u.unary.v.i64);
             break;
         case INIT_EXPR_TYPE_V128_CONST:
-            EMIT_V128(expr->u.v128);
+            EMIT_V128(expr->u.unary.v.v128);
             break;
         case INIT_EXPR_TYPE_GET_GLOBAL:
-            EMIT_U32(expr->u.global_index);
+            EMIT_U32(expr->u.unary.v.global_index);
             break;
         case INIT_EXPR_TYPE_FUNCREF_CONST:
         case INIT_EXPR_TYPE_REFNULL_CONST:
-            EMIT_U32(expr->u.ref_index);
+            EMIT_U32(expr->u.unary.v.ref_index);
             break;
 #if WASM_ENABLE_GC != 0
         case INIT_EXPR_TYPE_I31_NEW:
-            EMIT_U32(expr->u.i32);
+            EMIT_U32(expr->u.unary.v.i32);
             break;
         case INIT_EXPR_TYPE_STRUCT_NEW:
         {
             uint32 i;
             WASMStructNewInitValues *init_values =
-                (WASMStructNewInitValues *)expr->u.data;
+                (WASMStructNewInitValues *)expr->u.unary.v.data;
             WASMStructType *struct_type = NULL;
 
             EMIT_U32(init_values->type_idx);
@@ -1879,21 +1899,21 @@ aot_emit_init_expr(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
             break;
         }
         case INIT_EXPR_TYPE_STRUCT_NEW_DEFAULT:
-            EMIT_U32(expr->u.type_index);
+            EMIT_U32(expr->u.unary.v.type_index);
             break;
         case INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT:
         {
             WASMArrayType *array_type = NULL;
 
-            bh_assert(expr->u.array_new_default.type_index
+            bh_assert(expr->u.unary.v.array_new_default.type_index
                       < module->type_count);
             array_type =
                 (WASMArrayType *)
-                    module->types[expr->u.array_new_default.type_index];
+                    module->types[expr->u.unary.v.array_new_default.type_index];
 
             EMIT_U32(array_type->elem_type);
-            EMIT_U32(expr->u.array_new_default.type_index);
+            EMIT_U32(expr->u.unary.v.array_new_default.type_index);
-            EMIT_U32(expr->u.array_new_default.length);
+            EMIT_U32(expr->u.unary.v.array_new_default.length);
             break;
         }
         case INIT_EXPR_TYPE_ARRAY_NEW:
@@ -1901,7 +1921,7 @@ aot_emit_init_expr(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
         {
             uint32 value_count, i, field_size;
             WASMArrayNewInitValues *init_values =
-                (WASMArrayNewInitValues *)expr->u.data;
+                (WASMArrayNewInitValues *)expr->u.unary.v.data;
             WASMArrayType *array_type = NULL;
 
             bh_assert(init_values->type_idx < module->type_count);
@@ -1933,6 +1953,25 @@ aot_emit_init_expr(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
             break;
         }
 #endif /* end of WASM_ENABLE_GC != 0 */
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        case INIT_EXPR_TYPE_I32_ADD:
+        case INIT_EXPR_TYPE_I32_SUB:
+        case INIT_EXPR_TYPE_I32_MUL:
+        case INIT_EXPR_TYPE_I64_ADD:
+        case INIT_EXPR_TYPE_I64_SUB:
+        case INIT_EXPR_TYPE_I64_MUL:
+            if (comp_ctx->enable_extended_const) {
+                if (!aot_emit_init_expr(buf, buf_end, &offset, comp_ctx,
+                                        expr->u.binary.l_expr)) {
+                    return false;
+                }
+                if (!aot_emit_init_expr(buf, buf_end, &offset, comp_ctx,
+                                        expr->u.binary.r_expr)) {
+                    return false;
+                }
+            }
+            break;
+#endif
         default:
             aot_set_last_error("invalid init expr type.");
             return false;
@@ -2034,8 +2073,10 @@ aot_emit_table_info(uint8 *buf, uint8 *buf_end, uint32 *p_offset,
         EMIT_U32(init_datas[i]->mode);
         EMIT_U32(init_datas[i]->elem_type);
         EMIT_U32(init_datas[i]->table_index);
-        EMIT_U32(init_datas[i]->offset.init_expr_type);
+        if (!aot_emit_init_expr(buf, buf_end, &offset, comp_ctx,
-        EMIT_U64(init_datas[i]->offset.u.i64);
+                                &init_datas[i]->offset))
+            return false;
+
 #if WASM_ENABLE_GC != 0
         if (comp_ctx->enable_gc && init_datas[i]->elem_ref_type) {
             EMIT_U16(init_datas[i]->elem_ref_type->ref_ht_common.ref_type);

core/iwasm/compilation/aot_emit_function.c

@@ -347,7 +347,8 @@ call_aot_invoke_c_api_native(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
 
     /* Get &c_api_func_imports[func_idx], note size of CApiFuncImport
        is pointer_size * 3 */
-    offset = I32_CONST((comp_ctx->pointer_size * 3) * import_func_idx);
+    offset = I32_CONST((unsigned long long)comp_ctx->pointer_size * 3
+                       * import_func_idx);
     CHECK_LLVM_CONST(offset);
     c_api_func_import =
         LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE, c_api_func_imports,

core/iwasm/compilation/aot_emit_memory.c

@@ -10,6 +10,40 @@
 #include "aot_intrinsic.h"
 #include "aot_emit_control.h"
 
+#define BUILD_IS_NOT_NULL(value, res, name)                                \
+    do {                                                                   \
+        if (!(res = LLVMBuildIsNotNull(comp_ctx->builder, value, name))) { \
+            aot_set_last_error("llvm build is not null failed.");          \
+            goto fail;                                                     \
+        }                                                                  \
+    } while (0)
+
+#define BUILD_BR(llvm_block)                               \
+    do {                                                   \
+        if (!LLVMBuildBr(comp_ctx->builder, llvm_block)) { \
+            aot_set_last_error("llvm build br failed.");   \
+            goto fail;                                     \
+        }                                                  \
+    } while (0)
+
+#define BUILD_COND_BR(value_if, block_then, block_else)               \
+    do {                                                              \
+        if (!LLVMBuildCondBr(comp_ctx->builder, value_if, block_then, \
+                             block_else)) {                           \
+            aot_set_last_error("llvm build cond br failed.");         \
+            goto fail;                                                \
+        }                                                             \
+    } while (0)
+
+#define BUILD_TRUNC(value, data_type)                                     \
+    do {                                                                  \
+        if (!(value = LLVMBuildTrunc(comp_ctx->builder, value, data_type, \
+                                     "val_trunc"))) {                     \
+            aot_set_last_error("llvm build trunc failed.");               \
+            goto fail;                                                    \
+        }                                                                 \
+    } while (0)
+
 #define BUILD_ICMP(op, left, right, res, name)                                \
     do {                                                                      \
         if (!(res =                                                           \
@@ -111,6 +145,418 @@ ffs(int n)
 static LLVMValueRef
 get_memory_curr_page_count(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx);
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+uint32
+get_module_inst_extra_offset(AOTCompContext *comp_ctx);
+
+#define BUILD_LOAD_PTR(ptr, data_type, res)                           \
+    do {                                                              \
+        if (!(res = LLVMBuildLoad2(comp_ctx->builder, data_type, ptr, \
+                                   "load_value"))) {                  \
+            aot_set_last_error("llvm build load failed");             \
+            goto fail;                                                \
+        }                                                             \
+    } while (0)
+
+/* Update last used shared heap info(alloc ptr) in function ctx:
+ * 1. shared_heap_start_off 2. shared_heap_end_off 3. shared_heap_base_addr_adj
+ */
+bool
+aot_check_shared_heap_chain_and_update(AOTCompContext *comp_ctx,
+                                       AOTFuncContext *func_ctx,
+                                       LLVMBasicBlockRef check_succ,
+                                       LLVMValueRef start_offset,
+                                       LLVMValueRef bytes, bool is_memory64)
+{
+    LLVMValueRef param_values[7], ret_value, func, value, cmp;
+    LLVMTypeRef param_types[7], ret_type, func_type, func_ptr_type;
+
+    param_types[0] = INT8_PTR_TYPE;
+    param_types[1] = INTPTR_T_TYPE;
+    param_types[2] = SIZE_T_TYPE;
+    param_types[3] = INTPTR_T_PTR_TYPE;
+    param_types[4] = INTPTR_T_PTR_TYPE;
+    param_types[5] = INT8_PTR_TYPE;
+    param_types[6] = INT8_TYPE;
+    ret_type = INT8_TYPE;
+
+    GET_AOT_FUNCTION(wasm_runtime_check_and_update_last_used_shared_heap, 7);
+
+    /* Call function */
+    param_values[0] = func_ctx->aot_inst;
+    param_values[1] = start_offset;
+    param_values[2] = bytes;
+    /* pass alloc ptr */
+    param_values[3] = func_ctx->shared_heap_start_off;
+    param_values[4] = func_ctx->shared_heap_end_off;
+    param_values[5] = func_ctx->shared_heap_base_addr_adj;
+    param_values[6] = is_memory64 ? I8_ONE : I8_ZERO;
+
+    if (!(ret_value = LLVMBuildCall2(comp_ctx->builder, func_type, func,
+                                     param_values, 7, "call"))) {
+        aot_set_last_error("llvm build call failed.");
+        goto fail;
+    }
+
+    BUILD_ICMP(LLVMIntEQ, ret_value, I8_ZERO, cmp, "shared_heap_oob");
+    if (!aot_emit_exception(comp_ctx, func_ctx,
+                            EXCE_OUT_OF_BOUNDS_MEMORY_ACCESS, true, cmp,
+                            check_succ)) {
+        goto fail;
+    }
+
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Setup the basic blocks for shared heap and shared chain memory checks.
+ *
+ * Arguments:
+ *   block_curr: The current basic block.
+ *   app_addr_in_cache_shared_heap: Output, block for cache shared heap.
+ *   app_addr_in_linear_mem: Output, block for linear memory.
+ *   app_addr_in_shared_heap_chain: Output, block for shared heap chain
+ *                                  (only for shared heap chain).
+ *   check_shared_heap_chain: Output, block for checking shared heap chain
+ *                            (only for shared heap chain).
+ *
+ * Topology:
+ *   If enable_shared_heap:
+ *     block_curr -> app_addr_in_cache_shared_heap
+ *                -> app_addr_in_linear_mem
+ *   If enable_shared_chain:
+ *     block_curr -> app_addr_in_shared_heap_chain
+ *                   -> app_addr_in_cache_shared_heap
+ *                   -> check_shared_heap_chain
+ *                 -> app_addr_in_linear_mem
+ */
+static bool
+setup_shared_heap_blocks(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
+                         LLVMBasicBlockRef block_curr,
+                         LLVMBasicBlockRef *app_addr_in_cache_shared_heap,
+                         LLVMBasicBlockRef *app_addr_in_linear_mem,
+                         LLVMBasicBlockRef *app_addr_in_shared_heap_chain,
+                         LLVMBasicBlockRef *check_shared_heap_chain)
+{
+    ADD_BASIC_BLOCK(*app_addr_in_cache_shared_heap,
+                    "app_addr_in_cache_shared_heap");
+    ADD_BASIC_BLOCK(*app_addr_in_linear_mem, "app_addr_in_linear_mem");
+
+    if (comp_ctx->enable_shared_heap) {
+        LLVMMoveBasicBlockAfter(*app_addr_in_cache_shared_heap, block_curr);
+        LLVMMoveBasicBlockAfter(*app_addr_in_linear_mem,
+                                *app_addr_in_cache_shared_heap);
+    }
+    else if (comp_ctx->enable_shared_chain) {
+        ADD_BASIC_BLOCK(*app_addr_in_shared_heap_chain,
+                        "app_addr_in_shared_heap_chain");
+        ADD_BASIC_BLOCK(*check_shared_heap_chain, "check_shared_heap_chain");
+        LLVMMoveBasicBlockAfter(*app_addr_in_shared_heap_chain, block_curr);
+        LLVMMoveBasicBlockAfter(*app_addr_in_cache_shared_heap,
+                                *app_addr_in_shared_heap_chain);
+        LLVMMoveBasicBlockAfter(*check_shared_heap_chain,
+                                *app_addr_in_cache_shared_heap);
+        LLVMMoveBasicBlockAfter(*app_addr_in_linear_mem,
+                                *app_addr_in_cache_shared_heap);
+    }
+
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Build a branch to check if start_offset is in the shared heap chain region.
+ *
+ * Arguments:
+ *   start_offset: The offset to check.
+ *   app_addr_in_shared_heap_chain: Block to branch if in shared heap chain.
+ *   app_addr_in_linear_mem: Block to branch if not in shared heap chain.
+ */
+static bool
+build_check_app_addr_in_shared_heap_chain(
+    AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
+    LLVMValueRef start_offset, LLVMBasicBlockRef app_addr_in_shared_heap_chain,
+    LLVMBasicBlockRef app_addr_in_linear_mem)
+{
+    LLVMValueRef is_in_shared_heap = NULL;
+
+    /* Use start_offset > func_ctx->shared_heap_head_start_off to test
+     * start_off falls in shared heap chain memory region. The shared heap
+     * chain oob will be detected in app_addr_in_shared_heap block or
+     * aot_check_shared_heap_chain_and_update function
+     */
+    BUILD_ICMP(LLVMIntUGT, start_offset, func_ctx->shared_heap_head_start_off,
+               is_in_shared_heap, "shared_heap_lb_cmp");
+    BUILD_COND_BR(is_in_shared_heap, app_addr_in_shared_heap_chain,
+                  app_addr_in_linear_mem);
+
+    SET_BUILD_POS(app_addr_in_shared_heap_chain);
+
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Build the conditional branch for cache shared heap or shared heap chain.
+ *
+ * Arguments:
+ *   cmp: The condition for being in cache shared heap.
+ *   app_addr_in_cache_shared_heap: Block for cache shared heap.
+ *   app_addr_in_linear_mem: Block for linear memory.
+ *   check_shared_heap_chain: Block for checking shared heap chain.
+ *   bytes: The access size in bytes.
+ *   start_offset: The offset to check.
+ *   is_memory64: Whether memory is 64-bit.
+ */
+static bool
+build_shared_heap_conditional_branching(
+    AOTCompContext *comp_ctx, AOTFuncContext *func_ctx, LLVMValueRef cmp,
+    LLVMBasicBlockRef app_addr_in_cache_shared_heap,
+    LLVMBasicBlockRef app_addr_in_linear_mem,
+    LLVMBasicBlockRef check_shared_heap_chain, LLVMValueRef bytes,
+    LLVMValueRef start_offset, bool is_memory64)
+{
+    if (comp_ctx->enable_shared_heap) {
+        BUILD_COND_BR(cmp, app_addr_in_cache_shared_heap,
+                      app_addr_in_linear_mem);
+    }
+    else if (comp_ctx->enable_shared_chain) {
+        BUILD_COND_BR(cmp, app_addr_in_cache_shared_heap,
+                      check_shared_heap_chain);
+        SET_BUILD_POS(check_shared_heap_chain);
+        if (!aot_check_shared_heap_chain_and_update(
+                comp_ctx, func_ctx, app_addr_in_cache_shared_heap, start_offset,
+                bytes, is_memory64))
+            goto fail;
+    }
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Get the native address in the cache shared heap.
+ *
+ * Arguments:
+ *   start_offset: The offset to use for address calculation.
+ *   maddr: Output, the native address that in the cache shared heap.
+ */
+static bool
+build_get_maddr_in_cache_shared_heap(AOTCompContext *comp_ctx,
+                                     AOTFuncContext *func_ctx,
+                                     LLVMValueRef start_offset,
+                                     LLVMValueRef *maddr)
+{
+    LLVMValueRef shared_heap_base_addr_adj;
+    /* load the local variable */
+    BUILD_LOAD_PTR(func_ctx->shared_heap_base_addr_adj, INT8_PTR_TYPE,
+                   shared_heap_base_addr_adj);
+    if (!(*maddr = LLVMBuildInBoundsGEP2(
+              comp_ctx->builder, INT8_TYPE, shared_heap_base_addr_adj,
+              &start_offset, 1, "maddr_cache_shared_heap"))) {
+        aot_set_last_error("llvm build inbounds gep failed");
+        goto fail;
+    }
+
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Check for memory overflow in shared heap for normal memory access.
+ *
+ * Arguments:
+ *   block_curr: The current basic block.
+ *   block_maddr_phi: The phi block for memory address.
+ *   maddr_phi: The phi node for memory address.
+ *   start_offset: The first offset to check.
+ *   mem_base_addr: The base address of memory. Only used with segue.
+ *   bytes_u32: The access size in bytes.
+ *   is_memory64: Whether memory is wasm64 memory.
+ *   is_target_64bit: Whether target is 64-bit.
+ *   enable_segue: Whether to use segment register addressing.
+ */
+static bool
+aot_check_shared_heap_memory_overflow(
+    AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
+    LLVMBasicBlockRef block_curr, LLVMBasicBlockRef block_maddr_phi,
+    LLVMValueRef maddr_phi, LLVMValueRef start_offset,
+    LLVMValueRef mem_base_addr, uint32 bytes_u32, bool is_memory64,
+    bool is_target_64bit, bool enable_segue)
+{
+    LLVMBasicBlockRef app_addr_in_cache_shared_heap, app_addr_in_linear_mem;
+    LLVMBasicBlockRef app_addr_in_shared_heap_chain = NULL,
+                      check_shared_heap_chain = NULL;
+    LLVMValueRef cmp, cmp1, cmp2, shared_heap_start_off, shared_heap_end_off,
+        shared_heap_check_bound, maddr = NULL;
+    /* On 64/32-bit target, the offset is 64/32-bit */
+    LLVMTypeRef offset_type = is_target_64bit ? I64_TYPE : I32_TYPE;
+    LLVMValueRef length, bytes;
+
+    if (!setup_shared_heap_blocks(
+            comp_ctx, func_ctx, block_curr, &app_addr_in_cache_shared_heap,
+            &app_addr_in_linear_mem, &app_addr_in_shared_heap_chain,
+            &check_shared_heap_chain))
+        goto fail;
+    LLVMMoveBasicBlockAfter(block_maddr_phi, app_addr_in_linear_mem);
+
+    /* Early branching when it's not in shared heap chain at all */
+    if (comp_ctx->enable_shared_chain
+        && !build_check_app_addr_in_shared_heap_chain(
+            comp_ctx, func_ctx, start_offset, app_addr_in_shared_heap_chain,
+            app_addr_in_linear_mem))
+        goto fail;
+
+    /* Load the local variable of the function */
+    BUILD_LOAD_PTR(func_ctx->shared_heap_start_off, offset_type,
+                   shared_heap_start_off);
+    BUILD_LOAD_PTR(func_ctx->shared_heap_end_off, offset_type,
+                   shared_heap_end_off);
+    /* Check if the app address is in the cache shared heap range.
+     * If yes, branch to the cache branch; if not, check the shared heap chain
+     */
+    BUILD_ICMP(LLVMIntUGE, start_offset, shared_heap_start_off, cmp,
+               "cmp_cache_shared_heap_start");
+    length =
+        is_target_64bit ? I64_CONST(bytes_u32 - 1) : I32_CONST(bytes_u32 - 1);
+    CHECK_LLVM_CONST(length);
+    BUILD_OP(Sub, shared_heap_end_off, length, shared_heap_check_bound,
+             "cache_shared_heap_end_bound");
+    BUILD_ICMP(LLVMIntULE, start_offset, shared_heap_check_bound, cmp1,
+               "cmp_cache_shared_heap_end");
+    BUILD_OP(And, cmp, cmp1, cmp2, "is_in_cache_shared_heap");
+    /* Conditional branching based on whether in cached shared heap */
+    bytes = is_target_64bit ? I64_CONST(bytes_u32) : I32_CONST(bytes_u32);
+    if (!build_shared_heap_conditional_branching(
+            comp_ctx, func_ctx, cmp2, app_addr_in_cache_shared_heap,
+            app_addr_in_linear_mem, check_shared_heap_chain, bytes,
+            start_offset, is_memory64))
+        goto fail;
+
+    SET_BUILD_POS(app_addr_in_cache_shared_heap);
+    if (!build_get_maddr_in_cache_shared_heap(comp_ctx, func_ctx, start_offset,
+                                              &maddr))
+        goto fail;
+
+    if (enable_segue) {
+        LLVMValueRef mem_base_addr_u64, maddr_u64, offset_to_mem_base;
+        if (!(maddr_u64 = LLVMBuildPtrToInt(comp_ctx->builder, maddr, I64_TYPE,
+                                            "maddr_u64"))
+            || !(mem_base_addr_u64 =
+                     LLVMBuildPtrToInt(comp_ctx->builder, mem_base_addr,
+                                       I64_TYPE, "mem_base_addr_u64"))) {
+            aot_set_last_error("llvm build ptr to int failed");
+            goto fail;
+        }
+        if (!(offset_to_mem_base =
+                  LLVMBuildSub(comp_ctx->builder, maddr_u64, mem_base_addr_u64,
+                               "offset_to_mem_base"))) {
+            aot_set_last_error("llvm build sub failed");
+            goto fail;
+        }
+        if (!(maddr = LLVMBuildIntToPtr(comp_ctx->builder, offset_to_mem_base,
+                                        INT8_PTR_TYPE_GS,
+                                        "maddr_shared_heap_segue"))) {
+            aot_set_last_error("llvm build int to ptr failed.");
+            goto fail;
+        }
+    }
+
+    LLVMAddIncoming(maddr_phi, &maddr, &app_addr_in_cache_shared_heap, 1);
+    BUILD_BR(block_maddr_phi);
+    SET_BUILD_POS(app_addr_in_linear_mem);
+
+    return true;
+fail:
+    return false;
+}
+
+/*
+ * Check for memory overflow in shared heap for bulk memory access.
+ *
+ * Arguments:
+ *   block_curr: The current basic block.
+ *   block_maddr_phi: The phi block for memory address.
+ *   check_succ: The block to branch to on success.
+ *   maddr_phi: The phi node for memory address.
+ *   start_offset: The offset to check.
+ *   max_addr: The maximum address to check.
+ *   bytes: The access size in bytes (LLVMValueRef).
+ *   is_memory64: Whether memory is wasm64 memory.
+ *   is_target_64bit: Whether target is 64-bit.
+ */
+static bool
+aot_check_bulk_memory_shared_heap_memory_overflow(
+    AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
+    LLVMBasicBlockRef block_curr, LLVMBasicBlockRef block_maddr_phi,
+    LLVMBasicBlockRef check_succ, LLVMValueRef maddr_phi,
+    LLVMValueRef start_offset, LLVMValueRef max_addr, LLVMValueRef bytes,
+    bool is_memory64, bool is_target_64bit)
+{
+    LLVMBasicBlockRef app_addr_in_cache_shared_heap, app_addr_in_linear_mem;
+    LLVMBasicBlockRef app_addr_in_shared_heap_chain = NULL,
+                      check_shared_heap_chain = NULL;
+    LLVMValueRef cmp, cmp1, cmp2, shared_heap_start_off, shared_heap_end_off,
+        maddr = NULL, max_offset;
+    /* On 64/32-bit target, the offset is 64/32-bit */
+    LLVMTypeRef offset_type = is_target_64bit ? I64_TYPE : I32_TYPE;
+
+    if (!setup_shared_heap_blocks(
+            comp_ctx, func_ctx, block_curr, &app_addr_in_cache_shared_heap,
+            &app_addr_in_linear_mem, &app_addr_in_shared_heap_chain,
+            &check_shared_heap_chain))
+        goto fail;
+    LLVMMoveBasicBlockAfter(block_maddr_phi, check_succ);
+
+    /* Early branching when it's not in shared heap chain at all */
+    if (comp_ctx->enable_shared_chain
+        && !build_check_app_addr_in_shared_heap_chain(
+            comp_ctx, func_ctx, start_offset, app_addr_in_shared_heap_chain,
+            app_addr_in_linear_mem))
+        goto fail;
+
+    /* Load the local variable of the function */
+    BUILD_LOAD_PTR(func_ctx->shared_heap_start_off, offset_type,
+                   shared_heap_start_off);
+    BUILD_LOAD_PTR(func_ctx->shared_heap_end_off, offset_type,
+                   shared_heap_end_off);
+    /* Check if the app address is in the cache shared heap range.
+     * If yes, branch to the cache branch; if not, check the shared heap chain
+     */
+    BUILD_ICMP(LLVMIntUGE, start_offset, shared_heap_start_off, cmp,
+               "cmp_cache_shared_heap_start");
+    BUILD_OP(Add, max_addr, is_target_64bit ? I64_NEG_ONE : I32_NEG_ONE,
+             max_offset, "max_offset");
+    BUILD_ICMP(LLVMIntULE, max_offset, shared_heap_end_off, cmp1,
+               "cmp_cache_shared_heap_end");
+    BUILD_OP(And, cmp, cmp1, cmp2, "is_in_cache_shared_heap");
+    /* Conditional branching based on whether in cached shared heap */
+    if (!build_shared_heap_conditional_branching(
+            comp_ctx, func_ctx, cmp2, app_addr_in_cache_shared_heap,
+            app_addr_in_linear_mem, check_shared_heap_chain, bytes,
+            start_offset, is_memory64))
+        goto fail;
+
+    SET_BUILD_POS(app_addr_in_cache_shared_heap);
+    if (!build_get_maddr_in_cache_shared_heap(comp_ctx, func_ctx, start_offset,
+                                              &maddr))
+        goto fail;
+
+    LLVMAddIncoming(maddr_phi, &maddr, &app_addr_in_cache_shared_heap, 1);
+    BUILD_BR(block_maddr_phi);
+    SET_BUILD_POS(app_addr_in_linear_mem);
+
+    return true;
+fail:
+    return false;
+}
+#endif
+
 LLVMValueRef
 aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
                           mem_offset_t offset, uint32 bytes, bool enable_segue,
@@ -118,10 +564,10 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
 {
     LLVMValueRef offset_const =
         MEMORY64_COND_VALUE(I64_CONST(offset), I32_CONST(offset));
-    LLVMValueRef addr, maddr, maddr_phi = NULL, offset1, cmp1, cmp2, cmp;
+    LLVMValueRef addr, maddr, offset1, cmp1, cmp;
     LLVMValueRef mem_base_addr, mem_check_bound;
     LLVMBasicBlockRef block_curr = LLVMGetInsertBlock(comp_ctx->builder);
-    LLVMBasicBlockRef check_succ, block_maddr_phi = NULL;
+    LLVMBasicBlockRef check_succ;
     AOTValue *aot_value_top;
     uint32 local_idx_of_aot_value = 0;
     uint64 const_value;
@@ -136,6 +582,10 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
 #else
     bool is_memory64 = IS_MEMORY64;
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+    LLVMValueRef maddr_phi = NULL;
+    LLVMBasicBlockRef block_maddr_phi = NULL;
+#endif
 
     is_target_64bit = (comp_ctx->pointer_size == sizeof(uint64)) ? true : false;
 
@@ -262,6 +712,13 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
         *alignp = 1;
     }
 
+    /* The overflow check needs to be done under following conditions:
+     * 1. In 64-bit target, offset and addr will be extended to 64-bit
+     *    1.1 offset + addr can overflow when it's memory64
+     *    1.2 no overflow when it's memory32
+     * 2. In 32-bit target, offset and addr will be 32-bit
+     *    2.1 offset + addr can overflow when it's memory32
+     */
     if (is_target_64bit) {
         if (!(offset_const = LLVMBuildZExt(comp_ctx->builder, offset_const,
                                            I64_TYPE, "offset_i64"))
@@ -275,7 +732,9 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
     /* offset1 = offset + addr; */
     BUILD_OP(Add, offset_const, addr, offset1, "offset1");
 
-    if (is_memory64 && comp_ctx->enable_bound_check) {
+    /* 1.1 offset + addr can overflow when it's memory64
+     * 2.1 Or when it's on 32-bit platform */
+    if (is_memory64 || !is_target_64bit) {
         /* Check whether integer overflow occurs in offset + addr */
         LLVMBasicBlockRef check_integer_overflow_end;
         ADD_BASIC_BLOCK(check_integer_overflow_end,
@@ -289,23 +748,14 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
             goto fail;
         }
         SET_BUILD_POS(check_integer_overflow_end);
+        block_curr = check_integer_overflow_end;
     }
 
-    if (comp_ctx->enable_shared_heap /* TODO: && mem_idx == 0 */) {
+#if WASM_ENABLE_SHARED_HEAP != 0
-        LLVMBasicBlockRef app_addr_in_shared_heap, app_addr_in_linear_mem;
+    if (comp_ctx->enable_shared_heap
-        LLVMValueRef is_in_shared_heap, shared_heap_check_bound = NULL;
+        || comp_ctx->enable_shared_chain /* TODO: && mem_idx == 0 */) {
-
-        /* Add basic blocks */
-        ADD_BASIC_BLOCK(app_addr_in_shared_heap, "app_addr_in_shared_heap");
-        ADD_BASIC_BLOCK(app_addr_in_linear_mem, "app_addr_in_linear_mem");
         ADD_BASIC_BLOCK(block_maddr_phi, "maddr_phi");
-
+        SET_BUILD_POS(block_maddr_phi);
-        LLVMMoveBasicBlockAfter(app_addr_in_shared_heap, block_curr);
-        LLVMMoveBasicBlockAfter(app_addr_in_linear_mem,
-                                app_addr_in_shared_heap);
-        LLVMMoveBasicBlockAfter(block_maddr_phi, app_addr_in_linear_mem);
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, block_maddr_phi);
         if (!(maddr_phi =
                   LLVMBuildPhi(comp_ctx->builder,
                                enable_segue ? INT8_PTR_TYPE_GS : INT8_PTR_TYPE,
@@ -313,110 +763,16 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
             aot_set_last_error("llvm build phi failed");
             goto fail;
         }
+        SET_BUILD_POS(block_curr);
 
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, block_curr);
+        if (!aot_check_shared_heap_memory_overflow(
-
+                comp_ctx, func_ctx, block_curr, block_maddr_phi, maddr_phi,
-        if (!is_target_64bit) {
+                offset1, mem_base_addr, bytes, is_memory64, is_target_64bit,
-            /* Check whether integer overflow occurs in addr + offset */
+                enable_segue)) {
-            LLVMBasicBlockRef check_integer_overflow_end;
-            ADD_BASIC_BLOCK(check_integer_overflow_end,
-                            "check_integer_overflow_end");
-            LLVMMoveBasicBlockAfter(check_integer_overflow_end, block_curr);
-
-            BUILD_ICMP(LLVMIntULT, offset1, addr, cmp1, "cmp1");
-            if (!aot_emit_exception(comp_ctx, func_ctx,
-                                    EXCE_OUT_OF_BOUNDS_MEMORY_ACCESS, true,
-                                    cmp1, check_integer_overflow_end)) {
-                goto fail;
-            }
-            SET_BUILD_POS(check_integer_overflow_end);
-        }
-
-        shared_heap_check_bound =
-            is_memory64 ? I64_CONST(UINT64_MAX - bytes + 1)
-                        : (comp_ctx->pointer_size == sizeof(uint64)
-                               ? I64_CONST(UINT32_MAX - bytes + 1)
-                               : I32_CONST(UINT32_MAX - bytes + 1));
-        CHECK_LLVM_CONST(shared_heap_check_bound);
-
-        /* Check whether the bytes to access are in shared heap */
-        if (!comp_ctx->enable_bound_check) {
-            /* Use IntUGT but not IntUGE to compare, since (1) in the ems
-               memory allocator, the hmu node includes hmu header and hmu
-               memory, only the latter is returned to the caller as the
-               allocated memory, the hmu header isn't returned so the
-               first byte of the shared heap won't be accessed, (2) using
-               IntUGT gets better performance than IntUGE in some cases */
-            BUILD_ICMP(LLVMIntUGT, offset1, func_ctx->shared_heap_start_off,
-                       is_in_shared_heap, "is_in_shared_heap");
-            /* We don't check the shared heap's upper boundary if boundary
-               check isn't enabled, the runtime may also use the guard pages
-               of shared heap to check the boundary if hardware boundary
-               check feature is enabled. */
-        }
-        else {
-            /* Use IntUGT but not IntUGE to compare, same as above */
-            BUILD_ICMP(LLVMIntUGT, offset1, func_ctx->shared_heap_start_off,
-                       cmp1, "cmp1");
-            /* Check the shared heap's upper boundary if boundary check is
-               enabled */
-            BUILD_ICMP(LLVMIntULE, offset1, shared_heap_check_bound, cmp2,
-                       "cmp2");
-            BUILD_OP(And, cmp1, cmp2, is_in_shared_heap, "is_in_shared_heap");
-        }
-
-        if (!LLVMBuildCondBr(comp_ctx->builder, is_in_shared_heap,
-                             app_addr_in_shared_heap, app_addr_in_linear_mem)) {
-            aot_set_last_error("llvm build cond br failed");
-            goto fail;
-        }
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, app_addr_in_shared_heap);
-
-        /* Get native address inside shared heap */
-        if (!(maddr =
-                  LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,
-                                        func_ctx->shared_heap_base_addr_adj,
-                                        &offset1, 1, "maddr_shared_heap"))) {
-            aot_set_last_error("llvm build inbounds gep failed");
-            goto fail;
-        }
-
-        if (enable_segue) {
-            LLVMValueRef mem_base_addr_u64, maddr_u64, offset_to_mem_base;
-
-            if (!(maddr_u64 = LLVMBuildPtrToInt(comp_ctx->builder, maddr,
-                                                I64_TYPE, "maddr_u64"))
-                || !(mem_base_addr_u64 =
-                         LLVMBuildPtrToInt(comp_ctx->builder, mem_base_addr,
-                                           I64_TYPE, "mem_base_addr_u64"))) {
-                aot_set_last_error("llvm build ptr to int failed");
-                goto fail;
-            }
-            if (!(offset_to_mem_base =
-                      LLVMBuildSub(comp_ctx->builder, maddr_u64,
-                                   mem_base_addr_u64, "offset_to_mem_base"))) {
-                aot_set_last_error("llvm build sub failed");
-                goto fail;
-            }
-            if (!(maddr = LLVMBuildIntToPtr(
-                      comp_ctx->builder, offset_to_mem_base, INT8_PTR_TYPE_GS,
-                      "maddr_shared_heap_segue"))) {
-                aot_set_last_error("llvm build int to ptr failed.");
             goto fail;
         }
     }
-
+#endif
-        LLVMAddIncoming(maddr_phi, &maddr, &app_addr_in_shared_heap, 1);
-
-        if (!LLVMBuildBr(comp_ctx->builder, block_maddr_phi)) {
-            aot_set_last_error("llvm build br failed");
-            goto fail;
-        }
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, app_addr_in_linear_mem);
-        block_curr = LLVMGetInsertBlock(comp_ctx->builder);
-    }
 
     if (comp_ctx->enable_bound_check
         && !(is_local_of_aot_value
@@ -449,21 +805,7 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
             goto fail;
         }
 
-        if (is_target_64bit) {
         BUILD_ICMP(LLVMIntUGT, offset1, mem_check_bound, cmp, "cmp");
-        }
-        else {
-            if (comp_ctx->enable_shared_heap /* TODO: && mem_idx == 0 */) {
-                /* Check integer overflow has been checked above */
-                BUILD_ICMP(LLVMIntUGT, offset1, mem_check_bound, cmp, "cmp");
-            }
-            else {
-                /* Check integer overflow */
-                BUILD_ICMP(LLVMIntULT, offset1, addr, cmp1, "cmp1");
-                BUILD_ICMP(LLVMIntUGT, offset1, mem_check_bound, cmp2, "cmp2");
-                BUILD_OP(Or, cmp1, cmp2, cmp, "cmp");
-            }
-        }
 
         /* Add basic blocks */
         ADD_BASIC_BLOCK(check_succ, "check_succ");
@@ -509,17 +851,20 @@ aot_check_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
         }
     }
 
-    if (comp_ctx->enable_shared_heap /* TODO: && mem_idx == 0 */) {
+#if WASM_ENABLE_SHARED_HEAP != 0
+    if (comp_ctx->enable_shared_heap
+        || comp_ctx->enable_shared_chain /* TODO: && mem_idx == 0 */) {
         block_curr = LLVMGetInsertBlock(comp_ctx->builder);
         LLVMAddIncoming(maddr_phi, &maddr, &block_curr, 1);
         if (!LLVMBuildBr(comp_ctx->builder, block_maddr_phi)) {
             aot_set_last_error("llvm build br failed");
             goto fail;
         }
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, block_maddr_phi);
+        SET_BUILD_POS(block_maddr_phi);
         return maddr_phi;
     }
     else
+#endif
         return maddr;
 fail:
     return NULL;
@@ -544,15 +889,6 @@ fail:
         LLVMSetAlignment(value, known_align);                             \
     } while (0)
 
-#define BUILD_TRUNC(value, data_type)                                     \
-    do {                                                                  \
-        if (!(value = LLVMBuildTrunc(comp_ctx->builder, value, data_type, \
-                                     "val_trunc"))) {                     \
-            aot_set_last_error("llvm build trunc failed.");               \
-            goto fail;                                                    \
-        }                                                                 \
-    } while (0)
-
 #define BUILD_STORE()                                                   \
     do {                                                                \
         LLVMValueRef res;                                               \
@@ -1150,16 +1486,23 @@ LLVMValueRef
 check_bulk_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
                            LLVMValueRef offset, LLVMValueRef bytes)
 {
-    LLVMValueRef maddr, max_addr, cmp;
+    LLVMValueRef maddr, max_addr, cmp, cmp1;
-    LLVMValueRef mem_base_addr, maddr_phi = NULL;
+    LLVMValueRef mem_base_addr;
     LLVMBasicBlockRef block_curr = LLVMGetInsertBlock(comp_ctx->builder);
-    LLVMBasicBlockRef check_succ, block_maddr_phi = NULL;
+    LLVMBasicBlockRef check_succ;
     LLVMValueRef mem_size;
+    bool is_target_64bit;
 #if WASM_ENABLE_MEMORY64 == 0
     bool is_memory64 = false;
 #else
     bool is_memory64 = IS_MEMORY64;
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+    LLVMValueRef maddr_phi = NULL;
+    LLVMBasicBlockRef block_maddr_phi = NULL;
+#endif
+
+    is_target_64bit = (comp_ctx->pointer_size == sizeof(uint64)) ? true : false;
 
     /* Get memory base address and memory data size */
 #if WASM_ENABLE_SHARED_MEMORY != 0
@@ -1221,6 +1564,9 @@ check_bulk_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
     ADD_BASIC_BLOCK(check_succ, "check_succ");
     LLVMMoveBasicBlockAfter(check_succ, block_curr);
 
+    /* Same logic with aot_check_memory_overflow, offset and bytes are 32/64
+     * bits on 32/64 bits platform */
+    if (is_target_64bit) {
         offset =
             LLVMBuildZExt(comp_ctx->builder, offset, I64_TYPE, "extend_offset");
         bytes = LLVMBuildZExt(comp_ctx->builder, bytes, I64_TYPE, "extend_len");
@@ -1228,104 +1574,61 @@ check_bulk_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
             aot_set_last_error("llvm build zext failed.");
             goto fail;
         }
+    }
 
     BUILD_OP(Add, offset, bytes, max_addr, "max_addr");
 
-    if (is_memory64 && comp_ctx->enable_bound_check) {
+    /* Check overflow when it's memory64 or it's on 32 bits platform */
-        /* Check whether integer overflow occurs in offset + addr */
+    if (is_memory64 || !is_target_64bit) {
+        /* Check whether integer overflow occurs in offset + bytes */
         LLVMBasicBlockRef check_integer_overflow_end;
         ADD_BASIC_BLOCK(check_integer_overflow_end,
                         "check_integer_overflow_end");
         LLVMMoveBasicBlockAfter(check_integer_overflow_end, block_curr);
 
+        /* offset + bytes can overflow yet is valid(for example, 0xffffffff, 1),
+         * allow it to be 0(either 0, 0 or overflow and valid) */
         BUILD_ICMP(LLVMIntULT, max_addr, offset, cmp, "cmp");
+        BUILD_ICMP(LLVMIntNE, max_addr, is_target_64bit ? I64_ZERO : I32_ZERO,
+                   cmp1, "cmp1");
+        BUILD_OP(And, cmp, cmp1, cmp, "overflow");
         if (!aot_emit_exception(comp_ctx, func_ctx,
                                 EXCE_OUT_OF_BOUNDS_MEMORY_ACCESS, true, cmp,
                                 check_integer_overflow_end)) {
             goto fail;
         }
         SET_BUILD_POS(check_integer_overflow_end);
+        block_curr = check_integer_overflow_end;
     }
 
-    if (comp_ctx->enable_shared_heap /* TODO: && mem_idx == 0 */) {
+#if WASM_ENABLE_SHARED_HEAP != 0
-        LLVMBasicBlockRef app_addr_in_shared_heap, app_addr_in_linear_mem;
+    if (comp_ctx->enable_shared_heap
-        LLVMValueRef shared_heap_start_off, shared_heap_check_bound;
+        || comp_ctx->enable_shared_chain /* TODO: && mem_idx == 0 */) {
-        LLVMValueRef max_offset, cmp1, cmp2, is_in_shared_heap;
-
-        /* Add basic blocks */
-        ADD_BASIC_BLOCK(app_addr_in_shared_heap, "app_addr_in_shared_heap");
-        ADD_BASIC_BLOCK(app_addr_in_linear_mem, "app_addr_in_linear_mem");
         ADD_BASIC_BLOCK(block_maddr_phi, "maddr_phi");
-
+        SET_BUILD_POS(block_maddr_phi);
-        LLVMMoveBasicBlockAfter(app_addr_in_shared_heap, block_curr);
-        LLVMMoveBasicBlockAfter(app_addr_in_linear_mem,
-                                app_addr_in_shared_heap);
-        LLVMMoveBasicBlockAfter(block_maddr_phi, check_succ);
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, block_maddr_phi);
         if (!(maddr_phi = LLVMBuildPhi(comp_ctx->builder, INT8_PTR_TYPE,
                                        "maddr_phi"))) {
             aot_set_last_error("llvm build phi failed");
             goto fail;
         }
+        SET_BUILD_POS(block_curr);
 
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, block_curr);
+        if (!aot_check_bulk_memory_shared_heap_memory_overflow(
-
+                comp_ctx, func_ctx, block_curr, block_maddr_phi, check_succ,
-        shared_heap_start_off = func_ctx->shared_heap_start_off;
+                maddr_phi, offset, max_addr, bytes, is_memory64,
-        if (comp_ctx->pointer_size == sizeof(uint32)) {
+                is_target_64bit)) {
-            if (!(shared_heap_start_off =
-                      LLVMBuildZExt(comp_ctx->builder, shared_heap_start_off,
-                                    I64_TYPE, "shared_heap_start_off_u64"))) {
-                aot_set_last_error("llvm build zext failed");
             goto fail;
         }
     }
-        shared_heap_check_bound =
+#endif
-            is_memory64 ? I64_CONST(UINT64_MAX) : I64_CONST(UINT32_MAX);
-        CHECK_LLVM_CONST(shared_heap_check_bound);
 
-        /* Check whether the bytes to access are in shared heap */
+    /* mem_size is always 64-bit, extend max_addr on 32 bits platform */
-        if (!comp_ctx->enable_bound_check) {
+    if (!is_target_64bit
-            /* Use IntUGT but not IntUGE to compare, same as the check
+        && !(max_addr = LLVMBuildZExt(comp_ctx->builder, max_addr, I64_TYPE,
-               in aot_check_memory_overflow */
+                                      "extend_max_addr"))) {
-            BUILD_ICMP(LLVMIntUGT, offset, func_ctx->shared_heap_start_off,
+        aot_set_last_error("llvm build zext failed.");
-                       is_in_shared_heap, "is_in_shared_heap");
-        }
-        else {
-            BUILD_ICMP(LLVMIntUGT, offset, func_ctx->shared_heap_start_off,
-                       cmp1, "cmp1");
-            BUILD_OP(Add, max_addr, I64_NEG_ONE, max_offset, "max_offset");
-            BUILD_ICMP(LLVMIntULE, max_offset, shared_heap_check_bound, cmp2,
-                       "cmp2");
-            BUILD_OP(And, cmp1, cmp2, is_in_shared_heap, "is_in_shared_heap");
-        }
-
-        if (!LLVMBuildCondBr(comp_ctx->builder, is_in_shared_heap,
-                             app_addr_in_shared_heap, app_addr_in_linear_mem)) {
-            aot_set_last_error("llvm build cond br failed");
         goto fail;
     }
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, app_addr_in_shared_heap);
-
-        /* Get native address inside shared heap */
-        if (!(maddr = LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,
-                                            func_ctx->shared_heap_base_addr_adj,
-                                            &offset, 1, "maddr_shared_heap"))) {
-            aot_set_last_error("llvm build inbounds gep failed");
-            goto fail;
-        }
-        LLVMAddIncoming(maddr_phi, &maddr, &app_addr_in_shared_heap, 1);
-
-        if (!LLVMBuildBr(comp_ctx->builder, block_maddr_phi)) {
-            aot_set_last_error("llvm build br failed");
-            goto fail;
-        }
-
-        LLVMPositionBuilderAtEnd(comp_ctx->builder, app_addr_in_linear_mem);
-        block_curr = LLVMGetInsertBlock(comp_ctx->builder);
-    }
-
     BUILD_ICMP(LLVMIntUGT, max_addr, mem_size, cmp, "cmp_max_mem_addr");
 
     if (!aot_emit_exception(comp_ctx, func_ctx,
@@ -1341,7 +1644,9 @@ check_bulk_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
         goto fail;
     }
 
-    if (comp_ctx->enable_shared_heap /* TODO: && mem_idx == 0 */) {
+#if WASM_ENABLE_SHARED_HEAP != 0
+    if (comp_ctx->enable_shared_heap
+        || comp_ctx->enable_shared_chain /* TODO: && mem_idx == 0 */) {
         block_curr = LLVMGetInsertBlock(comp_ctx->builder);
         LLVMAddIncoming(maddr_phi, &maddr, &block_curr, 1);
         if (!LLVMBuildBr(comp_ctx->builder, block_maddr_phi)) {
@@ -1352,6 +1657,7 @@ check_bulk_memory_overflow(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
         return maddr_phi;
     }
     else
+#endif
         return maddr;
 fail:
     return NULL;

core/iwasm/compilation/aot_llvm.c

@@ -1517,73 +1517,154 @@ create_memory_info(const AOTCompContext *comp_ctx, AOTFuncContext *func_ctx,
     return true;
 }
 
+#define BUILD_IS_NOT_NULL(value, res, name)                                \
+    do {                                                                   \
+        if (!(res = LLVMBuildIsNotNull(comp_ctx->builder, value, name))) { \
+            aot_set_last_error("llvm build is not null failed.");          \
+            goto fail;                                                     \
+        }                                                                  \
+    } while (0)
+
+#define get_module_extra_field_offset(field)                        \
+    do {                                                            \
+        offset_u32 = get_module_inst_extra_offset(comp_ctx);        \
+        if (comp_ctx->is_jit_mode)                                  \
+            offset_u32 += offsetof(WASMModuleInstanceExtra, field); \
+        else                                                        \
+            offset_u32 += offsetof(AOTModuleInstanceExtra, field);  \
+    } while (0)
+
+#define LOAD_MODULE_EXTRA_FIELD_AND_ALLOCA(field, type)                        \
+    do {                                                                       \
+        get_module_extra_field_offset(field);                                  \
+        offset = I32_CONST(offset_u32);                                        \
+        CHECK_LLVM_CONST(offset);                                              \
+        if (!(field_p = LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,    \
+                                              func_ctx->aot_inst, &offset, 1,  \
+                                              #field "_p"))) {                 \
+            aot_set_last_error("llvm build inbounds gep failed");              \
+            goto fail;                                                         \
+        }                                                                      \
+        if (!(load_val =                                                       \
+                  LLVMBuildLoad2(comp_ctx->builder, type, field_p, #field))) { \
+            aot_set_last_error("llvm build load failed");                      \
+            goto fail;                                                         \
+        }                                                                      \
+        if (!(func_ctx->field =                                                \
+                  LLVMBuildAlloca(comp_ctx->builder, type, #field))) {         \
+            aot_set_last_error("llvm build alloca failed");                    \
+            goto fail;                                                         \
+        }                                                                      \
+        if (!LLVMBuildStore(comp_ctx->builder, load_val, func_ctx->field)) {   \
+            aot_set_last_error("llvm build store failed");                     \
+            goto fail;                                                         \
+        }                                                                      \
+    } while (0)
+
 static bool
 create_shared_heap_info(AOTCompContext *comp_ctx, AOTFuncContext *func_ctx)
 {
-    LLVMValueRef offset, base_addr_p, start_off_p, cmp;
+#if WASM_ENABLE_SHARED_HEAP != 0
+    LLVMValueRef offset, field_p, load_val, shared_heap_head_p,
+        shared_heap_head, cmp, field_p_or_default, shared_heap_head_start_off,
+        shared_heap_head_start_off_minus_one;
+    LLVMTypeRef shared_heap_offset_type;
     uint32 offset_u32;
-
+#if WASM_ENABLE_MEMORY64 == 0
-    /* Load aot_inst->e->shared_heap_base_addr_adj */
+    bool is_memory64 = false;
-    offset_u32 = get_module_inst_extra_offset(comp_ctx);
+#else
-#if WASM_ENABLE_JIT != 0 && WASM_ENABLE_SHARED_HEAP != 0
+    bool is_memory64 = IS_MEMORY64;
-    if (comp_ctx->is_jit_mode)
-        offset_u32 +=
-            offsetof(WASMModuleInstanceExtra, shared_heap_base_addr_adj);
-    else
 #endif
-        offset_u32 +=
+
-            offsetof(AOTModuleInstanceExtra, shared_heap_base_addr_adj);
+    shared_heap_offset_type =
+        comp_ctx->pointer_size == sizeof(uint64) ? I64_TYPE : I32_TYPE;
+
+    /* shared_heap_base_addr_adj, shared_heap_start_off, and
+     * shared_heap_end_off can be updated later, use local variable to
+     * represent them */
+    LOAD_MODULE_EXTRA_FIELD_AND_ALLOCA(shared_heap_base_addr_adj,
+                                       INT8_PTR_TYPE);
+    LOAD_MODULE_EXTRA_FIELD_AND_ALLOCA(shared_heap_start_off,
+                                       shared_heap_offset_type);
+    LOAD_MODULE_EXTRA_FIELD_AND_ALLOCA(shared_heap_end_off,
+                                       shared_heap_offset_type);
+
+    /* Shared Heap head start off won't be updated, no need to alloca */
+    get_module_extra_field_offset(shared_heap);
     offset = I32_CONST(offset_u32);
     CHECK_LLVM_CONST(offset);
-
+    if (!(shared_heap_head_p = LLVMBuildInBoundsGEP2(
-    if (!(base_addr_p = LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,
+              comp_ctx->builder, INT8_TYPE, func_ctx->aot_inst, &offset, 1,
-                                              func_ctx->aot_inst, &offset, 1,
+              "shared_heap_head_p"))) {
-                                              "shared_heap_base_addr_adj_p"))) {
         aot_set_last_error("llvm build inbounds gep failed");
-        return false;
+        goto fail;
     }
-    if (!(func_ctx->shared_heap_base_addr_adj =
+    if (!(shared_heap_head =
-              LLVMBuildLoad2(comp_ctx->builder, INT8_PTR_TYPE, base_addr_p,
+              LLVMBuildLoad2(comp_ctx->builder, INT8_PTR_TYPE,
-                             "shared_heap_base_addr_adj"))) {
+                             shared_heap_head_p, "shared_heap_head"))) {
         aot_set_last_error("llvm build load failed");
-        return false;
+        goto fail;
     }
+    BUILD_IS_NOT_NULL(shared_heap_head, cmp, "has_shared_heap");
 
-    /* Load aot_inst->e->shared_heap_start_off */
+    if (is_memory64) {
-    offset_u32 = get_module_inst_extra_offset(comp_ctx);
+        offset_u32 = offsetof(WASMSharedHeap, start_off_mem64);
-#if WASM_ENABLE_JIT != 0 && WASM_ENABLE_SHARED_HEAP != 0
+    }
-    if (comp_ctx->is_jit_mode)
+    else {
-        offset_u32 += offsetof(WASMModuleInstanceExtra, shared_heap_start_off);
+        offset_u32 = offsetof(WASMSharedHeap, start_off_mem32);
-    else
+    }
-#endif
-        offset_u32 += offsetof(AOTModuleInstanceExtra, shared_heap_start_off);
     offset = I32_CONST(offset_u32);
     CHECK_LLVM_CONST(offset);
-
+    if (!(field_p = LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,
-    if (!(start_off_p = LLVMBuildInBoundsGEP2(comp_ctx->builder, INT8_TYPE,
+                                          shared_heap_head, &offset, 1,
-                                              func_ctx->aot_inst, &offset, 1,
+                                          "head_start_off_p"))) {
-                                              "shared_heap_start_off_p"))) {
         aot_set_last_error("llvm build inbounds gep failed");
-        return false;
+        goto fail;
     }
-    if (!(func_ctx->shared_heap_start_off = LLVMBuildLoad2(
+
-              comp_ctx->builder,
+    /* Select a valid shared heap head ptr or safe alloca ptr stores
-              comp_ctx->pointer_size == sizeof(uint64) ? I64_TYPE : I32_TYPE,
+     * shared_heap_start_off(UINT32_MAX/UINT64_MAX) */
-              start_off_p, "shared_heap_start_off"))) {
+    if (!(field_p_or_default = LLVMBuildSelect(comp_ctx->builder, cmp, field_p,
+                                               func_ctx->shared_heap_start_off,
+                                               "ptr_or_default"))) {
+        aot_set_last_error("llvm build select failed");
+        goto fail;
+    }
+
+    if (!(shared_heap_head_start_off = LLVMBuildLoad2(
+              comp_ctx->builder, shared_heap_offset_type, field_p_or_default,
+              "shared_heap_head_start_off"))) {
         aot_set_last_error("llvm build load failed");
-        return false;
+        goto fail;
+    }
+    if (!(shared_heap_head_start_off_minus_one = LLVMBuildAdd(
+              comp_ctx->builder, shared_heap_head_start_off,
+              comp_ctx->pointer_size == sizeof(uint64) ? I64_NEG_ONE
+                                                       : I32_NEG_ONE,
+              "head_start_off_minus_one"))) {
+        aot_set_last_error("llvm build load failed");
+        goto fail;
     }
 
-    if (!(cmp = LLVMBuildIsNotNull(comp_ctx->builder,
+    /* if there is attached shared heap(s), the value will be valid start_off-1,
-                                   func_ctx->shared_heap_base_addr_adj,
+     * otherwise it will be UINT32_MAX/UINT64_MAX, so during the bounds checks,
-                                   "has_shared_heap"))) {
+     * when has attached shared heap:
-        aot_set_last_error("llvm build is not null failed");
+     *   offset > start_off - 1 => offset >= start_off
-        return false;
+     * when no attached shared heap:
+     *   offset > UINT32_MAX/UINT64_MAX is always false
+     * */
+    if (!(func_ctx->shared_heap_head_start_off = LLVMBuildSelect(
+              comp_ctx->builder, cmp, shared_heap_head_start_off_minus_one,
+              shared_heap_head_start_off, "head_start_off"))) {
+        aot_set_last_error("llvm build select failed");
+        goto fail;
     }
-
     return true;
 fail:
     return false;
+#else  /* else of WASM_ENABLE_SHARED_HEAP != 0 */
+    return true;
+#endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 }
 
 static bool
@@ -1877,7 +1958,7 @@ aot_create_func_context(const AOTCompData *comp_data, AOTCompContext *comp_ctx,
     }
 
     /* Load shared heap, shared heap start off mem32 or mem64 */
-    if (comp_ctx->enable_shared_heap
+    if ((comp_ctx->enable_shared_heap || comp_ctx->enable_shared_chain)
         && !create_shared_heap_info(comp_ctx, func_ctx)) {
         goto fail;
     }
@@ -2703,6 +2784,12 @@ aot_create_comp_context(const AOTCompData *comp_data, aot_comp_option_t option)
     if (option->enable_shared_heap)
         comp_ctx->enable_shared_heap = true;
 
+    if (option->enable_shared_chain)
+        comp_ctx->enable_shared_chain = true;
+
+    if (option->enable_extended_const)
+        comp_ctx->enable_extended_const = true;
+
     comp_ctx->opt_level = option->opt_level;
     comp_ctx->size_level = option->size_level;
 
@@ -3999,7 +4086,7 @@ aot_get_func_from_table(const AOTCompContext *comp_ctx, LLVMValueRef base,
 
     if (!(func =
               LLVMBuildBitCast(comp_ctx->builder, func, func_type, "func"))) {
-        aot_set_last_error("cast function fialed.");
+        aot_set_last_error("cast function failed.");
         goto fail;
     }
 
@@ -4068,7 +4155,7 @@ aot_load_const_from_table(AOTCompContext *comp_ctx, LLVMValueRef base,
 
     if (!(const_addr = LLVMBuildBitCast(comp_ctx->builder, const_addr,
                                         const_ptr_type, "const_addr"))) {
-        aot_set_last_error("cast const fialed.");
+        aot_set_last_error("cast const failed.");
         return NULL;
     }
 

core/iwasm/compilation/aot_llvm.h

@@ -254,8 +254,12 @@ typedef struct AOTFuncContext {
     bool mem_space_unchanged;
     AOTCheckedAddrList checked_addr_list;
 
+    /* The last accessed shared heap info */
     LLVMValueRef shared_heap_base_addr_adj;
     LLVMValueRef shared_heap_start_off;
+    LLVMValueRef shared_heap_end_off;
+    /* The start offset of the head of shared heap chain */
+    LLVMValueRef shared_heap_head_start_off;
 
     LLVMBasicBlockRef got_exception_block;
     LLVMBasicBlockRef func_return_block;
@@ -457,6 +461,9 @@ typedef struct AOTCompContext {
     /* Enable LLVM PGO (Profile-Guided Optimization) */
     bool enable_llvm_pgo;
 
+    /* Enable extended constant expression */
+    bool enable_extended_const;
+
     /* Treat unknown import function as wasm-c-api import function
        and allow to directly invoke it from AOT/JIT code */
     bool quick_invoke_c_api_import;
@@ -486,6 +493,7 @@ typedef struct AOTCompContext {
     bool enable_gc;
 
     bool enable_shared_heap;
+    bool enable_shared_chain;
 
     uint32 opt_level;
     uint32 size_level;

core/iwasm/fast-jit/fe/jit_emit_table.c

@@ -121,7 +121,8 @@ wasm_init_table(WASMModuleInstance *inst, uint32 tbl_idx, uint32 seg_idx,
                               + dst_offset * sizeof(table_elem_type_t));
     init_values = tbl_seg_init_values + src_offset;
     for (i = 0; i < len; i++) {
-        addr[i] = (table_elem_type_t)(uintptr_t)init_values[+i].u.ref_index;
+        addr[i] =
+            (table_elem_type_t)(uintptr_t)init_values[+i].u.unary.v.ref_index;
     }
 
     return 0;

core/iwasm/include/aot_comp_option.h

@@ -68,6 +68,7 @@ typedef struct AOTCompOption {
     bool enable_ref_types;
     bool enable_gc;
     bool enable_aux_stack_check;
+    bool enable_extended_const;
     AOTStackFrameType aux_stack_frame_type;
     AOTCallStackFeatures call_stack_features;
     bool enable_perf_profiling;
@@ -79,6 +80,7 @@ typedef struct AOTCompOption {
     bool enable_stack_estimation;
     bool quick_invoke_c_api_import;
     bool enable_shared_heap;
+    bool enable_shared_chain;
     char *use_prof_file;
     uint32_t opt_level;
     uint32_t size_level;

core/iwasm/include/aot_export.h

@@ -98,10 +98,10 @@ void
 aot_destroy_aot_file(uint8_t *aot_file);
 
 char *
-aot_get_last_error();
+aot_get_last_error(void);
 
 uint32_t
-aot_get_plt_table_size();
+aot_get_plt_table_size(void);
 
 #ifdef __cplusplus
 }

core/iwasm/include/wasm_export.h

@@ -139,8 +139,6 @@ typedef struct wasm_frame_t {
     uint32_t *lp;
 } WASMCApiFrame;
 
-typedef WASMCApiFrame wasm_frame_t;
-
 /* WASM section */
 typedef struct wasm_section_t {
     struct wasm_section_t *next;
@@ -291,6 +289,8 @@ typedef struct InstantiationArgs {
 } InstantiationArgs;
 #endif /* INSTANTIATION_ARGS_OPTION_DEFINED */
 
+struct InstantiationArgs2;
+
 #ifndef WASM_VALKIND_T_DEFINED
 #define WASM_VALKIND_T_DEFINED
 typedef uint8_t wasm_valkind_t;
@@ -351,6 +351,7 @@ typedef enum {
 
 typedef struct SharedHeapInitArgs {
     uint32_t size;
+    void *pre_allocated_addr;
 } SharedHeapInitArgs;
 
 /**
@@ -734,6 +735,46 @@ wasm_runtime_instantiate_ex(const wasm_module_t module,
                             const InstantiationArgs *args, char *error_buf,
                             uint32_t error_buf_size);
 
+/**
+ * Create an InstantiationArgs2 object with default parameters.
+ *
+ * @return true if success, false otherwise
+ */
+WASM_RUNTIME_API_EXTERN bool
+wasm_runtime_instantiation_args_create(struct InstantiationArgs2 **p);
+
+/**
+ * Dispose an InstantiationArgs2 object.
+ */
+WASM_RUNTIME_API_EXTERN void
+wasm_runtime_instantiation_args_destroy(struct InstantiationArgs2 *p);
+
+/**
+ * Setter functions for the InstantiationArgs2 object.
+ */
+WASM_RUNTIME_API_EXTERN void
+wasm_runtime_instantiation_args_set_default_stack_size(
+    struct InstantiationArgs2 *p, uint32_t v);
+
+WASM_RUNTIME_API_EXTERN void
+wasm_runtime_instantiation_args_set_host_managed_heap_size(
+    struct InstantiationArgs2 *p, uint32_t v);
+
+WASM_RUNTIME_API_EXTERN void
+wasm_runtime_instantiation_args_set_max_memory_pages(
+    struct InstantiationArgs2 *p, uint32_t v);
+
+/**
+ * Instantiate a WASM module, with specified instantiation arguments
+ *
+ * Same as wasm_runtime_instantiate_ex, but this version takes
+ * InstantiationArgs2, which can be extended without breaking the ABI.
+ */
+WASM_RUNTIME_API_EXTERN wasm_module_inst_t
+wasm_runtime_instantiate_ex2(const wasm_module_t module,
+                             const struct InstantiationArgs2 *args,
+                             char *error_buf, uint32_t error_buf_size);
+
 /**
  * Set the running mode of a WASM module instance, override the
  * default running mode of the runtime. Note that it only makes sense when
@@ -904,7 +945,7 @@ wasm_runtime_destroy_exec_env(wasm_exec_env_t exec_env);
  * @return number of copied frames
  */
 WASM_RUNTIME_API_EXTERN uint32_t
-wasm_copy_callstack(const wasm_exec_env_t exec_env, wasm_frame_t *buffer,
+wasm_copy_callstack(const wasm_exec_env_t exec_env, WASMCApiFrame *buffer,
                     const uint32_t length, const uint32_t skip_n,
                     char *error_buf, uint32_t error_buf_size);
 
@@ -2316,7 +2357,37 @@ WASM_RUNTIME_API_EXTERN wasm_shared_heap_t
 wasm_runtime_create_shared_heap(SharedHeapInitArgs *init_args);
 
 /**
- * Attach a shared heap to a module instance
+ * This function links two shared heap(lists), `head` and `body` in to a single
+ * shared heap list, where `head` becomes the new shared heap list head. The
+ * shared heap list remains one continuous shared heap in wasm app's point of
+ * view.  At most one shared heap in shared heap list can be dynamically
+ * allocated, the rest have to be the pre-allocated shared heap. *
+ *
+ * @param head The head of the shared heap chain.
+ * @param body The body of the shared heap chain to be appended.
+ * @return The new head of the shared heap chain. NULL if failed.
+ */
+WASM_RUNTIME_API_EXTERN wasm_shared_heap_t
+wasm_runtime_chain_shared_heaps(wasm_shared_heap_t head,
+                                wasm_shared_heap_t body);
+
+/**
+ * This function unchains the shared heaps from the given head. If
+ * `entire_chain` is true, it will unchain the entire chain of shared heaps.
+ * Otherwise, it will unchain only the first shared heap in the chain.
+ *
+ * @param head The head of the shared heap chain.
+ * @param entire_chain A boolean flag indicating whether to unchain the entire
+ * chain.
+ * @return The new head of the shared heap chain. Or the last shared heap in the
+ * chain if `entire_chain` is true.
+ */
+wasm_shared_heap_t
+wasm_runtime_unchain_shared_heaps(wasm_shared_heap_t head, bool entire_chain);
+
+/**
+ * Attach a shared heap, it can be the head of shared heap chain, in that case,
+ * attach the shared heap chain, to a module instance
  *
  * @param module_inst the module instance
  * @param shared_heap the shared heap
@@ -2335,7 +2406,8 @@ WASM_RUNTIME_API_EXTERN void
 wasm_runtime_detach_shared_heap(wasm_module_inst_t module_inst);
 
 /**
- * Allocate memory from a shared heap
+ * Allocate memory from a shared heap, or the non-preallocated shared heap from
+ * the shared heap chain
  *
  * @param module_inst the module instance
  * @param size required memory size
@@ -2352,7 +2424,8 @@ wasm_runtime_shared_heap_malloc(wasm_module_inst_t module_inst, uint64_t size,
                                 void **p_native_addr);
 
 /**
- * Free the memory allocated from shared heap
+ * Free the memory allocated from shared heap, or the non-preallocated shared
+ * heap from the shared heap chain
  *
  * @param module_inst the module instance
  * @param ptr the offset in wasm app

core/iwasm/interpreter/wasm.h

@@ -135,6 +135,12 @@ typedef void *table_elem_type_t;
 #define INIT_EXPR_TYPE_F64_CONST 0x44
 #define INIT_EXPR_TYPE_V128_CONST 0xFD
 #define INIT_EXPR_TYPE_GET_GLOBAL 0x23
+#define INIT_EXPR_TYPE_I32_ADD 0x6A
+#define INIT_EXPR_TYPE_I32_SUB 0x6B
+#define INIT_EXPR_TYPE_I32_MUL 0x6C
+#define INIT_EXPR_TYPE_I64_ADD 0x7C
+#define INIT_EXPR_TYPE_I64_SUB 0x7D
+#define INIT_EXPR_TYPE_I64_MUL 0x7E
 #define INIT_EXPR_TYPE_REFNULL_CONST 0xD0
 #define INIT_EXPR_TYPE_FUNCREF_CONST 0xD2
 #define INIT_EXPR_TYPE_STRUCT_NEW 0xD3
@@ -277,9 +283,41 @@ typedef struct InitializerExpression {
     /* type of INIT_EXPR_TYPE_XXX, which is an instruction of
        constant expression */
     uint8 init_expr_type;
-    WASMValue u;
+    union {
+        struct {
+            WASMValue v;
+        } unary;
+        struct {
+            struct InitializerExpression *l_expr;
+            struct InitializerExpression *r_expr;
+        } binary;
+    } u;
 } InitializerExpression;
 
+static inline bool
+is_expr_binary_op(uint8 flag)
+{
+    return flag == INIT_EXPR_TYPE_I32_ADD || flag == INIT_EXPR_TYPE_I32_SUB
+           || flag == INIT_EXPR_TYPE_I32_MUL || flag == INIT_EXPR_TYPE_I64_ADD
+           || flag == INIT_EXPR_TYPE_I64_SUB || flag == INIT_EXPR_TYPE_I64_MUL;
+}
+
+/* check if table or data offset is valid for i32 offset */
+static inline bool
+is_valid_i32_offset(uint8 flag)
+{
+    return flag == INIT_EXPR_TYPE_I32_CONST || flag == INIT_EXPR_TYPE_I32_ADD
+           || flag == INIT_EXPR_TYPE_I32_SUB || flag == INIT_EXPR_TYPE_I32_MUL;
+}
+
+/* check if table or data offset is valid for i64 offset */
+static inline bool
+is_valid_i64_offset(uint8 flag)
+{
+    return flag == INIT_EXPR_TYPE_I64_CONST || flag == INIT_EXPR_TYPE_I64_ADD
+           || flag == INIT_EXPR_TYPE_I64_SUB || flag == INIT_EXPR_TYPE_I64_MUL;
+}
+
 #if WASM_ENABLE_GC != 0
 /**
  * Reference type of (ref null ht) or (ref ht),

core/iwasm/interpreter/wasm_interp_classic.c

@@ -46,28 +46,6 @@ typedef float64 CellType_F64;
 #define get_linear_mem_size() GET_LINEAR_MEMORY_SIZE(memory)
 #endif
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-#if WASM_ENABLE_MULTI_MEMORY != 0
-/* Only enable shared heap for the default memory */
-#define is_default_memory (memidx == 0)
-#else
-#define is_default_memory true
-#endif
-#define app_addr_in_shared_heap(app_addr, bytes)                             \
-    (shared_heap && is_default_memory && (app_addr) >= shared_heap_start_off \
-     && (app_addr) <= shared_heap_end_off - bytes + 1)
-
-#define shared_heap_addr_app_to_native(app_addr, native_addr) \
-    native_addr = shared_heap_base_addr + ((app_addr)-shared_heap_start_off)
-
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr) \
-    if (app_addr_in_shared_heap(app_addr, bytes))                \
-        shared_heap_addr_app_to_native(app_addr, native_addr);   \
-    else
-#else
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr)
-#endif
-
 #if WASM_ENABLE_MEMORY64 == 0
 
 #if (!defined(OS_ENABLE_HW_BOUND_CHECK) \
@@ -1670,22 +1648,6 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
     if (memory)
         is_memory64 = memory->is_memory64;
 #endif
-#if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap = module->e->shared_heap;
-    uint8 *shared_heap_base_addr = shared_heap ? shared_heap->base_addr : NULL;
-#if WASM_ENABLE_MEMORY64 != 0
-    uint64 shared_heap_start_off =
-        shared_heap ? (is_memory64 ? shared_heap->start_off_mem64
-                                   : shared_heap->start_off_mem32)
-                    : 0;
-    uint64 shared_heap_end_off =
-        shared_heap ? (is_memory64 ? UINT64_MAX : UINT32_MAX) : 0;
-#else
-    uint64 shared_heap_start_off =
-        shared_heap ? shared_heap->start_off_mem32 : 0;
-    uint64 shared_heap_end_off = shared_heap ? UINT32_MAX : 0;
-#endif
-#endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 #if WASM_ENABLE_MULTI_MEMORY != 0
     uint32 memidx = 0;
     uint32 memidx_cached = (uint32)-1;
@@ -4088,7 +4050,7 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
                     case WASM_OP_STRING_ENCODE_LOSSY_UTF8_ARRAY:
                     case WASM_OP_STRING_ENCODE_WTF8_ARRAY:
                     {
-                        uint32 start, array_len, count;
+                        uint32 start, array_len;
                         int32 bytes_written;
                         EncodingFlag flag = WTF8;
                         WASMArrayType *array_type;
@@ -5996,12 +5958,14 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
                                       || init_values[i].init_expr_type
                                              == INIT_EXPR_TYPE_FUNCREF_CONST);
 #if WASM_ENABLE_GC == 0
-                            table_elems[i] =
+                            table_elems[i] = (table_elem_type_t)init_values[i]
-                                (table_elem_type_t)init_values[i].u.ref_index;
+                                                 .u.unary.v.ref_index;
 #else
-                            if (init_values[i].u.ref_index != UINT32_MAX) {
+                            if (init_values[i].u.unary.v.ref_index
+                                != UINT32_MAX) {
                                 if (!(func_obj = wasm_create_func_obj(
-                                          module, init_values[i].u.ref_index,
+                                          module,
+                                          init_values[i].u.unary.v.ref_index,
                                           true, NULL, 0))) {
                                     goto got_exception;
                                 }

core/iwasm/interpreter/wasm_interp_fast.c

@@ -41,22 +41,6 @@ typedef float64 CellType_F64;
 #define get_linear_mem_size() GET_LINEAR_MEMORY_SIZE(memory)
 #endif
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-#define app_addr_in_shared_heap(app_addr, bytes)        \
-    (shared_heap && (app_addr) >= shared_heap_start_off \
-     && (app_addr) <= shared_heap_end_off - bytes + 1)
-
-#define shared_heap_addr_app_to_native(app_addr, native_addr) \
-    native_addr = shared_heap_base_addr + ((app_addr)-shared_heap_start_off)
-
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr) \
-    if (app_addr_in_shared_heap(app_addr, bytes))                \
-        shared_heap_addr_app_to_native(app_addr, native_addr);   \
-    else
-#else
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr)
-#endif
-
 #if !defined(OS_ENABLE_HW_BOUND_CHECK) \
     || WASM_CPU_SUPPORTS_UNALIGNED_ADDR_ACCESS == 0
 #define CHECK_MEMORY_OVERFLOW(bytes)                                           \
@@ -1590,21 +1574,13 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
     bool is_return_call = false;
 #endif
 #if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap = module->e ? module->e->shared_heap : NULL;
+    /* TODO: currently flowing two variables are only dummy for shared heap
-    uint8 *shared_heap_base_addr = shared_heap ? shared_heap->base_addr : NULL;
+     * boundary check, need to be updated when multi-memory or memory64
-    /*
+     * proposals are to be implemented */
-#if WASM_ENABLE_MEMORY64 != 0
+    bool is_memory64 = false;
-    uint64 shared_heap_start_off =
+    uint32 memidx = 0;
-        shared_heap ? (is_memory64 ? shared_heap->start_off_mem64
+    (void)is_memory64;
-                                   : shared_heap->start_off_mem32)
+    (void)memidx;
-                    : 0;
-    uint64 shared_heap_end_off =
-        shared_heap ? (is_memory64 ? UINT64_MAX : UINT32_MAX) : 0;
-#else
-    */ /* TODO: uncomment the code when memory64 is enabled for fast-interp */
-    uint64 shared_heap_start_off =
-        shared_heap ? shared_heap->start_off_mem32 : 0;
-    uint64 shared_heap_end_off = shared_heap ? UINT32_MAX : 0;
 /* #endif */
 #endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 
@@ -1911,7 +1887,7 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
                 }
                 HANDLE_OP_END();
             }
-#if WASM_ENABLE_SIMD != 0
+#if WASM_ENABLE_SIMDE != 0
             HANDLE_OP(WASM_OP_SELECT_128)
             {
                 cond = frame_lp[GET_OFFSET()];
@@ -5374,12 +5350,14 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
                                       || init_values[i].init_expr_type
                                              == INIT_EXPR_TYPE_FUNCREF_CONST);
 #if WASM_ENABLE_GC == 0
-                            table_elems[i] =
+                            table_elems[i] = (table_elem_type_t)init_values[i]
-                                (table_elem_type_t)init_values[i].u.ref_index;
+                                                 .u.unary.v.ref_index;
 #else
-                            if (init_values[i].u.ref_index != UINT32_MAX) {
+                            if (init_values[i].u.unary.v.ref_index
+                                != UINT32_MAX) {
                                 if (!(func_obj = wasm_create_func_obj(
-                                          module, init_values[i].u.ref_index,
+                                          module,
+                                          init_values[i].u.unary.v.ref_index,
                                           true, NULL, 0))) {
                                     goto got_exception;
                                 }

core/iwasm/interpreter/wasm_loader.c

@@ -453,6 +453,9 @@ typedef struct InitValue {
     WASMRefType ref_type;
 #endif
     WASMValue value;
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    InitializerExpression *expr;
+#endif
 } InitValue;
 
 typedef struct ConstExprContext {
@@ -477,7 +480,11 @@ push_const_expr_stack(ConstExprContext *ctx, uint8 flag, uint8 type,
 #if WASM_ENABLE_GC != 0
                       WASMRefType *ref_type, uint8 gc_opcode,
 #endif
-                      WASMValue *value, char *error_buf, uint32 error_buf_size)
+                      WASMValue *value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                      InitializerExpression *expr,
+#endif
+                      char *error_buf, uint32 error_buf_size)
 {
     InitValue *cur_value;
 
@@ -503,6 +510,10 @@ push_const_expr_stack(ConstExprContext *ctx, uint8 flag, uint8 type,
     cur_value->flag = flag;
     cur_value->value = *value;
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    cur_value->expr = expr;
+#endif
+
 #if WASM_ENABLE_GC != 0
     cur_value->gc_opcode = gc_opcode;
     if (wasm_is_type_multi_byte_type(type)) {
@@ -587,7 +598,11 @@ pop_const_expr_stack(ConstExprContext *ctx, uint8 *p_flag, uint8 type,
 #if WASM_ENABLE_GC != 0
                      WASMRefType *ref_type, uint8 *p_gc_opcode,
 #endif
-                     WASMValue *p_value, char *error_buf, uint32 error_buf_size)
+                     WASMValue *p_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                     InitializerExpression **p_expr,
+#endif
+                     char *error_buf, uint32 error_buf_size)
 {
     InitValue *cur_value;
 
@@ -623,7 +638,10 @@ pop_const_expr_stack(ConstExprContext *ctx, uint8 *p_flag, uint8 type,
     if (p_gc_opcode)
         *p_gc_opcode = cur_value->gc_opcode;
 #endif
-
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (p_expr)
+        *p_expr = cur_value->expr;
+#endif
     return true;
 
 #if WASM_ENABLE_GC != 0
@@ -639,7 +657,7 @@ fail:
 }
 
 static void
-destroy_const_expr_stack(ConstExprContext *ctx)
+destroy_const_expr_stack(ConstExprContext *ctx, bool free_exprs)
 {
 #if WASM_ENABLE_GC != 0
     uint32 i;
@@ -654,24 +672,62 @@ destroy_const_expr_stack(ConstExprContext *ctx)
         }
     }
 #endif
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (free_exprs) {
+        for (uint32 j = 0; j < ctx->sp; j++) {
+            if (is_expr_binary_op(ctx->stack[j].expr->init_expr_type)) {
+                destroy_init_expr_recursive(ctx->stack[j].expr);
+                ctx->stack[j].expr = NULL;
+            }
+        }
+    }
+#endif
 
     if (ctx->stack != ctx->data) {
         wasm_runtime_free(ctx->stack);
     }
 }
 
-#if WASM_ENABLE_GC != 0
+#if WASM_ENABLE_GC != 0 || WASM_ENABLE_EXTENDED_CONST_EXPR != 0
 static void
 destroy_init_expr(WASMModule *module, InitializerExpression *expr)
 {
+#if WASM_ENABLE_GC != 0
     if (expr->init_expr_type == INIT_EXPR_TYPE_STRUCT_NEW
         || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW
         || expr->init_expr_type == INIT_EXPR_TYPE_ARRAY_NEW_FIXED) {
-        destroy_init_expr_data_recursive(module, expr->u.data);
+        destroy_init_expr_data_recursive(module, expr->u.unary.v.data);
     }
-}
+#endif
-#endif /* end of WASM_ENABLE_GC != 0 */
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    // free left expr and right exprs for binary oprand
+    if (!is_expr_binary_op(expr->init_expr_type)) {
+        return;
+    }
+    if (expr->u.binary.l_expr) {
+        destroy_init_expr_recursive(expr->u.binary.l_expr);
+    }
+    if (expr->u.binary.r_expr) {
+        destroy_init_expr_recursive(expr->u.binary.r_expr);
+    }
+    expr->u.binary.l_expr = expr->u.binary.r_expr = NULL;
+#endif
+}
+#endif
+
+/* for init expr
+ *    (data (i32.add (i32.const 0) (i32.sub (i32.const 1) (i32.const 2)))),
+ *   the binary format is
+ *       0x11: 41 00 ; i32.const 0
+ *       0x13: 41 01 ; i32.const 1
+ *       0x15: 41 02 ; i32.const 2
+ *       0x17: 6b    ; i32.sub
+ *       0x18: 6a    ; i32.add
+ *   for traversal: read opcodes and push them onto the stack. When encountering
+ *   a binary opcode, pop two values from the stack which become the left and
+ *  right child nodes of this binary operation node.
+ */
 static bool
 load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                InitializerExpression *init_expr, uint8 type, void *ref_type,
@@ -687,6 +743,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
     uint8 opcode;
     WASMRefType cur_ref_type = { 0 };
 #endif
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    InitializerExpression *cur_expr = NULL;
+#endif
 
     init_const_expr_stack(&const_expr_ctx, module);
 
@@ -699,24 +758,32 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
             case INIT_EXPR_TYPE_I32_CONST:
                 read_leb_int32(p, p_end, cur_value.i32);
 
-                if (!push_const_expr_stack(
+                if (!push_const_expr_stack(&const_expr_ctx, flag,
-                        &const_expr_ctx, flag, VALUE_TYPE_I32,
+                                           VALUE_TYPE_I32,
 #if WASM_ENABLE_GC != 0
                                            NULL, 0,
 #endif
-                        &cur_value, error_buf, error_buf_size))
+                                           &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
                 break;
             /* i64.const */
             case INIT_EXPR_TYPE_I64_CONST:
                 read_leb_int64(p, p_end, cur_value.i64);
 
-                if (!push_const_expr_stack(
+                if (!push_const_expr_stack(&const_expr_ctx, flag,
-                        &const_expr_ctx, flag, VALUE_TYPE_I64,
+                                           VALUE_TYPE_I64,
 #if WASM_ENABLE_GC != 0
                                            NULL, 0,
 #endif
-                        &cur_value, error_buf, error_buf_size))
+                                           &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
                 break;
             /* f32.const */
@@ -726,12 +793,16 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                 for (i = 0; i < sizeof(float32); i++)
                     *p_float++ = *p++;
 
-                if (!push_const_expr_stack(
+                if (!push_const_expr_stack(&const_expr_ctx, flag,
-                        &const_expr_ctx, flag, VALUE_TYPE_F32,
+                                           VALUE_TYPE_F32,
 #if WASM_ENABLE_GC != 0
                                            NULL, 0,
 #endif
-                        &cur_value, error_buf, error_buf_size))
+                                           &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
                 break;
             /* f64.const */
@@ -741,12 +812,16 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                 for (i = 0; i < sizeof(float64); i++)
                     *p_float++ = *p++;
 
-                if (!push_const_expr_stack(
+                if (!push_const_expr_stack(&const_expr_ctx, flag,
-                        &const_expr_ctx, flag, VALUE_TYPE_F64,
+                                           VALUE_TYPE_F64,
 #if WASM_ENABLE_GC != 0
                                            NULL, 0,
 #endif
-                        &cur_value, error_buf, error_buf_size))
+                                           &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
                 break;
 #if WASM_ENABLE_SIMD != 0
@@ -767,12 +842,16 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                 cur_value.v128.i64x2[0] = high;
                 cur_value.v128.i64x2[1] = low;
 
-                if (!push_const_expr_stack(
+                if (!push_const_expr_stack(&const_expr_ctx, flag,
-                        &const_expr_ctx, flag, VALUE_TYPE_V128,
+                                           VALUE_TYPE_V128,
 #if WASM_ENABLE_GC != 0
                                            NULL, 0,
 #endif
-                        &cur_value, error_buf, error_buf_size))
+                                           &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
 #if WASM_ENABLE_WAMR_COMPILER != 0
                 /* If any init_expr is v128.const, mark SIMD used */
@@ -783,7 +862,92 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 #endif /* end of (WASM_ENABLE_WAMR_COMPILER != 0) || (WASM_ENABLE_JIT != 0) || \
           (WASM_ENABLE_FAST_INTERP != 0) */
 #endif /* end of WASM_ENABLE_SIMD */
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            case INIT_EXPR_TYPE_I32_ADD:
+            case INIT_EXPR_TYPE_I32_SUB:
+            case INIT_EXPR_TYPE_I32_MUL:
+            case INIT_EXPR_TYPE_I64_ADD:
+            case INIT_EXPR_TYPE_I64_SUB:
+            case INIT_EXPR_TYPE_I64_MUL:
+            {
 
+                InitializerExpression *l_expr, *r_expr;
+                WASMValue l_value, r_value;
+                uint8 l_flag, r_flag;
+                uint8 value_type;
+
+                if (flag == INIT_EXPR_TYPE_I32_ADD
+                    || flag == INIT_EXPR_TYPE_I32_SUB
+                    || flag == INIT_EXPR_TYPE_I32_MUL) {
+                    value_type = VALUE_TYPE_I32;
+                }
+                else {
+                    value_type = VALUE_TYPE_I64;
+                }
+
+                /* If right flag indicates a binary operation, right expr will
+                 * be popped from stack. Otherwise, allocate a new expr for
+                 * right expr. Same for left expr.
+                 */
+                if (!(pop_const_expr_stack(&const_expr_ctx, &r_flag, value_type,
+#if WASM_ENABLE_GC != 0
+                                           NULL, NULL,
+#endif
+                                           &r_value, &r_expr, error_buf,
+                                           error_buf_size))) {
+                    goto fail;
+                }
+                if (!is_expr_binary_op(r_flag)) {
+                    if (!(r_expr = loader_malloc(sizeof(InitializerExpression),
+                                                 error_buf, error_buf_size))) {
+                        goto fail;
+                    }
+                    r_expr->init_expr_type = r_flag;
+                    r_expr->u.unary.v = r_value;
+                }
+
+                if (!(pop_const_expr_stack(&const_expr_ctx, &l_flag, value_type,
+#if WASM_ENABLE_GC != 0
+                                           NULL, NULL,
+#endif
+                                           &l_value, &l_expr, error_buf,
+                                           error_buf_size))) {
+                    destroy_init_expr_recursive(r_expr);
+                    goto fail;
+                }
+                if (!is_expr_binary_op(l_flag)) {
+                    if (!(l_expr = loader_malloc(sizeof(InitializerExpression),
+                                                 error_buf, error_buf_size))) {
+                        destroy_init_expr_recursive(r_expr);
+                        goto fail;
+                    }
+                    l_expr->init_expr_type = l_flag;
+                    l_expr->u.unary.v = l_value;
+                }
+
+                if (!(cur_expr = loader_malloc(sizeof(InitializerExpression),
+                                               error_buf, error_buf_size))) {
+                    destroy_init_expr_recursive(l_expr);
+                    destroy_init_expr_recursive(r_expr);
+                    goto fail;
+                }
+                cur_expr->init_expr_type = flag;
+                cur_expr->u.binary.l_expr = l_expr;
+                cur_expr->u.binary.r_expr = r_expr;
+
+                if (!push_const_expr_stack(&const_expr_ctx, flag, value_type,
+#if WASM_ENABLE_GC != 0
+                                           NULL, 0,
+#endif
+                                           &cur_value, cur_expr, error_buf,
+                                           error_buf_size)) {
+                    destroy_init_expr_recursive(cur_expr);
+                    goto fail;
+                }
+
+                break;
+            }
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR */
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
             /* ref.func */
             case INIT_EXPR_TYPE_FUNCREF_CONST:
@@ -799,6 +963,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 #if WASM_ENABLE_GC == 0
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_FUNCREF, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size))
                     goto fail;
 #else
@@ -816,8 +983,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                                              false, type_idx);
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            cur_ref_type.ref_type, &cur_ref_type,
-                                           0, &cur_value, error_buf,
+                                           0, &cur_value,
-                                           error_buf_size))
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
 #endif
 #if WASM_ENABLE_WAMR_COMPILER != 0
@@ -829,45 +999,71 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
             /* ref.null */
             case INIT_EXPR_TYPE_REFNULL_CONST:
             {
-                uint8 type1;
-
 #if WASM_ENABLE_GC == 0
+                uint8 type1;
                 CHECK_BUF(p, p_end, 1);
                 type1 = read_uint8(p);
-
                 cur_value.ref_index = NULL_REF;
                 if (!push_const_expr_stack(&const_expr_ctx, flag, type1,
-                                           &cur_value, error_buf,
+                                           &cur_value,
-                                           error_buf_size))
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
-                    goto fail;
+                                           NULL,
-#else
+#endif
-                int32 heap_type;
-                read_leb_int32(p, p_end, heap_type);
-                type1 = (uint8)((int32)0x80 + heap_type);
-
-                cur_value.gc_obj = NULL_REF;
-
-                if (!is_byte_a_type(type1)
-                    || !wasm_is_valid_heap_type(heap_type)
-                    || wasm_is_type_multi_byte_type(type1)) {
-                    p--;
-                    read_leb_uint32(p, p_end, type_idx);
-                    if (!check_type_index(module, module->type_count, type_idx,
                                            error_buf, error_buf_size))
                     goto fail;
+#else
+                /*
+                 * According to the current GC SPEC rules, the heap_type must be
+                 * validated when ref.null is used. It can be an absheaptype,
+                 * or the type C.types[type_idx] must be defined in the context.
+                 */
+                int32 heap_type;
+                read_leb_int32(p, p_end, heap_type);
+                cur_value.gc_obj = NULL_REF;
 
+                /*
+                 * The current check of heap_type can deterministically infer
+                 * the result of the previous condition
+                 * `(!is_byte_a_type(type1) ||
+                 * wasm_is_type_multi_byte_type(type1))`. Therefore, the
+                 * original condition is redundant and has been removed.
+                 *
+                 * This logic is consistent with the implementation of the
+                 * `WASM_OP_REF_NULL` case in the `wasm_loader_prepare_bytecode`
+                 * function.
+                 */
+
+                if (heap_type >= 0) {
+                    if (!check_type_index(module, module->type_count, heap_type,
+                                          error_buf, error_buf_size)) {
+                        goto fail;
+                    }
                     wasm_set_refheaptype_typeidx(&cur_ref_type.ref_ht_typeidx,
-                                                 true, type_idx);
+                                                 true, heap_type);
                     if (!push_const_expr_stack(&const_expr_ctx, flag,
                                                cur_ref_type.ref_type,
                                                &cur_ref_type, 0, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                               NULL,
+#endif
                                                error_buf, error_buf_size))
                         goto fail;
                 }
                 else {
-                    if (!push_const_expr_stack(&const_expr_ctx, flag, type1,
+                    if (!wasm_is_valid_heap_type(heap_type)) {
-                                               NULL, 0, &cur_value, error_buf,
+                        set_error_buf_v(error_buf, error_buf_size,
-                                               error_buf_size))
+                                        "unknown type %d", heap_type);
+                        goto fail;
+                    }
+                    cur_ref_type.ref_ht_common.ref_type =
+                        (uint8)((int32)0x80 + heap_type);
+                    if (!push_const_expr_stack(&const_expr_ctx, flag,
+                                               cur_ref_type.ref_type, NULL, 0,
+                                               &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                               NULL,
+#endif
+                                               error_buf, error_buf_size))
                         goto fail;
                 }
 #endif
@@ -956,8 +1152,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 #if WASM_ENABLE_GC != 0
                                            &cur_ref_type, 0,
 #endif
-                                           &cur_value, error_buf,
+                                           &cur_value,
-                                           error_buf_size))
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
                     goto fail;
 
                 break;
@@ -1020,6 +1219,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                                     &const_expr_ctx, NULL, field_type,
                                     field_ref_type, NULL,
                                     &struct_init_values->fields[field_idx],
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                    NULL,
+#endif
                                     error_buf, error_buf_size)) {
                                 destroy_init_expr_data_recursive(
                                     module, struct_init_values);
@@ -1033,6 +1235,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                         if (!push_const_expr_stack(
                                 &const_expr_ctx, flag, cur_ref_type.ref_type,
                                 &cur_ref_type, (uint8)opcode1, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                NULL,
+#endif
                                 error_buf, error_buf_size)) {
                             destroy_init_expr_data_recursive(
                                 module, struct_init_values);
@@ -1064,6 +1269,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                         if (!push_const_expr_stack(
                                 &const_expr_ctx, flag, cur_ref_type.ref_type,
                                 &cur_ref_type, (uint8)opcode1, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                NULL,
+#endif
                                 error_buf, error_buf_size)) {
                             goto fail;
                         }
@@ -1112,8 +1320,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                                 if (!pop_const_expr_stack(
                                         &const_expr_ctx, NULL, VALUE_TYPE_I32,
-                                        NULL, NULL, &len_val, error_buf,
+                                        NULL, NULL, &len_val,
-                                        error_buf_size)) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                        NULL,
+#endif
+                                        error_buf, error_buf_size)) {
                                     goto fail;
                                 }
 
@@ -1132,6 +1343,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                                         &const_expr_ctx, NULL, elem_type,
                                         elem_ref_type, NULL,
                                         &array_init_values->elem_data[0],
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                        NULL,
+#endif
                                         error_buf, error_buf_size)) {
                                     destroy_init_expr_data_recursive(
                                         module, array_init_values);
@@ -1164,6 +1378,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                                             elem_ref_type, NULL,
                                             &array_init_values
                                                  ->elem_data[i - 1],
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                            NULL,
+#endif
                                             error_buf, error_buf_size)) {
                                         destroy_init_expr_data_recursive(
                                             module, array_init_values);
@@ -1180,10 +1397,13 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                             uint32 len;
 
                             /* POP(i32) */
-                            if (!pop_const_expr_stack(&const_expr_ctx, NULL,
+                            if (!pop_const_expr_stack(
-                                                      VALUE_TYPE_I32, NULL,
+                                    &const_expr_ctx, NULL, VALUE_TYPE_I32, NULL,
-                                                      NULL, &len_val, error_buf,
+                                    NULL, &len_val,
-                                                      error_buf_size)) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                    NULL,
+#endif
+                                    error_buf, error_buf_size)) {
                                 goto fail;
                             }
                             len = len_val.i32;
@@ -1197,6 +1417,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                         if (!push_const_expr_stack(
                                 &const_expr_ctx, flag, cur_ref_type.ref_type,
                                 &cur_ref_type, (uint8)opcode1, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                NULL,
+#endif
                                 error_buf, error_buf_size)) {
                             if (array_init_values) {
                                 destroy_init_expr_data_recursive(
@@ -1223,9 +1446,13 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                     case WASM_OP_REF_I31:
                     {
                         /* POP(i32) */
-                        if (!pop_const_expr_stack(
+                        if (!pop_const_expr_stack(&const_expr_ctx, NULL,
-                                &const_expr_ctx, NULL, VALUE_TYPE_I32, NULL,
+                                                  VALUE_TYPE_I32, NULL, NULL,
-                                NULL, &cur_value, error_buf, error_buf_size)) {
+                                                  &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                                  NULL,
+#endif
+                                                  error_buf, error_buf_size)) {
                             goto fail;
                         }
 
@@ -1234,6 +1461,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                         if (!push_const_expr_stack(
                                 &const_expr_ctx, flag, cur_ref_type.ref_type,
                                 &cur_ref_type, (uint8)opcode1, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                NULL,
+#endif
                                 error_buf, error_buf_size)) {
                             goto fail;
                         }
@@ -1268,7 +1498,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 #if WASM_ENABLE_GC != 0
                               ref_type, &opcode,
 #endif
-                              &cur_value, error_buf, error_buf_size)) {
+                              &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                              &cur_expr,
+#endif
+                              error_buf, error_buf_size)) {
         goto fail;
     }
 
@@ -1278,8 +1512,21 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
         goto fail;
     }
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (cur_expr != NULL) {
+        bh_memcpy_s(init_expr, sizeof(InitializerExpression), cur_expr,
+                    sizeof(InitializerExpression));
+        wasm_runtime_free(cur_expr);
+    }
+    else {
         init_expr->init_expr_type = flag;
-    init_expr->u = cur_value;
+        init_expr->u.unary.v = cur_value;
+    }
+
+#else
+    init_expr->init_expr_type = flag;
+    init_expr->u.unary.v = cur_value;
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */
 
 #if WASM_ENABLE_GC != 0
     if (init_expr->init_expr_type == WASM_OP_GC_PREFIX) {
@@ -1310,11 +1557,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 #endif /* end of WASM_ENABLE_GC != 0 */
 
     *p_buf = p;
-    destroy_const_expr_stack(&const_expr_ctx);
+    destroy_const_expr_stack(&const_expr_ctx, false);
     return true;
 
 fail:
-    destroy_const_expr_stack(&const_expr_ctx);
+    destroy_const_expr_stack(&const_expr_ctx, true);
     return false;
 }
 
@@ -2042,9 +2289,9 @@ load_type_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
                                       "recursive type count too large");
                         return false;
                     }
-                    module->type_count += rec_count - 1;
                     new_total_size =
-                        sizeof(WASMFuncType *) * (uint64)module->type_count;
+                        sizeof(WASMFuncType *)
+                        * (uint64)(module->type_count + rec_count - 1);
                     if (new_total_size > UINT32_MAX) {
                         set_error_buf(error_buf, error_buf_size,
                                       "allocate memory failed");
@@ -2052,13 +2299,19 @@ load_type_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
                     }
                     MEM_REALLOC(module->types, (uint32)total_size,
                                 (uint32)new_total_size);
+                    module->type_count += rec_count - 1;
                     total_size = new_total_size;
                 }
 
+                if (rec_count < 1) {
+                    LOG_VERBOSE("Processing 0-entry rec group");
+                }
+                else {
                     LOG_VERBOSE("Processing rec group [%d-%d]",
                                 processed_type_count,
                                 processed_type_count + rec_count - 1);
                 }
+            }
             else {
                 p--;
             }
@@ -2243,6 +2496,13 @@ load_type_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
 #endif /* end of WASM_ENABLE_GC == 0 */
     }
 
+    for (i = 0; i < module->type_count; i++) {
+        if (module->types[i] == NULL) {
+            set_error_buf_v(error_buf, error_buf_size, "unknown type %d", i);
+            return false;
+        }
+    }
+
     if (p != p_end) {
         set_error_buf(error_buf, error_buf_size, "section size mismatch");
         return false;
@@ -3351,7 +3611,8 @@ load_import_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
                     /* valtype */
                     CHECK_BUF(p, p_end, 1);
                     global_type = read_uint8(p);
-                    if (wasm_is_reftype_htref_nullable(global_type)) {
+                    if (wasm_is_reftype_htref_nullable(global_type)
+                        || wasm_is_reftype_htref_non_nullable(global_type)) {
                         int32 heap_type;
                         read_leb_int32(p, p_end, heap_type);
                         (void)heap_type;
@@ -4070,9 +4331,9 @@ load_global_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
             if (global->init_expr.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL) {
                 uint8 global_type;
                 WASMRefType *global_ref_type;
-                uint32 global_idx = global->init_expr.u.global_index;
+                uint32 global_idx = global->init_expr.u.unary.v.global_index;
 
-                if (global->init_expr.u.global_index
+                if (global->init_expr.u.unary.v.global_index
                     >= module->import_global_count + i) {
                     set_error_buf(error_buf, error_buf_size, "unknown global");
                     return false;
@@ -4469,7 +4730,7 @@ load_func_index_vec(const uint8 **p_buf, const uint8 *buf_end,
         }
 
         init_expr->init_expr_type = INIT_EXPR_TYPE_FUNCREF_CONST;
-        init_expr->u.ref_index = function_index;
+        init_expr->u.unary.v.ref_index = function_index;
     }
 
     *p_buf = p;
@@ -4742,7 +5003,7 @@ load_table_segment_section(const uint8 *buf, const uint8 *buf_end,
 
 #if WASM_ENABLE_MEMORY64 != 0
             if (table_elem_idx_type == VALUE_TYPE_I64
-                && table_segment->base_offset.u.u64 > UINT32_MAX) {
+                && table_segment->base_offset.u.unary.v.u64 > UINT32_MAX) {
                 set_error_buf(error_buf, error_buf_size,
                               "In table64, table base offset can't be "
                               "larger than UINT32_MAX");
@@ -4902,6 +5163,9 @@ load_data_segment_section(const uint8 *buf, const uint8 *buf_end,
 
             if (!(dataseg = module->data_segments[i] = loader_malloc(
                       sizeof(WASMDataSeg), error_buf, error_buf_size))) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                destroy_init_expr(module, &init_expr);
+#endif
                 return false;
             }
 
@@ -6029,7 +6293,8 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                     && global->init_expr.init_expr_type
                            == INIT_EXPR_TYPE_I32_CONST) {
                     aux_heap_base_global = global;
-                    aux_heap_base = (uint64)(uint32)global->init_expr.u.i32;
+                    aux_heap_base =
+                        (uint64)(uint32)global->init_expr.u.unary.v.i32;
                     aux_heap_base_global_index = export->index;
                     LOG_VERBOSE("Found aux __heap_base global, value: %" PRIu64,
                                 aux_heap_base);
@@ -6050,7 +6315,8 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                     && global->init_expr.init_expr_type
                            == INIT_EXPR_TYPE_I32_CONST) {
                     aux_data_end_global = global;
-                    aux_data_end = (uint64)(uint32)global->init_expr.u.i32;
+                    aux_data_end =
+                        (uint64)(uint32)global->init_expr.u.unary.v.i32;
                     aux_data_end_global_index = export->index;
                     LOG_VERBOSE("Found aux __data_end global, value: %" PRIu64,
                                 aux_data_end);
@@ -6091,10 +6357,11 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                         && global->type.val_type == VALUE_TYPE_I32
                         && global->init_expr.init_expr_type
                                == INIT_EXPR_TYPE_I32_CONST
-                        && (uint64)(uint32)global->init_expr.u.i32
+                        && (uint64)(uint32)global->init_expr.u.unary.v.i32
                                <= aux_heap_base) {
                         aux_stack_top_global = global;
-                        aux_stack_top = (uint64)(uint32)global->init_expr.u.i32;
+                        aux_stack_top =
+                            (uint64)(uint32)global->init_expr.u.unary.v.i32;
                         module->aux_stack_top_global_index =
                             module->import_global_count + global_index;
                         module->aux_stack_bottom = aux_stack_top;
@@ -6945,7 +7212,7 @@ wasm_loader_unload(WASMModule *module)
         wasm_runtime_free(module->memories);
 
     if (module->globals) {
-#if WASM_ENABLE_GC != 0
+#if WASM_ENABLE_GC != 0 || WASM_ENABLE_EXTENDED_CONST_EXPR != 0
         for (i = 0; i < module->global_count; i++) {
             destroy_init_expr(module, &module->globals[i].init_expr);
         }
@@ -6978,6 +7245,9 @@ wasm_loader_unload(WASMModule *module)
 #endif
                 wasm_runtime_free(module->table_segments[i].init_values);
             }
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            destroy_init_expr(module, &module->table_segments[i].base_offset);
+#endif
         }
         wasm_runtime_free(module->table_segments);
     }
@@ -6987,6 +7257,10 @@ wasm_loader_unload(WASMModule *module)
             if (module->data_segments[i]) {
                 if (module->data_segments[i]->is_data_cloned)
                     wasm_runtime_free(module->data_segments[i]->data);
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                destroy_init_expr(module,
+                                  &(module->data_segments[i]->base_offset));
+#endif
                 wasm_runtime_free(module->data_segments[i]);
             }
         }
@@ -7356,7 +7630,7 @@ wasm_loader_find_block_addr(WASMExecEnv *exec_env, BlockAddr *block_addr_cache,
             case WASM_OP_SELECT:
             case WASM_OP_DROP_64:
             case WASM_OP_SELECT_64:
-#if WASM_ENABLE_FAST_INTERP != 0 && WASM_ENABLE_SIMD != 0
+#if WASM_ENABLE_SIMDE != 0
             case WASM_OP_SELECT_128:
 #endif
                 break;
@@ -12881,9 +13155,7 @@ re_scan:
 #if (WASM_ENABLE_WAMR_COMPILER != 0) || (WASM_ENABLE_JIT != 0) \
     || (WASM_ENABLE_FAST_INTERP != 0)
                         case VALUE_TYPE_V128:
-#if WASM_ENABLE_FAST_INTERP == 0
+#if WASM_ENABLE_SIMDE != 0
-                            *(p - 1) = WASM_OP_SELECT_128;
-#else
                             if (loader_ctx->p_code_compiled) {
                                 uint8 opcode_tmp = WASM_OP_SELECT_128;
 #if WASM_ENABLE_LABELS_AS_VALUES != 0
@@ -13011,8 +13283,7 @@ re_scan:
                     uint8 opcode_tmp = WASM_OP_SELECT;
 
                     if (type == VALUE_TYPE_V128) {
-#if WASM_ENABLE_JIT != 0 \
+#if WASM_ENABLE_SIMDE != 0
-    || WASM_ENABLE_FAST_INTERP != 0 && WASM_ENABLE_SIMD != 0
                         opcode_tmp = WASM_OP_SELECT_128;
 #else
                         set_error_buf(error_buf, error_buf_size,
@@ -13258,7 +13529,8 @@ re_scan:
                                 == VALUE_TYPE_FUNCREF
                             && module->globals[i].init_expr.init_expr_type
                                    == INIT_EXPR_TYPE_FUNCREF_CONST
-                            && module->globals[i].init_expr.u.u32 == func_idx) {
+                            && module->globals[i].init_expr.u.unary.v.u32
+                                   == func_idx) {
                             func_declared = true;
                             break;
                         }
@@ -13287,7 +13559,8 @@ re_scan:
 #endif
                             ) {
                                 for (j = 0; j < table_seg->value_count; j++) {
-                                    if (table_seg->init_values[j].u.ref_index
+                                    if (table_seg->init_values[j]
+                                            .u.unary.v.ref_index
                                         == func_idx) {
                                         func_declared = true;
                                         break;
@@ -15023,8 +15296,6 @@ re_scan:
                     case WASM_OP_STRING_NEW_LOSSY_UTF8:
                     case WASM_OP_STRING_NEW_WTF8:
                     {
-                        uint32 memidx;
-
 #if WASM_ENABLE_JIT != 0 || WASM_ENABLE_WAMR_COMPILER != 0
                         func->has_memory_operations = true;
 #endif
@@ -15036,7 +15307,6 @@ re_scan:
                         POP_I32();
                         POP_I32();
                         PUSH_REF(REF_TYPE_STRINGREF);
-                        (void)memidx;
                         break;
                     }
                     case WASM_OP_STRING_CONST:
@@ -15064,8 +15334,6 @@ re_scan:
                     case WASM_OP_STRING_ENCODE_LOSSY_UTF8:
                     case WASM_OP_STRING_ENCODE_WTF8:
                     {
-                        uint32 memidx;
-
 #if WASM_ENABLE_JIT != 0 || WASM_ENABLE_WAMR_COMPILER != 0
                         func->has_memory_operations = true;
 #endif
@@ -15077,7 +15345,6 @@ re_scan:
                         POP_I32();
                         POP_STRINGREF();
                         PUSH_I32();
-                        (void)memidx;
                         break;
                     }
                     case WASM_OP_STRING_CONCAT:
@@ -15118,8 +15385,6 @@ re_scan:
                     case WASM_OP_STRINGVIEW_WTF8_ENCODE_LOSSY_UTF8:
                     case WASM_OP_STRINGVIEW_WTF8_ENCODE_WTF8:
                     {
-                        uint32 memidx;
-
 #if WASM_ENABLE_JIT != 0 || WASM_ENABLE_WAMR_COMPILER != 0
                         func->has_memory_operations = true;
 #endif
@@ -15134,7 +15399,6 @@ re_scan:
                         POP_REF(REF_TYPE_STRINGVIEWWTF8);
                         PUSH_I32();
                         PUSH_I32();
-                        (void)memidx;
                         break;
                     }
                     case WASM_OP_STRINGVIEW_WTF8_SLICE:
@@ -15166,8 +15430,6 @@ re_scan:
                     }
                     case WASM_OP_STRINGVIEW_WTF16_ENCODE:
                     {
-                        uint32 memidx;
-
 #if WASM_ENABLE_JIT != 0 || WASM_ENABLE_WAMR_COMPILER != 0
                         func->has_memory_operations = true;
 #endif
@@ -15181,7 +15443,6 @@ re_scan:
                         POP_I32();
                         POP_REF(REF_TYPE_STRINGVIEWWTF16);
                         PUSH_I32();
-                        (void)memidx;
                         break;
                     }
                     case WASM_OP_STRINGVIEW_WTF16_SLICE:

core/iwasm/interpreter/wasm_mini_loader.c

@@ -261,6 +261,9 @@ typedef struct InitValue {
     uint8 type;
     uint8 flag;
     WASMValue value;
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    InitializerExpression *expr;
+#endif
 } InitValue;
 
 typedef struct ConstExprContext {
@@ -282,7 +285,11 @@ init_const_expr_stack(ConstExprContext *ctx, WASMModule *module)
 
 static bool
 push_const_expr_stack(ConstExprContext *ctx, uint8 flag, uint8 type,
-                      WASMValue *value, char *error_buf, uint32 error_buf_size)
+                      WASMValue *value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                      InitializerExpression *expr,
+#endif
+                      char *error_buf, uint32 error_buf_size)
 {
     InitValue *cur_value;
 
@@ -305,6 +312,9 @@ push_const_expr_stack(ConstExprContext *ctx, uint8 flag, uint8 type,
     cur_value->type = type;
     cur_value->flag = flag;
     cur_value->value = *value;
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    cur_value->expr = expr;
+#endif
 
     return true;
 fail:
@@ -313,7 +323,11 @@ fail:
 
 static bool
 pop_const_expr_stack(ConstExprContext *ctx, uint8 *p_flag, uint8 type,
-                     WASMValue *p_value, char *error_buf, uint32 error_buf_size)
+                     WASMValue *p_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                     InitializerExpression **p_expr,
+#endif
+                     char *error_buf, uint32 error_buf_size)
 {
     InitValue *cur_value;
 
@@ -331,18 +345,50 @@ pop_const_expr_stack(ConstExprContext *ctx, uint8 *p_flag, uint8 type,
         *p_flag = cur_value->flag;
     if (p_value)
         *p_value = cur_value->value;
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (p_expr)
+        *p_expr = cur_value->expr;
+#endif
 
     return true;
 }
 
 static void
-destroy_const_expr_stack(ConstExprContext *ctx)
+destroy_const_expr_stack(ConstExprContext *ctx, bool free_exprs)
 {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (free_exprs) {
+        for (uint32 j = 0; j < ctx->sp; j++) {
+            if (is_expr_binary_op(ctx->stack[j].expr->init_expr_type)) {
+                destroy_init_expr_recursive(ctx->stack[j].expr);
+                ctx->stack[j].expr = NULL;
+            }
+        }
+    }
+#endif
     if (ctx->stack != ctx->data) {
         wasm_runtime_free(ctx->stack);
     }
 }
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+static void
+destroy_init_expr(InitializerExpression *expr)
+{
+    // free left expr and right exprs for binary oprand
+    if (is_expr_binary_op(expr->init_expr_type)) {
+        return;
+    }
+    if (expr->u.binary.l_expr) {
+        destroy_init_expr_recursive(expr->u.binary.l_expr);
+    }
+    if (expr->u.binary.r_expr) {
+        destroy_init_expr_recursive(expr->u.binary.r_expr);
+    }
+    expr->u.binary.l_expr = expr->u.binary.r_expr = NULL;
+}
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */
+
 static bool
 load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                InitializerExpression *init_expr, uint8 type, char *error_buf,
@@ -353,6 +399,9 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
     uint32 i;
     ConstExprContext const_expr_ctx = { 0 };
     WASMValue cur_value = { 0 };
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    InitializerExpression *cur_expr = NULL;
+#endif
 
     init_const_expr_stack(&const_expr_ctx, module);
 
@@ -367,8 +416,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_I32, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
                 break;
             /* i64.const */
@@ -377,8 +429,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_I64, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
                 break;
             /* f32.const */
@@ -390,8 +445,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_F32, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
                 break;
             /* f64.const */
@@ -403,8 +461,11 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_F64, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
                 break;
 
@@ -417,13 +478,16 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                 cur_value.ref_index = func_idx;
                 if (!check_function_index(module, func_idx, error_buf,
                                           error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag,
                                            VALUE_TYPE_FUNCREF, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                                           NULL,
+#endif
                                            error_buf, error_buf_size)) {
-                    bh_assert(0);
+                    goto fail;
                 }
                 break;
             }
@@ -438,9 +502,12 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
                 cur_value.ref_index = UINT32_MAX;
                 if (!push_const_expr_stack(&const_expr_ctx, flag, type1,
-                                           &cur_value, error_buf,
+                                           &cur_value,
-                                           error_buf_size)) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
-                    bh_assert(0);
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size)) {
+                    goto fail;
                 }
                 break;
             }
@@ -471,15 +538,93 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
                 }
 
                 if (!push_const_expr_stack(&const_expr_ctx, flag, global_type,
-                                           &cur_value, error_buf,
+                                           &cur_value,
-                                           error_buf_size))
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
-                    bh_assert(0);
+                                           NULL,
+#endif
+                                           error_buf, error_buf_size))
+                    goto fail;
 
                 break;
             }
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            case INIT_EXPR_TYPE_I32_ADD:
+            case INIT_EXPR_TYPE_I64_ADD:
+            case INIT_EXPR_TYPE_I32_SUB:
+            case INIT_EXPR_TYPE_I64_SUB:
+            case INIT_EXPR_TYPE_I32_MUL:
+            case INIT_EXPR_TYPE_I64_MUL:
+            {
+                InitializerExpression *l_expr, *r_expr;
+                WASMValue l_value, r_value;
+                uint8 l_flag, r_flag;
+                uint8 value_type;
+
+                if (flag == INIT_EXPR_TYPE_I32_ADD
+                    || flag == INIT_EXPR_TYPE_I32_SUB
+                    || flag == INIT_EXPR_TYPE_I32_MUL) {
+                    value_type = VALUE_TYPE_I32;
+                }
+                else {
+                    value_type = VALUE_TYPE_I64;
+                }
+
+                /* If right flag indicates a binary operation, right expr will
+                 * be popped from stack. Otherwise, allocate a new expr for
+                 * right expr. Same for left expr.
+                 */
+                if (!(pop_const_expr_stack(&const_expr_ctx, &r_flag, value_type,
+                                           &r_value, &r_expr, error_buf,
+                                           error_buf_size))) {
+                    goto fail;
+                }
+                if (!is_expr_binary_op(r_flag)) {
+                    if (!(r_expr = loader_malloc(sizeof(InitializerExpression),
+                                                 error_buf, error_buf_size))) {
+                        goto fail;
+                    }
+                    r_expr->init_expr_type = r_flag;
+                    r_expr->u.unary.v = r_value;
+                }
+
+                if (!(pop_const_expr_stack(&const_expr_ctx, &l_flag, value_type,
+                                           &l_value, &l_expr, error_buf,
+                                           error_buf_size))) {
+                    destroy_init_expr_recursive(r_expr);
+                    goto fail;
+                }
+                if (!is_expr_binary_op(l_flag)) {
+                    if (!(l_expr = loader_malloc(sizeof(InitializerExpression),
+                                                 error_buf, error_buf_size))) {
+                        destroy_init_expr_recursive(r_expr);
+                        goto fail;
+                    }
+                    l_expr->init_expr_type = l_flag;
+                    l_expr->u.unary.v = l_value;
+                }
+
+                if (!(cur_expr = loader_malloc(sizeof(InitializerExpression),
+                                               error_buf, error_buf_size))) {
+                    destroy_init_expr_recursive(l_expr);
+                    destroy_init_expr_recursive(r_expr);
+                    goto fail;
+                }
+                cur_expr->init_expr_type = flag;
+                cur_expr->u.binary.l_expr = l_expr;
+                cur_expr->u.binary.r_expr = r_expr;
+
+                if (!push_const_expr_stack(&const_expr_ctx, flag, value_type,
+                                           &cur_value, cur_expr, error_buf,
+                                           error_buf_size)) {
+                    destroy_init_expr_recursive(cur_expr);
+                    goto fail;
+                }
+                break;
+            }
+#endif
             default:
             {
-                bh_assert(0);
+                goto fail;
             }
         }
 
@@ -489,18 +634,42 @@ load_init_expr(WASMModule *module, const uint8 **p_buf, const uint8 *buf_end,
 
     /* There should be only one value left on the init value stack */
     if (!pop_const_expr_stack(&const_expr_ctx, &flag, type, &cur_value,
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                              &cur_expr,
+#endif
                               error_buf, error_buf_size)) {
-        bh_assert(0);
+        goto fail;
     }
 
-    bh_assert(const_expr_ctx.sp == 0);
+    if (const_expr_ctx.sp != 0) {
+        set_error_buf(error_buf, error_buf_size,
+                      "type mismatch: illegal constant opcode sequence");
+        goto fail;
+    }
 
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+    if (cur_expr != NULL) {
+        bh_memcpy_s(init_expr, sizeof(InitializerExpression), cur_expr,
+                    sizeof(InitializerExpression));
+        wasm_runtime_free(cur_expr);
+    }
+    else {
         init_expr->init_expr_type = flag;
-    init_expr->u = cur_value;
+        init_expr->u.unary.v = cur_value;
+    }
+
+#else
+    init_expr->init_expr_type = flag;
+    init_expr->u.unary.v = cur_value;
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */
 
     *p_buf = p;
-    destroy_const_expr_stack(&const_expr_ctx);
+    destroy_const_expr_stack(&const_expr_ctx, false);
     return true;
+
+fail:
+    destroy_const_expr_stack(&const_expr_ctx, true);
+    return false;
 }
 
 static bool
@@ -1385,13 +1554,14 @@ load_global_section(const uint8 *buf, const uint8 *buf_end, WASMModule *module,
                  * global.get instructions are
                  * only allowed to refer to imported globals.
                  */
-                uint32 target_global_index = global->init_expr.u.global_index;
+                uint32 target_global_index =
+                    global->init_expr.u.unary.v.global_index;
                 bh_assert(target_global_index < module->import_global_count);
                 (void)target_global_index;
             }
             else if (INIT_EXPR_TYPE_FUNCREF_CONST
                      == global->init_expr.init_expr_type) {
-                bh_assert(global->init_expr.u.ref_index
+                bh_assert(global->init_expr.u.unary.v.ref_index
                           < module->import_function_count
                                 + module->function_count);
             }
@@ -1575,7 +1745,7 @@ load_func_index_vec(const uint8 **p_buf, const uint8 *buf_end,
         }
 
         init_expr->init_expr_type = INIT_EXPR_TYPE_FUNCREF_CONST;
-        init_expr->u.ref_index = function_index;
+        init_expr->u.unary.v.ref_index = function_index;
     }
 
     *p_buf = p;
@@ -1890,6 +2060,9 @@ load_data_segment_section(const uint8 *buf, const uint8 *buf_end,
 
             if (!(dataseg = module->data_segments[i] = loader_malloc(
                       sizeof(WASMDataSeg), error_buf, error_buf_size))) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                destroy_init_expr(&init_expr);
+#endif
                 return false;
             }
 
@@ -2778,7 +2951,8 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                     && global->init_expr.init_expr_type
                            == INIT_EXPR_TYPE_I32_CONST) {
                     aux_heap_base_global = global;
-                    aux_heap_base = (uint64)(uint32)global->init_expr.u.i32;
+                    aux_heap_base =
+                        (uint64)(uint32)global->init_expr.u.unary.v.i32;
                     aux_heap_base_global_index = export->index;
                     LOG_VERBOSE("Found aux __heap_base global, value: %" PRIu64,
                                 aux_heap_base);
@@ -2798,7 +2972,8 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                     && global->init_expr.init_expr_type
                            == INIT_EXPR_TYPE_I32_CONST) {
                     aux_data_end_global = global;
-                    aux_data_end = (uint64)(uint32)global->init_expr.u.i32;
+                    aux_data_end =
+                        (uint64)(uint32)global->init_expr.u.unary.v.i32;
                     aux_data_end_global_index = export->index;
                     LOG_VERBOSE("Found aux __data_end global, value: %" PRIu64,
                                 aux_data_end);
@@ -2838,10 +3013,11 @@ load_from_sections(WASMModule *module, WASMSection *sections,
                         && global->type.val_type == VALUE_TYPE_I32
                         && global->init_expr.init_expr_type
                                == INIT_EXPR_TYPE_I32_CONST
-                        && (uint64)(uint32)global->init_expr.u.i32
+                        && (uint64)(uint32)global->init_expr.u.unary.v.i32
                                <= aux_heap_base) {
                         aux_stack_top_global = global;
-                        aux_stack_top = (uint64)(uint32)global->init_expr.u.i32;
+                        aux_stack_top =
+                            (uint64)(uint32)global->init_expr.u.unary.v.i32;
                         module->aux_stack_top_global_index =
                             module->import_global_count + global_index;
                         module->aux_stack_bottom = aux_stack_top;
@@ -3448,8 +3624,14 @@ wasm_loader_unload(WASMModule *module)
     if (module->memories)
         wasm_runtime_free(module->memories);
 
-    if (module->globals)
+    if (module->globals) {
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        for (i = 0; i < module->global_count; i++) {
+            destroy_init_expr(&module->globals[i].init_expr);
+        }
+#endif
         wasm_runtime_free(module->globals);
+    }
 
     if (module->exports)
         wasm_runtime_free(module->exports);
@@ -3458,6 +3640,9 @@ wasm_loader_unload(WASMModule *module)
         for (i = 0; i < module->table_seg_count; i++) {
             if (module->table_segments[i].init_values)
                 wasm_runtime_free(module->table_segments[i].init_values);
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            destroy_init_expr(&module->table_segments[i].base_offset);
+#endif
         }
         wasm_runtime_free(module->table_segments);
     }
@@ -3467,6 +3652,9 @@ wasm_loader_unload(WASMModule *module)
             if (module->data_segments[i]) {
                 if (module->data_segments[i]->is_data_cloned)
                     wasm_runtime_free(module->data_segments[i]->data);
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+                destroy_init_expr(&module->data_segments[i]->base_offset);
+#endif
                 wasm_runtime_free(module->data_segments[i]);
             }
         }
@@ -7320,7 +7508,8 @@ re_scan:
                                 == VALUE_TYPE_FUNCREF
                             && module->globals[i].init_expr.init_expr_type
                                    == INIT_EXPR_TYPE_FUNCREF_CONST
-                            && module->globals[i].init_expr.u.u32 == func_idx) {
+                            && module->globals[i].init_expr.u.unary.v.ref_index
+                                   == func_idx) {
                             func_declared = true;
                             break;
                         }
@@ -7334,7 +7523,8 @@ re_scan:
                              i++, table_seg++) {
                             if (table_seg->elem_type == VALUE_TYPE_FUNCREF) {
                                 for (j = 0; j < table_seg->value_count; j++) {
-                                    if (table_seg->init_values[j].u.ref_index
+                                    if (table_seg->init_values[j]
+                                            .u.unary.v.ref_index
                                         == func_idx) {
                                         func_declared = true;
                                         break;

core/iwasm/interpreter/wasm_opcode.h

@@ -278,8 +278,7 @@ typedef enum WASMOpcode {
     DEBUG_OP_BREAK = 0xdc, /* debug break point */
 #endif
 
-#if WASM_ENABLE_JIT != 0 || WASM_ENABLE_FAST_INTERP != 0 \
+#if WASM_ENABLE_SIMDE != 0
-    || WASM_ENABLE_WAMR_COMPILER != 0 && WASM_ENABLE_SIMD != 0
     EXT_OP_SET_LOCAL_FAST_V128 = 0xdd,
     EXT_OP_TEE_LOCAL_FAST_V128 = 0xde,
     EXT_OP_COPY_STACK_TOP_V128 = 0xdf,
@@ -798,7 +797,7 @@ typedef enum WASMAtomicEXTOpcode {
 #define SET_GOTO_TABLE_SIMD_PREFIX_ELEM()
 #endif
 
-#if WASM_ENABLE_SIMD != 0 && WASM_ENABLE_SIMDE != 0
+#if WASM_ENABLE_SIMDE != 0
 #define DEF_EXT_V128_HANDLE()                                       \
     SET_GOTO_TABLE_ELEM(EXT_OP_SET_LOCAL_FAST_V128),     /* 0xdd */ \
         SET_GOTO_TABLE_ELEM(EXT_OP_TEE_LOCAL_FAST_V128), /* 0xde */ \

core/iwasm/interpreter/wasm_runtime.c

@@ -1165,6 +1165,81 @@ instantiate_array_global_recursive(WASMModule *module,
 }
 #endif
 
+static bool
+get_init_value_recursive(WASMModule *module, InitializerExpression *expr,
+                         WASMGlobalInstance *globals, WASMValue *value,
+                         char *error_buf, uint32 error_buf_size)
+{
+    uint8 flag = expr->init_expr_type;
+    switch (flag) {
+        case INIT_EXPR_TYPE_GET_GLOBAL:
+        {
+            if (!check_global_init_expr(module, expr->u.unary.v.global_index,
+                                        error_buf, error_buf_size)) {
+                goto fail;
+            }
+
+            *value = globals[expr->u.unary.v.global_index].initial_value;
+            break;
+        }
+        case INIT_EXPR_TYPE_I32_CONST:
+        case INIT_EXPR_TYPE_I64_CONST:
+        {
+            *value = expr->u.unary.v;
+            break;
+        }
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+        case INIT_EXPR_TYPE_I32_ADD:
+        case INIT_EXPR_TYPE_I32_SUB:
+        case INIT_EXPR_TYPE_I32_MUL:
+        case INIT_EXPR_TYPE_I64_ADD:
+        case INIT_EXPR_TYPE_I64_SUB:
+        case INIT_EXPR_TYPE_I64_MUL:
+        {
+            WASMValue l_value, r_value;
+            if (!expr->u.binary.l_expr || !expr->u.binary.r_expr) {
+                goto fail;
+            }
+            if (!get_init_value_recursive(module, expr->u.binary.l_expr,
+                                          globals, &l_value, error_buf,
+                                          error_buf_size)) {
+                goto fail;
+            }
+            if (!get_init_value_recursive(module, expr->u.binary.r_expr,
+                                          globals, &r_value, error_buf,
+                                          error_buf_size)) {
+                goto fail;
+            }
+
+            if (flag == INIT_EXPR_TYPE_I32_ADD) {
+                value->i32 = l_value.i32 + r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I32_SUB) {
+                value->i32 = l_value.i32 - r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I32_MUL) {
+                value->i32 = l_value.i32 * r_value.i32;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_ADD) {
+                value->i64 = l_value.i64 + r_value.i64;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_SUB) {
+                value->i64 = l_value.i64 - r_value.i64;
+            }
+            else if (flag == INIT_EXPR_TYPE_I64_MUL) {
+                value->i64 = l_value.i64 * r_value.i64;
+            }
+            break;
+        }
+#endif /* end of WASM_ENABLE_EXTENDED_CONST_EXPR != 0 */
+        default:
+            goto fail;
+    }
+    return true;
+fail:
+    return false;
+}
+
 /**
  * Instantiate globals in a module.
  */
@@ -1209,7 +1284,7 @@ globals_instantiate(WASMModule *module, WASMModuleInstance *module_inst,
             /* The linked global instance has been initialized, we
                just need to copy the value. */
             global->initial_value =
-                global_import->import_global_linked->init_expr.u;
+                global_import->import_global_linked->init_expr.u.unary.v;
         }
         else
 #endif
@@ -1245,17 +1320,23 @@ globals_instantiate(WASMModule *module, WASMModuleInstance *module_inst,
 #endif
 
         switch (flag) {
+            case INIT_EXPR_TYPE_I32_CONST:
+            case INIT_EXPR_TYPE_I64_CONST:
             case INIT_EXPR_TYPE_GET_GLOBAL:
+#if WASM_ENABLE_EXTENDED_CONST_EXPR != 0
+            case INIT_EXPR_TYPE_I32_ADD:
+            case INIT_EXPR_TYPE_I32_SUB:
+            case INIT_EXPR_TYPE_I32_MUL:
+            case INIT_EXPR_TYPE_I64_ADD:
+            case INIT_EXPR_TYPE_I64_SUB:
+            case INIT_EXPR_TYPE_I64_MUL:
+#endif
             {
-                if (!check_global_init_expr(module, init_expr->u.global_index,
+                if (!get_init_value_recursive(module, init_expr, globals,
-                                            error_buf, error_buf_size)) {
+                                              &global->initial_value, error_buf,
+                                              error_buf_size)) {
                     goto fail;
                 }
-
-                bh_memcpy_s(
-                    &(global->initial_value), sizeof(WASMValue),
-                    &(globals[init_expr->u.global_index].initial_value),
-                    sizeof(globals[init_expr->u.global_index].initial_value));
                 break;
             }
 #if WASM_ENABLE_GC != 0
@@ -1267,11 +1348,12 @@ globals_instantiate(WASMModule *module, WASMModuleInstance *module_inst,
                 uint32 type_idx;
 
                 if (flag == INIT_EXPR_TYPE_STRUCT_NEW) {
-                    init_values = (WASMStructNewInitValues *)init_expr->u.data;
+                    init_values =
+                        (WASMStructNewInitValues *)init_expr->u.unary.v.data;
                     type_idx = init_values->type_idx;
                 }
                 else {
-                    type_idx = init_expr->u.type_index;
+                    type_idx = init_expr->u.unary.v.type_index;
                 }
 
                 struct_obj = instantiate_struct_global_recursive(
@@ -1294,12 +1376,14 @@ globals_instantiate(WASMModule *module, WASMModuleInstance *module_inst,
                 uint32 type_idx, len;
 
                 if (flag == INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT) {
-                    type_idx = init_expr->u.array_new_default.type_index;
+                    type_idx =
-                    len = init_expr->u.array_new_default.length;
+                        init_expr->u.unary.v.array_new_default.type_index;
+                    len = init_expr->u.unary.v.array_new_default.length;
                     array_init_value = &empty_value;
                 }
                 else {
-                    init_values = (WASMArrayNewInitValues *)init_expr->u.data;
+                    init_values =
+                        (WASMArrayNewInitValues *)init_expr->u.unary.v.data;
                     type_idx = init_values->type_idx;
                     len = init_values->length;
 
@@ -1318,13 +1402,12 @@ globals_instantiate(WASMModule *module, WASMModuleInstance *module_inst,
             case INIT_EXPR_TYPE_I31_NEW:
             {
                 global->initial_value.gc_obj =
-                    (wasm_obj_t)wasm_i31_obj_new(init_expr->u.i32);
+                    (wasm_obj_t)wasm_i31_obj_new(init_expr->u.unary.v.i32);
                 break;
             }
 #endif /* end of WASM_ENABLE_GC != 0 */
             default:
-                bh_memcpy_s(&(global->initial_value), sizeof(WASMValue),
+                global->initial_value = init_expr->u.unary.v;
-                            &(init_expr->u), sizeof(init_expr->u));
                 break;
         }
 
@@ -2668,7 +2751,7 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
                         }
                         STORE_PTR((void **)global_data, func_obj);
                         global_data += sizeof(void *);
-                        /* Also update the inital_value since other globals may
+                        /* Also update the initial_value since other globals may
                          * refer to this */
                         global->initial_value.gc_obj = (wasm_obj_t)func_obj;
                         break;
@@ -2698,6 +2781,7 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
         uint8 *memory_data = NULL;
         uint64 memory_size = 0;
         WASMDataSeg *data_seg = module->data_segments[i];
+        WASMValue offset_value;
 
 #if WASM_ENABLE_BULK_MEMORY != 0
         if (data_seg->is_passive)
@@ -2717,54 +2801,37 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
             (uint64)memory->num_bytes_per_page * memory->cur_page_count;
         bh_assert(memory_data || memory_size == 0);
 
-        bh_assert(data_seg->base_offset.init_expr_type
+        uint8 offset_flag = data_seg->base_offset.init_expr_type;
-                      == INIT_EXPR_TYPE_GET_GLOBAL
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                  || data_seg->base_offset.init_expr_type
+                  || (memory->is_memory64 ? is_valid_i64_offset(offset_flag)
-                         == (memory->is_memory64 ? INIT_EXPR_TYPE_I64_CONST
+                                          : is_valid_i32_offset(offset_flag)));
-                                                 : INIT_EXPR_TYPE_I32_CONST));
 
-        if (data_seg->base_offset.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL) {
+        if (!get_init_value_recursive(module, &data_seg->base_offset, globals,
-            if (!check_global_init_expr(module,
+                                      &offset_value, error_buf,
-                                        data_seg->base_offset.u.global_index,
+                                      error_buf_size)) {
-                                        error_buf, error_buf_size)) {
             goto fail;
         }
 
+        if (offset_flag == INIT_EXPR_TYPE_GET_GLOBAL) {
             if (!globals
-                || globals[data_seg->base_offset.u.global_index].type
+                || globals[data_seg->base_offset.u.unary.v.global_index].type
                        != (memory->is_memory64 ? VALUE_TYPE_I64
                                                : VALUE_TYPE_I32)) {
                 set_error_buf(error_buf, error_buf_size,
                               "data segment does not fit");
                 goto fail;
             }
+        }
 
 #if WASM_ENABLE_MEMORY64 != 0
         if (memory->is_memory64) {
-                base_offset =
+            base_offset = (uint64)offset_value.i64;
-                    (uint64)globals[data_seg->base_offset.u.global_index]
-                        .initial_value.i64;
         }
         else
 #endif
         {
-                base_offset =
+            base_offset = (uint32)offset_value.i32;
-                    (uint32)globals[data_seg->base_offset.u.global_index]
-                        .initial_value.i32;
         }
-        }
-        else {
-#if WASM_ENABLE_MEMORY64 != 0
-            if (memory->is_memory64) {
-                base_offset = (uint64)data_seg->base_offset.u.i64;
-            }
-            else
-#endif
-            {
-                base_offset = (uint32)data_seg->base_offset.u.i32;
-            }
-        }
-
         /* check offset */
         if (base_offset > memory_size) {
 #if WASM_ENABLE_MEMORY64 != 0
@@ -2818,6 +2885,7 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
 #else
     module_inst->e->shared_heap_start_off.u32[0] = UINT32_MAX;
 #endif
+    module_inst->e->shared_heap = NULL;
 #endif
 
 #if WASM_ENABLE_GC != 0
@@ -2841,36 +2909,39 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
             || table->init_expr.init_expr_type == INIT_EXPR_TYPE_REFNULL_CONST);
 
         if (table->init_expr.init_expr_type == INIT_EXPR_TYPE_GET_GLOBAL) {
-            if (!check_global_init_expr(module, table->init_expr.u.global_index,
+            if (!check_global_init_expr(module,
+                                        table->init_expr.u.unary.v.global_index,
                                         error_buf, error_buf_size)) {
                 goto fail;
             }
 
-            table->init_expr.u.gc_obj =
+            table->init_expr.u.unary.v.gc_obj =
-                globals[table->init_expr.u.global_index].initial_value.gc_obj;
+                globals[table->init_expr.u.unary.v.global_index]
+                    .initial_value.gc_obj;
         }
         else if (table->init_expr.init_expr_type
                  == INIT_EXPR_TYPE_FUNCREF_CONST) {
-            uint32 func_idx = table->init_expr.u.ref_index;
+            uint32 func_idx = table->init_expr.u.unary.v.ref_index;
             if (func_idx != UINT32_MAX) {
-                if (!(table->init_expr.u.gc_obj =
+                if (!(table->init_expr.u.unary.v.gc_obj =
                           wasm_create_func_obj(module_inst, func_idx, false,
                                                error_buf, error_buf_size)))
                     goto fail;
             }
             else {
-                table->init_expr.u.gc_obj = NULL_REF;
+                table->init_expr.u.unary.v.gc_obj = NULL_REF;
             }
         }
         else if (table->init_expr.init_expr_type
                  == INIT_EXPR_TYPE_REFNULL_CONST) {
-            table->init_expr.u.gc_obj = NULL_REF;
+            table->init_expr.u.unary.v.gc_obj = NULL_REF;
         }
 
         LOG_DEBUG("Init table [%d] elements from [%d] to [%d] as: %p", i, 0,
-                  table_inst->cur_size, (void *)table->init_expr.u.gc_obj);
+                  table_inst->cur_size,
+                  (void *)table->init_expr.u.unary.v.gc_obj);
         for (j = 0; j < table_inst->cur_size; j++) {
-            *(table_data + j) = table->init_expr.u.gc_obj;
+            *(table_data + j) = table->init_expr.u.unary.v.gc_obj;
         }
     }
 #endif /* end of WASM_ENABLE_GC != 0 */
@@ -2882,6 +2953,7 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
         /* has check it in loader */
         WASMTableInstance *table = module_inst->tables[table_seg->table_index];
         table_elem_type_t *table_data;
+        WASMValue offset_value;
         uint32 j;
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
         uint8 tbl_elem_type;
@@ -2940,48 +3012,37 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
             continue;
 #endif
 
+        uint8 offset_flag = table_seg->base_offset.init_expr_type;
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
-        bh_assert(table_seg->base_offset.init_expr_type
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                      == INIT_EXPR_TYPE_I32_CONST
+                  || offset_flag == INIT_EXPR_TYPE_FUNCREF_CONST
-                  || table_seg->base_offset.init_expr_type
+                  || offset_flag == INIT_EXPR_TYPE_REFNULL_CONST
-                         == INIT_EXPR_TYPE_GET_GLOBAL
+                  || is_valid_i32_offset(offset_flag));
-                  || table_seg->base_offset.init_expr_type
-                         == INIT_EXPR_TYPE_FUNCREF_CONST
-                  || table_seg->base_offset.init_expr_type
-                         == INIT_EXPR_TYPE_REFNULL_CONST);
 #else
-        bh_assert(table_seg->base_offset.init_expr_type
+        bh_assert(offset_flag == INIT_EXPR_TYPE_GET_GLOBAL
-                      == INIT_EXPR_TYPE_I32_CONST
+                  || is_valid_i32_offset(offset_flag));
-                  || table_seg->base_offset.init_expr_type
-                         == INIT_EXPR_TYPE_GET_GLOBAL);
 #endif
 
-        /* init vec(funcidx) or vec(expr) */
+        if (!get_init_value_recursive(module, &table_seg->base_offset, globals,
-        if (table_seg->base_offset.init_expr_type
+                                      &offset_value, error_buf,
-            == INIT_EXPR_TYPE_GET_GLOBAL) {
+                                      error_buf_size)) {
-            if (!check_global_init_expr(module,
-                                        table_seg->base_offset.u.global_index,
-                                        error_buf, error_buf_size)) {
             goto fail;
         }
 
+        if (offset_flag == INIT_EXPR_TYPE_GET_GLOBAL) {
             if (!globals
-                || globals[table_seg->base_offset.u.global_index].type
+                || globals[table_seg->base_offset.u.unary.v.global_index].type
                        != VALUE_TYPE_I32) {
                 set_error_buf(error_buf, error_buf_size,
                               "type mismatch: elements segment does not fit");
                 goto fail;
             }
-
-            table_seg->base_offset.u.i32 =
-                globals[table_seg->base_offset.u.global_index]
-                    .initial_value.i32;
         }
 
         /* check offset since length might negative */
-        if ((uint32)table_seg->base_offset.u.i32 > table->cur_size) {
+        if ((uint32)offset_value.i32 > table->cur_size) {
-            LOG_DEBUG("base_offset(%d) > table->cur_size(%d)",
+            LOG_DEBUG("base_offset(%d) > table->cur_size(%d)", offset_value.i32,
-                      table_seg->base_offset.u.i32, table->cur_size);
+                      table->cur_size);
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
             set_error_buf(error_buf, error_buf_size,
                           "out of bounds table access");
@@ -2994,9 +3055,9 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
 
         /* check offset + length(could be zero) */
         length = table_seg->value_count;
-        if ((uint32)table_seg->base_offset.u.i32 + length > table->cur_size) {
+        if ((uint32)offset_value.i32 + length > table->cur_size) {
             LOG_DEBUG("base_offset(%d) + length(%d)> table->cur_size(%d)",
-                      table_seg->base_offset.u.i32, length, table->cur_size);
+                      offset_value.i32, length, table->cur_size);
 #if WASM_ENABLE_REF_TYPES != 0 || WASM_ENABLE_GC != 0
             set_error_buf(error_buf, error_buf_size,
                           "out of bounds table access");
@@ -3026,10 +3087,10 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
                 case INIT_EXPR_TYPE_FUNCREF_CONST:
                 {
 #if WASM_ENABLE_GC == 0
-                    ref = (void *)(uintptr_t)init_expr->u.ref_index;
+                    ref = (void *)(uintptr_t)init_expr->u.unary.v.ref_index;
 #else
                     WASMFuncObjectRef func_obj;
-                    uint32 func_idx = init_expr->u.ref_index;
+                    uint32 func_idx = init_expr->u.unary.v.ref_index;
                     /* UINT32_MAX indicates that it is a null reference */
                     if (func_idx != UINT32_MAX) {
                         if (!(func_obj = wasm_create_func_obj(
@@ -3048,14 +3109,14 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
 #if WASM_ENABLE_GC != 0
                 case INIT_EXPR_TYPE_GET_GLOBAL:
                 {
-                    if (!check_global_init_expr(module,
+                    if (!check_global_init_expr(
-                                                init_expr->u.global_index,
+                            module, init_expr->u.unary.v.global_index,
                             error_buf, error_buf_size)) {
                         goto fail;
                     }
 
-                    ref =
+                    ref = globals[init_expr->u.unary.v.global_index]
-                        globals[init_expr->u.global_index].initial_value.gc_obj;
+                              .initial_value.gc_obj;
                     break;
                 }
                 case INIT_EXPR_TYPE_STRUCT_NEW:
@@ -3068,12 +3129,12 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
                     uint32 type_idx;
 
                     if (flag == INIT_EXPR_TYPE_STRUCT_NEW) {
-                        init_values =
+                        init_values = (WASMStructNewInitValues *)
-                            (WASMStructNewInitValues *)init_expr->u.data;
+                                          init_expr->u.unary.v.data;
                         type_idx = init_values->type_idx;
                     }
                     else {
-                        type_idx = init_expr->u.type_index;
+                        type_idx = init_expr->u.unary.v.type_index;
                     }
 
                     struct_type = (WASMStructType *)module->types[type_idx];
@@ -3124,13 +3185,14 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
                     uint32 type_idx, len;
 
                     if (flag == INIT_EXPR_TYPE_ARRAY_NEW_DEFAULT) {
-                        type_idx = init_expr->u.array_new_default.type_index;
+                        type_idx =
-                        len = init_expr->u.array_new_default.length;
+                            init_expr->u.unary.v.array_new_default.type_index;
+                        len = init_expr->u.unary.v.array_new_default.length;
                         arr_init_val = &empty_val;
                     }
                     else {
                         init_values =
-                            (WASMArrayNewInitValues *)init_expr->u.data;
+                            (WASMArrayNewInitValues *)init_expr->u.unary.v.data;
                         type_idx = init_values->type_idx;
                         len = init_values->length;
 
@@ -3176,14 +3238,14 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
                 }
                 case INIT_EXPR_TYPE_I31_NEW:
                 {
-                    ref = (wasm_obj_t)wasm_i31_obj_new(init_expr->u.i32);
+                    ref =
+                        (wasm_obj_t)wasm_i31_obj_new(init_expr->u.unary.v.i32);
                     break;
                 }
 #endif /* end of WASM_ENABLE_GC != 0 */
             }
 
-            *(table_data + table_seg->base_offset.u.i32 + j) =
+            *(table_data + offset_value.i32 + j) = (table_elem_type_t)ref;
-                (table_elem_type_t)ref;
         }
     }
 
@@ -4161,7 +4223,7 @@ wasm_get_module_inst_mem_consumption(const WASMModuleInstance *module_inst,
         sizeof(WASMMemoryInstance *) * module_inst->memory_count;
     for (i = 0; i < module_inst->memory_count; i++) {
         WASMMemoryInstance *memory = module_inst->memories[i];
-        size = memory->num_bytes_per_page * memory->cur_page_count;
+        size = (uint64)memory->num_bytes_per_page * memory->cur_page_count;
         mem_conspn->memories_size += size;
         mem_conspn->app_heap_size += memory->heap_data_end - memory->heap_data;
         /* size of app heap structure */
@@ -4195,9 +4257,9 @@ wasm_get_module_inst_mem_consumption(const WASMModuleInstance *module_inst,
 #endif /* end of (WASM_ENABLE_MEMORY_PROFILING != 0) \
                  || (WASM_ENABLE_MEMORY_TRACING != 0) */
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 uint32
-wasm_interp_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+wasm_interp_copy_callstack(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                            uint32 length, uint32 skip_n, char *error_buf,
                            uint32_t error_buf_size)
 {
@@ -4242,7 +4304,7 @@ wasm_interp_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
     }
     return count >= skip_n ? count - skip_n : 0;
 }
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 #if WASM_ENABLE_DUMP_CALL_STACK != 0
 bool
@@ -4705,10 +4767,10 @@ llvm_jit_table_init(WASMModuleInstance *module_inst, uint32 tbl_idx,
     for (i = 0; i < length; i++) {
 #if WASM_ENABLE_GC != 0
         /* UINT32_MAX indicates that it is a null ref */
-        if (init_values[i].u.ref_index != UINT32_MAX) {
+        if (init_values[i].u.unary.v.ref_index != UINT32_MAX) {
-            if (!(func_obj = wasm_create_func_obj(module_inst,
+            if (!(func_obj = wasm_create_func_obj(
-                                                  init_values[i].u.ref_index,
+                      module_inst, init_values[i].u.unary.v.ref_index, true,
-                                                  true, NULL, 0))) {
+                      NULL, 0))) {
                 wasm_set_exception(module_inst, "null function reference");
                 return;
             }
@@ -4718,7 +4780,7 @@ llvm_jit_table_init(WASMModuleInstance *module_inst, uint32 tbl_idx,
             table_elems[i] = NULL_REF;
         }
 #else
-        table_elems[i] = init_values[i].u.ref_index;
+        table_elems[i] = init_values[i].u.unary.v.ref_index;
 #endif
     }
 }

core/iwasm/interpreter/wasm_runtime.h

@@ -93,12 +93,21 @@ typedef union {
 } MemBound;
 
 typedef struct WASMSharedHeap {
-    struct WASMSharedHeap *next;
+    /* The global shared heap list maintained in runtime, used for runtime
-    void *heap_handle;
+     * destroy */
-    uint8 *base_addr;
+    DefPointer(struct WASMSharedHeap *, next);
+    /* The logical shared heap chain the shared heap in */
+    DefPointer(struct WASMSharedHeap *, chain_next);
+    /* Will be null if shared heap is created from pre allocated memory chunk
+     * and don't need to dynamic malloc and free */
+    DefPointer(void *, heap_handle);
+    DefPointer(uint8 *, base_addr);
     uint64 size;
     uint64 start_off_mem64;
     uint64 start_off_mem32;
+    /* The number of wasm apps it attached to, for a shared heap chain, only the
+     * list head need to maintain the valid attached_count */
+    uint8 attached_count;
 } WASMSharedHeap;
 
 struct WASMMemoryInstance {
@@ -364,8 +373,6 @@ typedef struct WASMModuleInstanceExtra {
 #endif
 
 #if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap;
-#if WASM_ENABLE_JIT != 0
     /*
      * Adjusted shared heap based addr to simple the calculation
      * in the aot code. The value is:
@@ -373,7 +380,8 @@ typedef struct WASMModuleInstanceExtra {
      */
     uint8 *shared_heap_base_addr_adj;
     MemBound shared_heap_start_off;
-#endif
+    MemBound shared_heap_end_off;
+    WASMSharedHeap *shared_heap;
 #endif
 
 #if WASM_ENABLE_DEBUG_INTERP != 0                         \
@@ -731,12 +739,12 @@ wasm_get_table_inst(const WASMModuleInstance *module_inst, uint32 tbl_idx)
 
 #if WASM_ENABLE_DUMP_CALL_STACK != 0
 
-#if WAMR_ENABLE_COPY_CALLSTACK != 0
+#if WASM_ENABLE_COPY_CALL_STACK != 0
 uint32
-wasm_interp_copy_callstack(WASMExecEnv *exec_env, wasm_frame_t *buffer,
+wasm_interp_copy_callstack(WASMExecEnv *exec_env, WASMCApiFrame *buffer,
                            uint32 length, uint32 skip_n, char *error_buf,
                            uint32_t error_buf_size);
-#endif // WAMR_ENABLE_COPY_CALLSTACK
+#endif // WASM_ENABLE_COPY_CALL_STACK
 
 bool
 wasm_interp_create_call_stack(struct WASMExecEnv *exec_env);

core/iwasm/libraries/lib-socket/inc/wasi_socket_ext.h

@@ -135,6 +135,28 @@ typedef struct __wasi_addr_info_hints_t {
 #define IPV6_LEAVE_GROUP 21
 #define IPV6_V6ONLY 26
 
+/* getaddrinfo error codes.
+ *
+ * we use values compatible with wasi-libc/musl netdb.h.
+ * https://github.com/WebAssembly/wasi-libc/blob/4ea6fdfa288e15a57c02fe31dda78e5ddc87c3c7/libc-top-half/musl/include/netdb.h#L43-L53
+ *
+ * for now, non-posix error codes are excluded:
+ * EAI_PROTOCOL and EAI_BADHINTS (BSDs)
+ * EAI_ADDRFAMILY, EAI_NODATA
+ * https://github.com/WebAssembly/wasi-libc/blob/4ea6fdfa288e15a57c02fe31dda78e5ddc87c3c7/libc-top-half/musl/include/netdb.h#L145-L152
+ */
+
+#define EAI_AGAIN -3
+#define EAI_BADFLAGS -1
+#define EAI_FAIL -4
+#define EAI_FAMILY -6
+#define EAI_MEMORY -10
+#define EAI_NONAME -2
+#define EAI_OVERFLOW -12
+#define EAI_SERVICE -8
+#define EAI_SOCKTYPE -7
+#define EAI_SYSTEM -11
+
 struct addrinfo {
     int ai_flags;             /* Input flags.  */
     int ai_family;            /* Protocol family for socket.  */
@@ -197,6 +219,9 @@ getaddrinfo(const char *node, const char *service, const struct addrinfo *hints,
 
 void
 freeaddrinfo(struct addrinfo *res);
+
+const char *
+gai_strerror(int code);
 #endif
 
 /**

core/iwasm/libraries/lib-socket/src/wasi/wasi_socket_ext.c

@@ -38,6 +38,13 @@ __errno_location(void);
         return -1;                       \
     }
 
+/* REVISIT: in many cases, EAI_SYSTEM may not be an ideal error code */
+#define GAI_HANDLE_ERROR(error)          \
+    if (error != __WASI_ERRNO_SUCCESS) { \
+        errno = error;                   \
+        return EAI_SYSTEM;               \
+    }
+
 static void
 ipv4_addr_to_wasi_ip4_addr(uint32_t addr_num, __wasi_addr_ip4_t *out)
 {
@@ -191,7 +198,7 @@ bind(int sockfd, const struct sockaddr *addr, socklen_t addrlen)
     error = __wasi_sock_bind(sockfd, &wasi_addr);
     HANDLE_ERROR(error)
 
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 int
@@ -212,7 +219,7 @@ connect(int sockfd, const struct sockaddr *addr, socklen_t addrlen)
     error = __wasi_sock_connect(sockfd, &wasi_addr);
     HANDLE_ERROR(error)
 
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 int
@@ -220,7 +227,7 @@ listen(int sockfd, int backlog)
 {
     __wasi_errno_t error = __wasi_sock_listen(sockfd, backlog);
     HANDLE_ERROR(error)
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 ssize_t
@@ -375,7 +382,7 @@ socket(int domain, int type, int protocol)
         af = INET6;
     }
     else {
-        return __WASI_ERRNO_NOPROTOOPT;
+        HANDLE_ERROR(__WASI_ERRNO_NOPROTOOPT)
     }
 
     if (SOCK_DGRAM == type) {
@@ -385,7 +392,7 @@ socket(int domain, int type, int protocol)
         socktype = SOCKET_STREAM;
     }
     else {
-        return __WASI_ERRNO_NOPROTOOPT;
+        HANDLE_ERROR(__WASI_ERRNO_NOPROTOOPT)
     }
 
     error = __wasi_sock_open(poolfd, af, socktype, &sockfd);
@@ -408,7 +415,7 @@ getsockname(int sockfd, struct sockaddr *addr, socklen_t *addrlen)
     error = wasi_addr_to_sockaddr(&wasi_addr, addr, addrlen);
     HANDLE_ERROR(error)
 
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 int
@@ -425,7 +432,7 @@ getpeername(int sockfd, struct sockaddr *addr, socklen_t *addrlen)
     error = wasi_addr_to_sockaddr(&wasi_addr, addr, addrlen);
     HANDLE_ERROR(error)
 
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 struct aibuf {
@@ -518,7 +525,7 @@ getaddrinfo(const char *node, const char *service, const struct addrinfo *hints,
     struct aibuf *aibuf_res;
 
     error = addrinfo_hints_to_wasi_hints(hints, &wasi_hints);
-    HANDLE_ERROR(error)
+    GAI_HANDLE_ERROR(error)
 
     do {
         if (addr_info)
@@ -529,7 +536,7 @@ getaddrinfo(const char *node, const char *service, const struct addrinfo *hints,
                                                  * sizeof(__wasi_addr_info_t));
 
         if (!addr_info) {
-            HANDLE_ERROR(__WASI_ERRNO_NOMEM)
+            return EAI_MEMORY;
         }
 
         error = __wasi_sock_addr_resolve(node, service == NULL ? "" : service,
@@ -537,29 +544,25 @@ getaddrinfo(const char *node, const char *service, const struct addrinfo *hints,
                                          &max_info_size);
         if (error != __WASI_ERRNO_SUCCESS) {
             free(addr_info);
-            HANDLE_ERROR(error);
+            GAI_HANDLE_ERROR(error);
         }
     } while (max_info_size > addr_info_size);
 
+    addr_info_size = max_info_size;
     if (addr_info_size == 0) {
         free(addr_info);
-        *res = NULL;
+        return EAI_NONAME;
-        return __WASI_ERRNO_SUCCESS;
     }
 
     aibuf_res =
         (struct aibuf *)calloc(1, addr_info_size * sizeof(struct aibuf));
     if (!aibuf_res) {
         free(addr_info);
-        HANDLE_ERROR(__WASI_ERRNO_NOMEM)
+        return EAI_MEMORY;
     }
 
     *res = &aibuf_res[0].ai;
 
-    if (addr_info_size) {
-        addr_info_size = max_info_size;
-    }
-
     for (i = 0; i < addr_info_size; i++) {
         struct addrinfo *ai = &aibuf_res[i].ai;
         ai->ai_addr = (struct sockaddr *)&aibuf_res[i].sa;
@@ -568,14 +571,14 @@ getaddrinfo(const char *node, const char *service, const struct addrinfo *hints,
         if (error != __WASI_ERRNO_SUCCESS) {
             free(addr_info);
             free(aibuf_res);
-            HANDLE_ERROR(error)
+            GAI_HANDLE_ERROR(error)
         }
         ai->ai_next = i == addr_info_size - 1 ? NULL : &aibuf_res[i + 1].ai;
     }
 
     free(addr_info);
 
-    return __WASI_ERRNO_SUCCESS;
+    return 0;
 }
 
 void
@@ -587,6 +590,28 @@ freeaddrinfo(struct addrinfo *res)
     free(res);
 }
 
+const char *
+gai_strerror(int code)
+{
+    switch (code) {
+#define ERR(a) \
+    case a:    \
+        return #a
+        ERR(EAI_AGAIN);
+        ERR(EAI_BADFLAGS);
+        ERR(EAI_FAIL);
+        ERR(EAI_FAMILY);
+        ERR(EAI_MEMORY);
+        ERR(EAI_NONAME);
+        ERR(EAI_OVERFLOW);
+        ERR(EAI_SERVICE);
+        ERR(EAI_SOCKTYPE);
+        ERR(EAI_SYSTEM);
+#undef ERR
+    }
+    return "Unknown error";
+}
+
 static struct timeval
 time_us_to_timeval(uint64_t time_us)
 {
@@ -610,6 +635,7 @@ get_sol_socket_option(int sockfd, int optname, void *__restrict optval,
     uint64_t timeout_us;
     bool is_linger_enabled;
     int linger_s;
+    __wasi_fdstat_t sb;
 
     switch (optname) {
         case SO_RCVTIMEO:
@@ -617,38 +643,38 @@ get_sol_socket_option(int sockfd, int optname, void *__restrict optval,
             error = __wasi_sock_get_recv_timeout(sockfd, &timeout_us);
             HANDLE_ERROR(error);
             *(struct timeval *)optval = time_us_to_timeval(timeout_us);
-            return error;
+            return 0;
         case SO_SNDTIMEO:
             assert(*optlen == sizeof(struct timeval));
             error = __wasi_sock_get_send_timeout(sockfd, &timeout_us);
             HANDLE_ERROR(error);
             *(struct timeval *)optval = time_us_to_timeval(timeout_us);
-            return error;
+            return 0;
         case SO_SNDBUF:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_send_buf_size(sockfd, (size_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case SO_RCVBUF:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_recv_buf_size(sockfd, (size_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case SO_KEEPALIVE:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_keep_alive(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case SO_REUSEADDR:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_reuse_addr(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case SO_REUSEPORT:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_reuse_port(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case SO_LINGER:
             assert(*optlen == sizeof(struct linger));
             error =
@@ -656,16 +682,32 @@ get_sol_socket_option(int sockfd, int optname, void *__restrict optval,
             HANDLE_ERROR(error);
             ((struct linger *)optval)->l_onoff = (int)is_linger_enabled;
             ((struct linger *)optval)->l_linger = linger_s;
-            return error;
+            return 0;
         case SO_BROADCAST:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_broadcast(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
+        case SO_TYPE:
+            assert(*optlen == sizeof(int));
+            error = __wasi_fd_fdstat_get(sockfd, &sb);
+            HANDLE_ERROR(error);
+            switch (sb.fs_filetype) {
+                case __WASI_FILETYPE_SOCKET_DGRAM:
+                    *(int *)optval = SOCK_DGRAM;
+                    break;
+                case __WASI_FILETYPE_SOCKET_STREAM:
+                    *(int *)optval = SOCK_STREAM;
+                    break;
+                default:
+                    errno = __WASI_ERRNO_NOTSOCK;
+                    return -1;
+            }
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -679,32 +721,32 @@ get_ipproto_tcp_option(int sockfd, int optname, void *__restrict optval,
             assert(*optlen == sizeof(uint32_t));
             error = __wasi_sock_get_tcp_keep_idle(sockfd, (uint32_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_KEEPINTVL:
             assert(*optlen == sizeof(uint32_t));
             error = __wasi_sock_get_tcp_keep_intvl(sockfd, (uint32_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_FASTOPEN_CONNECT:
             assert(*optlen == sizeof(int));
             error =
                 __wasi_sock_get_tcp_fastopen_connect(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_NODELAY:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_tcp_no_delay(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_QUICKACK:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_tcp_quick_ack(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -720,21 +762,21 @@ get_ipproto_ip_option(int sockfd, int optname, void *__restrict optval,
             error = __wasi_sock_get_ip_multicast_loop(sockfd, false,
                                                       (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_TTL:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_ip_ttl(sockfd, (uint8_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_MULTICAST_TTL:
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_ip_multicast_ttl(sockfd, (uint8_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -749,17 +791,17 @@ get_ipproto_ipv6_option(int sockfd, int optname, void *__restrict optval,
             assert(*optlen == sizeof(int));
             error = __wasi_sock_get_ipv6_only(sockfd, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IPV6_MULTICAST_LOOP:
             assert(*optlen == sizeof(int));
             error =
                 __wasi_sock_get_ip_multicast_loop(sockfd, true, (bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -781,7 +823,7 @@ getsockopt(int sockfd, int level, int optname, void *__restrict optval,
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -799,7 +841,7 @@ set_sol_socket_option(int sockfd, int optname, const void *optval,
             timeout_us = timeval_to_time_us(*(struct timeval *)optval);
             error = __wasi_sock_set_recv_timeout(sockfd, timeout_us);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_SNDTIMEO:
         {
@@ -807,42 +849,42 @@ set_sol_socket_option(int sockfd, int optname, const void *optval,
             timeout_us = timeval_to_time_us(*(struct timeval *)optval);
             error = __wasi_sock_set_send_timeout(sockfd, timeout_us);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_SNDBUF:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_send_buf_size(sockfd, *(size_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_RCVBUF:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_recv_buf_size(sockfd, *(size_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_KEEPALIVE:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_keep_alive(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_REUSEADDR:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_reuse_addr(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_REUSEPORT:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_reuse_port(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_LINGER:
         {
@@ -851,20 +893,20 @@ set_sol_socket_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_linger(sockfd, (bool)linger_opt->l_onoff,
                                            linger_opt->l_linger);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         case SO_BROADCAST:
         {
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_broadcast(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
         default:
         {
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         }
     }
 }
@@ -880,32 +922,32 @@ set_ipproto_tcp_option(int sockfd, int optname, const void *optval,
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_tcp_no_delay(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_KEEPIDLE:
             assert(optlen == sizeof(uint32_t));
             error = __wasi_sock_set_tcp_keep_idle(sockfd, *(uint32_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_KEEPINTVL:
             assert(optlen == sizeof(uint32_t));
             error = __wasi_sock_set_tcp_keep_intvl(sockfd, *(uint32_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_FASTOPEN_CONNECT:
             assert(optlen == sizeof(int));
             error =
                 __wasi_sock_set_tcp_fastopen_connect(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case TCP_QUICKACK:
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_tcp_quick_ack(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -923,7 +965,7 @@ set_ipproto_ip_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_ip_multicast_loop(sockfd, false,
                                                       *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_ADD_MEMBERSHIP:
             assert(optlen == sizeof(struct ip_mreq));
             ip_mreq_opt = (struct ip_mreq *)optval;
@@ -933,7 +975,7 @@ set_ipproto_ip_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_ip_add_membership(
                 sockfd, &imr_multiaddr, ip_mreq_opt->imr_interface.s_addr);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_DROP_MEMBERSHIP:
             assert(optlen == sizeof(struct ip_mreq));
             ip_mreq_opt = (struct ip_mreq *)optval;
@@ -943,22 +985,22 @@ set_ipproto_ip_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_ip_drop_membership(
                 sockfd, &imr_multiaddr, ip_mreq_opt->imr_interface.s_addr);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_TTL:
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_ip_ttl(sockfd, *(uint8_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IP_MULTICAST_TTL:
             assert(optlen == sizeof(int));
             error =
                 __wasi_sock_set_ip_multicast_ttl(sockfd, *(uint8_t *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -975,13 +1017,13 @@ set_ipproto_ipv6_option(int sockfd, int optname, const void *optval,
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_ipv6_only(sockfd, *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IPV6_MULTICAST_LOOP:
             assert(optlen == sizeof(int));
             error = __wasi_sock_set_ip_multicast_loop(sockfd, true,
                                                       *(bool *)optval);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IPV6_JOIN_GROUP:
             assert(optlen == sizeof(struct ipv6_mreq));
             ipv6_mreq_opt = (struct ipv6_mreq *)optval;
@@ -992,7 +1034,7 @@ set_ipproto_ipv6_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_ip_add_membership(
                 sockfd, &imr_multiaddr, ipv6_mreq_opt->ipv6mr_interface);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         case IPV6_LEAVE_GROUP:
             assert(optlen == sizeof(struct ipv6_mreq));
             ipv6_mreq_opt = (struct ipv6_mreq *)optval;
@@ -1003,11 +1045,11 @@ set_ipproto_ipv6_option(int sockfd, int optname, const void *optval,
             error = __wasi_sock_set_ip_drop_membership(
                 sockfd, &imr_multiaddr, ipv6_mreq_opt->ipv6mr_interface);
             HANDLE_ERROR(error);
-            return error;
+            return 0;
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }
 
@@ -1029,6 +1071,6 @@ setsockopt(int sockfd, int level, int optname, const void *optval,
         default:
             error = __WASI_ERRNO_NOTSUP;
             HANDLE_ERROR(error);
-            return error;
+            return 0;
     }
 }

core/iwasm/libraries/libc-wasi/libc_wasi_wrapper.c

@@ -375,6 +375,9 @@ wasi_fd_pread(wasm_exec_env_t exec_env, wasi_fd_t fd, iovec_app_t *iovec_app,
         return (wasi_errno_t)-1;
 
     total_size = sizeof(wasi_iovec_t) * (uint64)iovs_len;
+    if (total_size == 0) {
+        total_size = 1; /* avoid user-triggered 0-sized allocation */
+    }
     if (total_size >= UINT32_MAX
         || !(iovec_begin = wasm_runtime_malloc((uint32)total_size)))
         return (wasi_errno_t)-1;
@@ -430,6 +433,9 @@ wasi_fd_pwrite(wasm_exec_env_t exec_env, wasi_fd_t fd,
         return (wasi_errno_t)-1;
 
     total_size = sizeof(wasi_ciovec_t) * (uint64)iovs_len;
+    if (total_size == 0) {
+        total_size = 1; /* avoid user-triggered 0-sized allocation */
+    }
     if (total_size >= UINT32_MAX
         || !(ciovec_begin = wasm_runtime_malloc((uint32)total_size)))
         return (wasi_errno_t)-1;
@@ -484,6 +490,9 @@ wasi_fd_read(wasm_exec_env_t exec_env, wasi_fd_t fd,
         return (wasi_errno_t)-1;
 
     total_size = sizeof(wasi_iovec_t) * (uint64)iovs_len;
+    if (total_size == 0) {
+        total_size = 1; /* avoid user-triggered 0-sized allocation */
+    }
     if (total_size >= UINT32_MAX
         || !(iovec_begin = wasm_runtime_malloc((uint32)total_size)))
         return (wasi_errno_t)-1;
@@ -654,6 +663,9 @@ wasi_fd_write(wasm_exec_env_t exec_env, wasi_fd_t fd,
         return (wasi_errno_t)-1;
 
     total_size = sizeof(wasi_ciovec_t) * (uint64)iovs_len;
+    if (total_size == 0) {
+        total_size = 1; /* avoid user-triggered 0-sized allocation */
+    }
     if (total_size >= UINT32_MAX
         || !(ciovec_begin = wasm_runtime_malloc((uint32)total_size)))
         return (wasi_errno_t)-1;

core/iwasm/libraries/thread-mgr/thread_manager.c

@@ -301,7 +301,8 @@ wasm_cluster_create(WASMExecEnv *exec_env)
         aux_stack_start -= cluster->stack_size;
 
         for (i = 0; i < cluster_max_thread_num; i++) {
-            cluster->stack_tops[i] = aux_stack_start - cluster->stack_size * i;
+            cluster->stack_tops[i] =
+                aux_stack_start - (uint64)cluster->stack_size * i;
         }
     }
 #endif

core/iwasm/libraries/wasi-nn/include/wasi_nn.h

@@ -21,6 +21,7 @@
 #else
 #define WASI_NN_IMPORT(name) \
     __attribute__((import_module("wasi_nn"), import_name(name)))
+#warning You are using "wasi_nn", which is a legacy WAMR-specific ABI. It's deperecated and will likely be removed in future versions of WAMR. Please use "wasi_ephemeral_nn" instead. (For a WASM module, use the wasi_ephemeral_nn.h header instead. For the runtime configurations, enable WASM_ENABLE_WASI_EPHEMERAL_NN/WAMR_BUILD_WASI_EPHEMERAL_NN.)
 #endif
 
 /**
@@ -108,14 +109,13 @@ WASI_NN_NAME(compute)
 WASI_NN_ERROR_TYPE
 WASI_NN_NAME(get_output)
 (WASI_NN_NAME(graph_execution_context) ctx, uint32_t index,
- WASI_NN_NAME(tensor_data) output_tensor, uint32_t output_tensor_max_size,
+ uint8_t *output_tensor, uint32_t output_tensor_max_size,
  uint32_t *output_tensor_size) WASI_NN_IMPORT("get_output");
 #else
 WASI_NN_ERROR_TYPE
 WASI_NN_NAME(get_output)
-(graph_execution_context ctx, uint32_t index,
+(graph_execution_context ctx, uint32_t index, uint8_t *output_tensor,
- WASI_NN_NAME(tensor_data) output_tensor, uint32_t *output_tensor_size)
+ uint32_t *output_tensor_size) WASI_NN_IMPORT("get_output");
-    WASI_NN_IMPORT("get_output");
 #endif
 
 #endif

core/iwasm/libraries/wasi-nn/include/wasi_nn_types.h

@@ -99,7 +99,14 @@ typedef enum {
 // 4-byte f32 elements would have a data array of length 16). Naturally, this
 // representation requires some knowledge of how to lay out data in
 // memory--e.g., using row-major ordering--and could perhaps be improved.
+#if !defined(__wasm__) || WASM_ENABLE_WASI_EPHEMERAL_NN != 0
+typedef struct {
+    uint8_t *buf;
+    uint32_t size;
+} WASI_NN_NAME(tensor_data);
+#else
 typedef uint8_t *WASI_NN_NAME(tensor_data);
+#endif
 
 // A tensor.
 typedef struct {

core/iwasm/libraries/wasi-nn/src/utils/wasi_nn_app_native.c

@@ -99,7 +99,8 @@ graph_builder_array_app_native(wasm_module_inst_t instance,
 
 static wasi_nn_error
 tensor_data_app_native(wasm_module_inst_t instance, uint32_t total_elements,
-                       tensor_wasm *input_tensor_wasm, tensor_data *data)
+                       tensor_wasm *input_tensor_wasm, void **data,
+                       uint32_t *size)
 {
 #if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
 #define data_size input_tensor_wasm->data_size
@@ -113,8 +114,9 @@ tensor_data_app_native(wasm_module_inst_t instance, uint32_t total_elements,
         NN_ERR_PRINTF("input_tensor_wasm->data_offset is invalid");
         return invalid_argument;
     }
-    *data = (tensor_data)wasm_runtime_addr_app_to_native(
+    *data = wasm_runtime_addr_app_to_native(
         instance, (uint64)input_tensor_wasm->data_offset);
+    *size = data_size;
     return success;
 #undef data_size
 }
@@ -188,16 +190,19 @@ tensor_app_native(wasm_module_inst_t instance, tensor_wasm *input_tensor_wasm,
     NN_DBG_PRINTF("Tensor type: %d", input_tensor_wasm->type);
     NN_DBG_PRINTF("Total number of elements: %d", total_elements);
 
-    tensor_data data = NULL;
+    void *data = NULL;
+    uint32_t datasize;
     if (success
-        != (res = tensor_data_app_native(instance, total_elements,
+        != (res =
-                                         input_tensor_wasm, &data))) {
+                tensor_data_app_native(instance, total_elements,
+                                       input_tensor_wasm, &data, &datasize))) {
         wasm_runtime_free(dimensions);
         return res;
     }
 
     input_tensor->type = input_tensor_wasm->type;
     input_tensor->dimensions = dimensions;
-    input_tensor->data = data;
+    input_tensor->data.buf = data;
+    input_tensor->data.size = datasize;
     return success;
 }

core/iwasm/libraries/wasi-nn/src/wasi_nn.c

@@ -20,6 +20,10 @@
 #include "wasi_nn_types.h"
 #include "wasm_export.h"
 
+#if WASM_ENABLE_WASI_EPHEMERAL_NN == 0
+#warning You are using "wasi_nn", which is a legacy WAMR-specific ABI. It's deperecated and will likely be removed in future versions of WAMR. Please use "wasi_ephemeral_nn" instead. (For a WASM module, use the wasi_ephemeral_nn.h header instead. For the runtime configurations, enable WASM_ENABLE_WASI_EPHEMERAL_NN/WAMR_BUILD_WASI_EPHEMERAL_NN.)
+#endif
+
 #define HASHMAP_INITIAL_SIZE 20
 #if defined(__APPLE__)
 #define LIB_EXTENTION ".dylib"
@@ -51,53 +55,21 @@ struct backends_api_functions {
             NN_ERR_PRINTF("Error %s() -> %d", #func, wasi_error);          \
     } while (0)
 
-/* HashMap utils */
+static void *wasi_nn_key;
-static HashMap *hashmap;
-
-static uint32
-hash_func(const void *key)
-{
-    // fnv1a_hash
-    const uint32 FNV_PRIME = 16777619;
-    const uint32 FNV_OFFSET_BASIS = 2166136261U;
-
-    uint32 hash = FNV_OFFSET_BASIS;
-    const unsigned char *bytes = (const unsigned char *)key;
-
-    for (size_t i = 0; i < sizeof(uintptr_t); ++i) {
-        hash ^= bytes[i];
-        hash *= FNV_PRIME;
-    }
-
-    return hash;
-}
-
-static bool
-key_equal_func(void *key1, void *key2)
-{
-    return key1 == key2;
-}
-
-static void
-key_destroy_func(void *key1)
-{
-    /* key type is wasm_module_inst_t*. do nothing */
-}
 
 static void
 wasi_nn_ctx_destroy(WASINNContext *wasi_nn_ctx)
 {
-    NN_DBG_PRINTF("[WASI NN] DEINIT...");
-
     if (wasi_nn_ctx == NULL) {
-        NN_ERR_PRINTF(
-            "Error when deallocating memory. WASI-NN context is NULL");
         return;
     }
+    NN_DBG_PRINTF("[WASI NN] DEINIT...");
     NN_DBG_PRINTF("Freeing wasi-nn");
     NN_DBG_PRINTF("-> is_model_loaded: %d", wasi_nn_ctx->is_model_loaded);
     NN_DBG_PRINTF("-> current_encoding: %d", wasi_nn_ctx->backend);
 
+    bh_assert(!wasi_nn_ctx->busy);
+
     /* deinit() the backend */
     if (wasi_nn_ctx->is_backend_ctx_initialized) {
         wasi_nn_error res;
@@ -105,13 +77,14 @@ wasi_nn_ctx_destroy(WASINNContext *wasi_nn_ctx)
                           wasi_nn_ctx->backend_ctx);
     }
 
+    os_mutex_destroy(&wasi_nn_ctx->lock);
     wasm_runtime_free(wasi_nn_ctx);
 }
 
 static void
-value_destroy_func(void *value)
+dtor(wasm_module_inst_t inst, void *ctx)
 {
-    wasi_nn_ctx_destroy((WASINNContext *)value);
+    wasi_nn_ctx_destroy(ctx);
 }
 
 bool
@@ -124,12 +97,9 @@ wasi_nn_initialize()
         return false;
     }
 
-    // hashmap { instance: wasi_nn_ctx }
+    wasi_nn_key = wasm_runtime_create_context_key(dtor);
-    hashmap = bh_hash_map_create(HASHMAP_INITIAL_SIZE, true, hash_func,
+    if (wasi_nn_key == NULL) {
-                                 key_equal_func, key_destroy_func,
+        NN_ERR_PRINTF("Failed to create context key");
-                                 value_destroy_func);
-    if (hashmap == NULL) {
-        NN_ERR_PRINTF("Error while initializing hashmap");
         os_mutex_destroy(&wasi_nn_lock);
         return false;
     }
@@ -150,6 +120,11 @@ wasi_nn_initialize_context()
     }
 
     memset(wasi_nn_ctx, 0, sizeof(WASINNContext));
+    if (os_mutex_init(&wasi_nn_ctx->lock)) {
+        NN_ERR_PRINTF("Error when initializing a lock for WASI-NN context");
+        wasm_runtime_free(wasi_nn_ctx);
+        return NULL;
+    }
     return wasi_nn_ctx;
 }
 
@@ -158,29 +133,59 @@ static WASINNContext *
 wasm_runtime_get_wasi_nn_ctx(wasm_module_inst_t instance)
 {
     WASINNContext *wasi_nn_ctx =
-        (WASINNContext *)bh_hash_map_find(hashmap, (void *)instance);
+        wasm_runtime_get_context(instance, wasi_nn_key);
     if (wasi_nn_ctx == NULL) {
-        wasi_nn_ctx = wasi_nn_initialize_context();
+        WASINNContext *newctx = wasi_nn_initialize_context();
-        if (wasi_nn_ctx == NULL)
+        if (newctx == NULL)
-            return NULL;
-
-        bool ok =
-            bh_hash_map_insert(hashmap, (void *)instance, (void *)wasi_nn_ctx);
-        if (!ok) {
-            NN_ERR_PRINTF("Error while storing context");
-            wasi_nn_ctx_destroy(wasi_nn_ctx);
             return NULL;
+        os_mutex_lock(&wasi_nn_lock);
+        wasi_nn_ctx = wasm_runtime_get_context(instance, wasi_nn_key);
+        if (wasi_nn_ctx == NULL) {
+            wasm_runtime_set_context_spread(instance, wasi_nn_key, newctx);
+            wasi_nn_ctx = newctx;
+            newctx = NULL;
+        }
+        os_mutex_unlock(&wasi_nn_lock);
+        if (newctx != NULL) {
+            wasi_nn_ctx_destroy(newctx);
         }
     }
-
     return wasi_nn_ctx;
 }
 
+static WASINNContext *
+lock_ctx(wasm_module_inst_t instance)
+{
+    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        return NULL;
+    }
+    os_mutex_lock(&wasi_nn_ctx->lock);
+    if (wasi_nn_ctx->busy) {
+        os_mutex_unlock(&wasi_nn_ctx->lock);
+        return NULL;
+    }
+    wasi_nn_ctx->busy = true;
+    os_mutex_unlock(&wasi_nn_ctx->lock);
+    return wasi_nn_ctx;
+}
+
+static void
+unlock_ctx(WASINNContext *wasi_nn_ctx)
+{
+    if (wasi_nn_ctx == NULL) {
+        return;
+    }
+    os_mutex_lock(&wasi_nn_ctx->lock);
+    bh_assert(wasi_nn_ctx->busy);
+    wasi_nn_ctx->busy = false;
+    os_mutex_unlock(&wasi_nn_ctx->lock);
+}
+
 void
 wasi_nn_destroy()
 {
-    // destroy hashmap will destroy keys and values
+    wasm_runtime_destroy_context_key(wasi_nn_key);
-    bh_hash_map_destroy(hashmap);
 
     // close backends' libraries and registered functions
     for (unsigned i = 0; i < sizeof(lookup) / sizeof(lookup[0]); i++) {
@@ -401,7 +406,7 @@ detect_and_load_backend(graph_encoding backend_hint,
 
 static wasi_nn_error
 ensure_backend(wasm_module_inst_t instance, graph_encoding encoding,
-               WASINNContext **wasi_nn_ctx_ptr)
+               WASINNContext *wasi_nn_ctx)
 {
     wasi_nn_error res;
 
@@ -412,7 +417,6 @@ ensure_backend(wasm_module_inst_t instance, graph_encoding encoding,
         goto fail;
     }
 
-    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
     if (wasi_nn_ctx->is_backend_ctx_initialized) {
         if (wasi_nn_ctx->backend != loaded_backend) {
             res = unsupported_operation;
@@ -430,7 +434,6 @@ ensure_backend(wasm_module_inst_t instance, graph_encoding encoding,
 
         wasi_nn_ctx->is_backend_ctx_initialized = true;
     }
-    *wasi_nn_ctx_ptr = wasi_nn_ctx;
     return success;
 fail:
     return res;
@@ -458,17 +461,23 @@ wasi_nn_load(wasm_exec_env_t exec_env, graph_builder_array_wasm *builder,
     if (!instance)
         return runtime_error;
 
+    WASINNContext *wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
     graph_builder_array builder_native = { 0 };
 #if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
     if (success
         != (res = graph_builder_array_app_native(
                 instance, builder, builder_wasm_size, &builder_native)))
-        return res;
+        goto fail;
 #else  /* WASM_ENABLE_WASI_EPHEMERAL_NN == 0 */
     if (success
         != (res = graph_builder_array_app_native(instance, builder,
                                                  &builder_native)))
-        return res;
+        goto fail;
 #endif /* WASM_ENABLE_WASI_EPHEMERAL_NN != 0 */
 
     if (!wasm_runtime_validate_native_addr(instance, g,
@@ -478,8 +487,7 @@ wasi_nn_load(wasm_exec_env_t exec_env, graph_builder_array_wasm *builder,
         goto fail;
     }
 
-    WASINNContext *wasi_nn_ctx;
+    res = ensure_backend(instance, encoding, wasi_nn_ctx);
-    res = ensure_backend(instance, encoding, &wasi_nn_ctx);
     if (res != success)
         goto fail;
 
@@ -494,14 +502,40 @@ fail:
     // XXX: Free intermediate structure pointers
     if (builder_native.buf)
         wasm_runtime_free(builder_native.buf);
+    unlock_ctx(wasi_nn_ctx);
 
     return res;
 }
 
+static wasi_nn_error
+copyin_and_nul_terminate(wasm_module_inst_t inst, char *name, uint32_t name_len,
+                         char **resultp)
+{
+    char *nul_terminated_name;
+    if (!wasm_runtime_validate_native_addr(inst, name, name_len)) {
+        return invalid_argument;
+    }
+    nul_terminated_name = wasm_runtime_malloc(name_len + 1);
+    if (nul_terminated_name == NULL) {
+        return runtime_error;
+    }
+    bh_memcpy_s(nul_terminated_name, name_len + 1, name, name_len);
+    nul_terminated_name[name_len] = '\0'; /* ensure NUL termination */
+    if (strlen(nul_terminated_name) != name_len) {
+        /* reject names containing '\0' for now */
+        wasm_runtime_free(nul_terminated_name);
+        return invalid_argument;
+    }
+    *resultp = nul_terminated_name;
+    return success;
+}
+
 wasi_nn_error
 wasi_nn_load_by_name(wasm_exec_env_t exec_env, char *name, uint32_t name_len,
                      graph *g)
 {
+    WASINNContext *wasi_nn_ctx = NULL;
+    char *nul_terminated_name = NULL;
     wasi_nn_error res;
 
     wasm_module_inst_t instance = wasm_runtime_get_module_inst(exec_env);
@@ -509,36 +543,46 @@ wasi_nn_load_by_name(wasm_exec_env_t exec_env, char *name, uint32_t name_len,
         return runtime_error;
     }
 
-    if (!wasm_runtime_validate_native_addr(instance, name, name_len)) {
-        NN_ERR_PRINTF("name is invalid");
-        return invalid_argument;
-    }
-
     if (!wasm_runtime_validate_native_addr(instance, g,
                                            (uint64)sizeof(graph))) {
         NN_ERR_PRINTF("graph is invalid");
         return invalid_argument;
     }
 
-    if (name_len == 0 || name[name_len] != '\0') {
+    res = copyin_and_nul_terminate(instance, name, name_len,
-        NN_ERR_PRINTF("Invalid filename");
+                                   &nul_terminated_name);
-        return invalid_argument;
+    if (res != success) {
+        goto fail;
     }
 
-    NN_DBG_PRINTF("[WASI NN] LOAD_BY_NAME %s...", name);
+    NN_DBG_PRINTF("[WASI NN] LOAD_BY_NAME %s...", nul_terminated_name);
 
-    WASINNContext *wasi_nn_ctx;
+    wasi_nn_ctx = lock_ctx(instance);
-    res = ensure_backend(instance, autodetect, &wasi_nn_ctx);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
+    res = ensure_backend(instance, autodetect, wasi_nn_ctx);
     if (res != success)
-        return res;
+        goto fail;
 
     call_wasi_nn_func(wasi_nn_ctx->backend, load_by_name, res,
-                      wasi_nn_ctx->backend_ctx, name, name_len, g);
+                      wasi_nn_ctx->backend_ctx, nul_terminated_name, name_len,
+                      g);
     if (res != success)
-        return res;
+        goto fail;
 
     wasi_nn_ctx->is_model_loaded = true;
-    return success;
+    res = success;
+fail:
+    if (nul_terminated_name != NULL) {
+        wasm_runtime_free(nul_terminated_name);
+    }
+    if (wasi_nn_ctx != NULL) {
+        unlock_ctx(wasi_nn_ctx);
+    }
+    return res;
 }
 
 wasi_nn_error
@@ -546,6 +590,9 @@ wasi_nn_load_by_name_with_config(wasm_exec_env_t exec_env, char *name,
                                  int32_t name_len, char *config,
                                  int32_t config_len, graph *g)
 {
+    WASINNContext *wasi_nn_ctx = NULL;
+    char *nul_terminated_name = NULL;
+    char *nul_terminated_config = NULL;
     wasi_nn_error res;
 
     wasm_module_inst_t instance = wasm_runtime_get_module_inst(exec_env);
@@ -553,42 +600,56 @@ wasi_nn_load_by_name_with_config(wasm_exec_env_t exec_env, char *name,
         return runtime_error;
     }
 
-    if (!wasm_runtime_validate_native_addr(instance, name, name_len)) {
-        NN_ERR_PRINTF("name is invalid");
-        return invalid_argument;
-    }
-
     if (!wasm_runtime_validate_native_addr(instance, g,
                                            (uint64)sizeof(graph))) {
         NN_ERR_PRINTF("graph is invalid");
         return invalid_argument;
     }
 
-    if (name_len == 0 || name[name_len] != '\0') {
+    res = copyin_and_nul_terminate(instance, name, name_len,
-        NN_ERR_PRINTF("Invalid filename");
+                                   &nul_terminated_name);
-        return invalid_argument;
+    if (res != success) {
+        goto fail;
+    }
+    res = copyin_and_nul_terminate(instance, config, config_len,
+                                   &nul_terminated_config);
+    if (res != success) {
+        goto fail;
     }
 
-    if (!config || config_len == 0 || config[config_len] != '\0') {
+    NN_DBG_PRINTF("[WASI NN] LOAD_BY_NAME_WITH_CONFIG %s %s...",
-        NN_ERR_PRINTF("Invalid config");
+                  nul_terminated_name, nul_terminated_config);
-        return invalid_argument;
+
+    wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
     }
 
-    NN_DBG_PRINTF("[WASI NN] LOAD_BY_NAME_WITH_CONFIG %s %s...", name, config);
+    res = ensure_backend(instance, autodetect, wasi_nn_ctx);
-
-    WASINNContext *wasi_nn_ctx;
-    res = ensure_backend(instance, autodetect, &wasi_nn_ctx);
     if (res != success)
-        return res;
+        goto fail;
+    ;
 
     call_wasi_nn_func(wasi_nn_ctx->backend, load_by_name_with_config, res,
-                      wasi_nn_ctx->backend_ctx, name, name_len, config,
+                      wasi_nn_ctx->backend_ctx, nul_terminated_name, name_len,
-                      config_len, g);
+                      nul_terminated_config, config_len, g);
     if (res != success)
-        return res;
+        goto fail;
 
     wasi_nn_ctx->is_model_loaded = true;
-    return success;
+    res = success;
+fail:
+    if (nul_terminated_name != NULL) {
+        wasm_runtime_free(nul_terminated_name);
+    }
+    if (nul_terminated_config != NULL) {
+        wasm_runtime_free(nul_terminated_config);
+    }
+    if (wasi_nn_ctx != NULL) {
+        unlock_ctx(wasi_nn_ctx);
+    }
+    return res;
 }
 
 wasi_nn_error
@@ -602,20 +663,27 @@ wasi_nn_init_execution_context(wasm_exec_env_t exec_env, graph g,
         return runtime_error;
     }
 
-    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
-
     wasi_nn_error res;
+    WASINNContext *wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
     if (success != (res = is_model_initialized(wasi_nn_ctx)))
-        return res;
+        goto fail;
 
     if (!wasm_runtime_validate_native_addr(
             instance, ctx, (uint64)sizeof(graph_execution_context))) {
         NN_ERR_PRINTF("ctx is invalid");
-        return invalid_argument;
+        res = invalid_argument;
+        goto fail;
     }
 
     call_wasi_nn_func(wasi_nn_ctx->backend, init_execution_context, res,
                       wasi_nn_ctx->backend_ctx, g, ctx);
+fail:
+    unlock_ctx(wasi_nn_ctx);
     return res;
 }
 
@@ -630,17 +698,21 @@ wasi_nn_set_input(wasm_exec_env_t exec_env, graph_execution_context ctx,
         return runtime_error;
     }
 
-    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
-
     wasi_nn_error res;
+    WASINNContext *wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
     if (success != (res = is_model_initialized(wasi_nn_ctx)))
-        return res;
+        goto fail;
 
     tensor input_tensor_native = { 0 };
     if (success
         != (res = tensor_app_native(instance, input_tensor,
                                     &input_tensor_native)))
-        return res;
+        goto fail;
 
     call_wasi_nn_func(wasi_nn_ctx->backend, set_input, res,
                       wasi_nn_ctx->backend_ctx, ctx, index,
@@ -648,7 +720,8 @@ wasi_nn_set_input(wasm_exec_env_t exec_env, graph_execution_context ctx,
     // XXX: Free intermediate structure pointers
     if (input_tensor_native.dimensions)
         wasm_runtime_free(input_tensor_native.dimensions);
-
+fail:
+    unlock_ctx(wasi_nn_ctx);
     return res;
 }
 
@@ -662,26 +735,32 @@ wasi_nn_compute(wasm_exec_env_t exec_env, graph_execution_context ctx)
         return runtime_error;
     }
 
-    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
-
     wasi_nn_error res;
+    WASINNContext *wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
     if (success != (res = is_model_initialized(wasi_nn_ctx)))
-        return res;
+        goto fail;
 
     call_wasi_nn_func(wasi_nn_ctx->backend, compute, res,
                       wasi_nn_ctx->backend_ctx, ctx);
+fail:
+    unlock_ctx(wasi_nn_ctx);
     return res;
 }
 
 #if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
 wasi_nn_error
 wasi_nn_get_output(wasm_exec_env_t exec_env, graph_execution_context ctx,
-                   uint32_t index, tensor_data output_tensor,
+                   uint32_t index, void *output_tensor,
                    uint32_t output_tensor_len, uint32_t *output_tensor_size)
 #else  /* WASM_ENABLE_WASI_EPHEMERAL_NN == 0 */
 wasi_nn_error
 wasi_nn_get_output(wasm_exec_env_t exec_env, graph_execution_context ctx,
-                   uint32_t index, tensor_data output_tensor,
+                   uint32_t index, void *output_tensor,
                    uint32_t *output_tensor_size)
 #endif /* WASM_ENABLE_WASI_EPHEMERAL_NN != 0 */
 {
@@ -692,28 +771,36 @@ wasi_nn_get_output(wasm_exec_env_t exec_env, graph_execution_context ctx,
         return runtime_error;
     }
 
-    WASINNContext *wasi_nn_ctx = wasm_runtime_get_wasi_nn_ctx(instance);
-
     wasi_nn_error res;
+    WASINNContext *wasi_nn_ctx = lock_ctx(instance);
+    if (wasi_nn_ctx == NULL) {
+        res = busy;
+        goto fail;
+    }
+
     if (success != (res = is_model_initialized(wasi_nn_ctx)))
-        return res;
+        goto fail;
 
     if (!wasm_runtime_validate_native_addr(instance, output_tensor_size,
                                            (uint64)sizeof(uint32_t))) {
         NN_ERR_PRINTF("output_tensor_size is invalid");
-        return invalid_argument;
+        res = invalid_argument;
+        goto fail;
     }
 
+    tensor_data tensor = {
+        .buf = output_tensor,
 #if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
+        .size = output_tensor_len,
+#else
+        .size = *output_tensor_size,
+#endif
+    };
     call_wasi_nn_func(wasi_nn_ctx->backend, get_output, res,
-                      wasi_nn_ctx->backend_ctx, ctx, index, output_tensor,
+                      wasi_nn_ctx->backend_ctx, ctx, index, &tensor,
-                      &output_tensor_len);
-    *output_tensor_size = output_tensor_len;
-#else  /* WASM_ENABLE_WASI_EPHEMERAL_NN == 0 */
-    call_wasi_nn_func(wasi_nn_ctx->backend, get_output, res,
-                      wasi_nn_ctx->backend_ctx, ctx, index, output_tensor,
                       output_tensor_size);
-#endif /* WASM_ENABLE_WASI_EPHEMERAL_NN != 0 */
+fail:
+    unlock_ctx(wasi_nn_ctx);
     return res;
 }
 
@@ -728,6 +815,12 @@ static NativeSymbol native_symbols_wasi_nn[] = {
 #if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
     REG_NATIVE_FUNC(load, "(*iii*)i"),
     REG_NATIVE_FUNC(load_by_name, "(*i*)i"),
+
+    /* load_by_name_with_config is intented to be compatible with
+     * a wasmedge extension.
+     * https://github.com/second-state/wasmedge-wasi-nn/pull/2
+     * https://github.com/WasmEdge/WasmEdge/blob/5553924e8cdccdc2cbd2a6a6d0ed9b11250c353e/plugins/wasi_nn/wasinnmodule.cpp#L13-L14
+     */
     REG_NATIVE_FUNC(load_by_name_with_config, "(*i*i*)i"),
     REG_NATIVE_FUNC(init_execution_context, "(i*)i"),
     REG_NATIVE_FUNC(set_input, "(ii*)i"),

core/iwasm/libraries/wasi-nn/src/wasi_nn_backend.h

@@ -3,15 +3,26 @@
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  */
 
-#ifndef WASI_NN_OPENVINO_HPP
+#ifndef WASI_NN_BACKEND_H
-#define WASI_NN_OPENVINO_HPP
+#define WASI_NN_BACKEND_H
 
 #include "wasi_nn_types.h"
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
 __attribute__((visibility("default"))) wasi_nn_error
 load(void *ctx, graph_builder_array *builder, graph_encoding encoding,
      execution_target target, graph *g);
 
+__attribute__((visibility("default"))) wasi_nn_error
+load_by_name(void *tflite_ctx, const char *name, uint32_t namelen, graph *g);
+
+__attribute__((visibility("default"))) wasi_nn_error
+load_by_name_with_config(void *ctx, const char *name, uint32_t namelen,
+                         const char *config, uint32_t config_len, graph *g);
+
 __attribute__((visibility("default"))) wasi_nn_error
 init_execution_context(void *ctx, graph g, graph_execution_context *exec_ctx);
 
@@ -24,7 +35,7 @@ compute(void *ctx, graph_execution_context exec_ctx);
 
 __attribute__((visibility("default"))) wasi_nn_error
 get_output(void *ctx, graph_execution_context exec_ctx, uint32_t index,
-           tensor_data output_tensor, uint32_t *output_tensor_size);
+           tensor_data *output_tensor, uint32_t *output_tensor_size);
 
 __attribute__((visibility("default"))) wasi_nn_error
 init_backend(void **ctx);
@@ -32,4 +43,8 @@ init_backend(void **ctx);
 __attribute__((visibility("default"))) wasi_nn_error
 deinit_backend(void *ctx);
 
-#endif /* WASI_NN_OPENVINO_HPP */
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* WASI_NN_BACKEND_H */

core/iwasm/libraries/wasi-nn/src/wasi_nn_llamacpp.c

@@ -2,7 +2,10 @@
  * Copyright (C) 2019 Intel Corporation.  All rights reserved.
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  */
-#include "wasi_nn_types.h"
+
+#include <stdlib.h>
+
+#include "wasi_nn_backend.h"
 #include "utils/logger.h"
 #include "llama.h"
 #include "ggml.h"
@@ -14,6 +17,10 @@ extern char const *LLAMA_COMMIT;
 extern char const *LLAMA_COMPILER;
 extern char const *LLAMA_BUILD_TARGET;
 
+#if WASM_ENABLE_WASI_EPHEMERAL_NN == 0
+#error This backend doesn't support legacy "wasi_nn" abi. Please enable WASM_ENABLE_WASI_EPHEMERAL_NN.
+#endif
+
 // compatible with WasmEdge
 // https://github.com/second-state/WasmEdge-WASINN-examples/blob/master/wasmedge-ggml/README.md#parameters
 // https://github.com/WasmEdge/WasmEdge/blob/master/plugins/wasi_nn/ggml.cpp
@@ -286,7 +293,7 @@ deinit_backend(void *ctx)
 
     llama_backend_free();
 
-    os_free(backend_ctx);
+    free(backend_ctx);
     return success;
 }
 
@@ -302,6 +309,11 @@ __load_by_name_with_configuration(void *ctx, const char *filename, graph *g)
 {
     struct LlamaContext *backend_ctx = (struct LlamaContext *)ctx;
 
+    if (backend_ctx->model != NULL) {
+        // we only implement a single graph
+        return unsupported_operation;
+    }
+
     // make sure backend_ctx->config is initialized
 
     struct llama_model_params model_params =
@@ -320,6 +332,7 @@ __load_by_name_with_configuration(void *ctx, const char *filename, graph *g)
 #endif
 
     backend_ctx->model = model;
+    *g = 0;
 
     return success;
 }
@@ -360,6 +373,16 @@ init_execution_context(void *ctx, graph g, graph_execution_context *exec_ctx)
 {
     struct LlamaContext *backend_ctx = (struct LlamaContext *)ctx;
 
+    if (g != 0 || backend_ctx->model == NULL) {
+        // we only implement a single graph
+        return runtime_error;
+    }
+
+    if (backend_ctx->ctx != NULL) {
+        // we only implement a single context
+        return unsupported_operation;
+    }
+
     struct llama_context_params ctx_params =
         llama_context_params_from_wasi_nn_llama_config(&backend_ctx->config);
     struct llama_context *llama_ctx =
@@ -370,6 +393,7 @@ init_execution_context(void *ctx, graph g, graph_execution_context *exec_ctx)
     }
 
     backend_ctx->ctx = llama_ctx;
+    *exec_ctx = 0;
 
     NN_INFO_PRINTF("n_predict = %d, n_ctx = %d", backend_ctx->config.n_predict,
                    llama_n_ctx(backend_ctx->ctx));
@@ -381,18 +405,41 @@ set_input(void *ctx, graph_execution_context exec_ctx, uint32_t index,
           tensor *wasi_nn_tensor)
 {
     struct LlamaContext *backend_ctx = (struct LlamaContext *)ctx;
-    // tensor->data is the prompt string. ends with \0
+
-    char *prompt_text = (char *)wasi_nn_tensor->data;
+    if (exec_ctx != 0 || backend_ctx->ctx == NULL) {
+        // we only implement a single context
+        return runtime_error;
+    }
+
+    if (index != 0) {
+        NN_ERR_PRINTF("Invalid input index %d", index);
+        return invalid_argument;
+    }
+
+    // tensor->data is the prompt string.
+    char *prompt_text = (char *)wasi_nn_tensor->data.buf;
+    uint32_t prompt_text_len = wasi_nn_tensor->data.size;
+
+    // note: buf[0] == 1 is a workaround for
+    // https://github.com/second-state/WasmEdge-WASINN-examples/issues/196.
+    // we may remove it in future.
+    if (wasi_nn_tensor->type != u8 || wasi_nn_tensor->dimensions->size != 1
+        || !(wasi_nn_tensor->dimensions->buf[0] == 1
+             || wasi_nn_tensor->dimensions->buf[0] == prompt_text_len)) {
+        return invalid_argument;
+    }
+    if (wasi_nn_tensor->dimensions->buf[0] == 1 && prompt_text_len != 1) {
+        NN_WARN_PRINTF("Ignoring seemingly wrong input tensor dimensions.");
+    }
 
 #ifndef NDEBUG
     NN_DBG_PRINTF("--------------------------------------------------");
-    NN_DBG_PRINTF("prompt_text: %s", prompt_text);
+    NN_DBG_PRINTF("prompt_text: %.*s", (int)prompt_text_len, prompt_text);
     NN_DBG_PRINTF("--------------------------------------------------");
 #endif
 
     // tokenize the prompt
     uint32_t n_token_max = llama_n_ctx(backend_ctx->ctx);
-    uint32_t prompt_text_len = strlen(prompt_text);
 
     if (backend_ctx->prompt == NULL) {
         backend_ctx->prompt = calloc(n_token_max, sizeof(llama_token));
@@ -430,6 +477,11 @@ compute(void *ctx, graph_execution_context exec_ctx)
     struct LlamaContext *backend_ctx = (struct LlamaContext *)ctx;
     wasi_nn_error ret = runtime_error;
 
+    if (exec_ctx != 0 || backend_ctx->ctx == NULL) {
+        // we only implement a single context
+        return runtime_error;
+    }
+
     // reset the generation buffer
     if (backend_ctx->generation == NULL) {
         backend_ctx->generation =
@@ -477,7 +529,6 @@ compute(void *ctx, graph_execution_context exec_ctx)
 
     // main loop
     int32_t n_cur = batch.n_tokens;
-    int n_decode = 0;
     int32_t n_vocab = llama_n_vocab(backend_ctx->model);
     llama_token_data *candidates = NULL;
 
@@ -528,7 +579,6 @@ compute(void *ctx, graph_execution_context exec_ctx)
         // push this new token for next evaluation
         llama_batch_add(&batch, new_token_id, n_cur, seq_ids,
                         sizeof(seq_ids) / sizeof(seq_ids[0]), true);
-        n_decode++;
         n_cur++;
 
         if (llama_decode(backend_ctx->ctx, batch) != 0) {
@@ -549,10 +599,15 @@ fail:
 
 __attribute__((visibility("default"))) wasi_nn_error
 get_output(void *ctx, graph_execution_context exec_ctx, uint32_t index,
-           tensor_data output_tensor, uint32_t *output_tensor_size)
+           tensor_data *output_tensor, uint32_t *output_tensor_size)
 {
     struct LlamaContext *backend_ctx = (struct LlamaContext *)ctx;
 
+    if (exec_ctx != 0 || backend_ctx->ctx == NULL) {
+        // we only implement a single context
+        return runtime_error;
+    }
+
     // Compatibility with WasmEdge
     if (index > 1) {
         NN_ERR_PRINTF("Invalid output index %d", index);
@@ -568,7 +623,7 @@ get_output(void *ctx, graph_execution_context exec_ctx, uint32_t index,
             printf("%s\n", output_metadata);
         }
 
-        memcpy(output_tensor, output_metadata, strlen(output_metadata));
+        memcpy(output_tensor->buf, output_metadata, strlen(output_metadata));
         *output_tensor_size = strlen(output_metadata);
         return success;
     }
@@ -588,7 +643,7 @@ get_output(void *ctx, graph_execution_context exec_ctx, uint32_t index,
             printf("%s", buf);
         }
 
-        memcpy(output_tensor + end_pos, buf, strlen(buf));
+        memcpy(output_tensor->buf + end_pos, buf, strlen(buf));
         end_pos += strlen(buf);
     }
 

core/iwasm/libraries/wasi-nn/src/wasi_nn_openvino.c

@@ -3,13 +3,16 @@
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  */
 
-#include "wasi_nn_types.h"
+#include "wasi_nn_backend.h"
-#include "wasi_nn_openvino.h"
 #include "utils/logger.h"
 #include "bh_platform.h"
 
 #include "openvino/c/openvino.h"
 
+#if WASM_ENABLE_WASI_EPHEMERAL_NN == 0
+#error This backend doesn't support legacy "wasi_nn" abi. Please enable WASM_ENABLE_WASI_EPHEMERAL_NN.
+#endif
+
 /*
  * refer to
  * https://docs.openvino.ai/2024/openvino-workflow/running-inference/integrate-openvino-with-your-application.html
@@ -26,15 +29,25 @@
  * from 4. to 6. is the Inference Loop
  */
 
+/* these limits are arbitrary. */
+#define MAX_GRAPHS 4
+#define MAX_EXECUTION_CONTEXTS 4
+
 typedef struct {
     ov_core_t *core;
     /* keep input model files */
+    struct OpenVINOGraph {
         void *weight_data;
         ov_tensor_t *weights_tensor;
         ov_model_t *model;
         ov_compiled_model_t *compiled_model;
+    } graphs[MAX_GRAPHS];
+    struct OpenVINOExecutionContext {
+        struct OpenVINOGraph *graph;
         ov_infer_request_t *infer_request;
-    ov_tensor_t *input_tensor;
+    } execution_contexts[MAX_EXECUTION_CONTEXTS];
+    unsigned int n_graphs;
+    unsigned int n_execution_contexts;
 } OpenVINOContext;
 
 /*
@@ -134,7 +147,7 @@ print_model_input_output_info(ov_model_t *model)
         output_port = NULL;
     }
 
-    ov_error = ov_error;
+    (void)ov_error;
 fail:
     if (friendly_name)
         ov_free(friendly_name);
@@ -179,6 +192,29 @@ wasi_nn_tensor_type_to_openvino_element_type(tensor_type wasi_nn_type)
     return UNDEFINED;
 }
 
+static void
+free_graph(struct OpenVINOGraph *graph)
+{
+    if (graph->weight_data)
+        os_free(graph->weight_data);
+
+    if (graph->weights_tensor)
+        ov_tensor_free(graph->weights_tensor);
+
+    if (graph->model)
+        ov_model_free(graph->model);
+
+    if (graph->compiled_model)
+        ov_compiled_model_free(graph->compiled_model);
+}
+
+static void
+free_execution_context(struct OpenVINOExecutionContext *c)
+{
+    if (c->infer_request)
+        ov_infer_request_free(c->infer_request);
+}
+
 static wasi_nn_error
 uint32_array_to_int64_array(uint32_t array_size, uint32_t *src, int64_t **dst)
 {
@@ -198,6 +234,8 @@ load(void *ctx, graph_builder_array *builder, graph_encoding encoding,
      execution_target target, graph *g)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOGraph *graph;
+    unsigned int graph_idx;
     wasi_nn_error ret = unsupported_operation;
 
     if (encoding != openvino) {
@@ -223,33 +261,47 @@ load(void *ctx, graph_builder_array *builder, graph_encoding encoding,
     graph_builder xml = builder->buf[0];
     graph_builder weight = builder->buf[1];
 
+    graph_idx = ov_ctx->n_graphs;
+    if (graph_idx >= MAX_GRAPHS) {
+        return runtime_error;
+    }
+    graph = &ov_ctx->graphs[graph_idx];
+    memset(graph, 0, sizeof(*graph));
+
     /* transfer weight to an ov tensor */
     {
-        ov_ctx->weight_data = os_malloc(weight.size);
+        graph->weight_data = os_malloc(weight.size);
-        if (!ov_ctx->weight_data)
+        if (!graph->weight_data)
             goto fail;
-        memcpy(ov_ctx->weight_data, weight.buf, weight.size);
+        memcpy(graph->weight_data, weight.buf, weight.size);
 
         ov_element_type_e type = U8;
         int64_t dims[1] = { weight.size };
         ov_shape_t shape = { 1, dims };
         CHECK_OV_STATUS(ov_tensor_create_from_host_ptr(type, shape,
-                                                       ov_ctx->weight_data,
+                                                       graph->weight_data,
-                                                       &ov_ctx->weights_tensor),
+                                                       &graph->weights_tensor),
                         ret);
     }
 
     /* load model from buffer */
     CHECK_OV_STATUS(ov_core_read_model_from_memory_buffer(
                         ov_ctx->core, (char *)xml.buf, xml.size,
-                        ov_ctx->weights_tensor, &ov_ctx->model),
+                        graph->weights_tensor, &graph->model),
                     ret);
 #ifndef NDEBUG
-    print_model_input_output_info(ov_ctx->model);
+    print_model_input_output_info(graph->model);
 #endif
 
-    ret = success;
+    CHECK_OV_STATUS(ov_core_compile_model(ov_ctx->core, graph->model, "CPU", 0,
+                                          &graph->compiled_model),
+                    ret);
+
+    *g = graph_idx;
+    ov_ctx->n_graphs++;
+    return success;
 fail:
+    free_graph(graph);
     return ret;
 }
 
@@ -257,20 +309,62 @@ __attribute__((visibility("default"))) wasi_nn_error
 load_by_name(void *ctx, const char *filename, uint32_t filename_len, graph *g)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOGraph *graph;
+    unsigned int graph_idx;
     wasi_nn_error ret = unsupported_operation;
 
-    CHECK_OV_STATUS(
+    graph_idx = ov_ctx->n_graphs;
-        ov_core_read_model(ov_ctx->core, filename, NULL, &ov_ctx->model), ret);
+    if (graph_idx >= MAX_GRAPHS) {
+        return runtime_error;
+    }
+    graph = &ov_ctx->graphs[graph_idx];
 
-    ret = success;
+    memset(graph, 0, sizeof(*graph));
+    CHECK_OV_STATUS(
+        ov_core_read_model(ov_ctx->core, filename, NULL, &graph->model), ret);
+
+    CHECK_OV_STATUS(ov_core_compile_model(ov_ctx->core, graph->model, "CPU", 0,
+                                          &graph->compiled_model),
+                    ret);
+
+    *g = graph_idx;
+    ov_ctx->n_graphs++;
+    return success;
 fail:
+    free_graph(graph);
     return ret;
 }
 
 __attribute__((visibility("default"))) wasi_nn_error
 init_execution_context(void *ctx, graph g, graph_execution_context *exec_ctx)
 {
+    OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOGraph *graph;
+    struct OpenVINOExecutionContext *exec;
+    unsigned int exec_idx;
+    wasi_nn_error ret;
+
+    if (g >= ov_ctx->n_graphs)
+        return runtime_error;
+    graph = &ov_ctx->graphs[g];
+
+    exec_idx = ov_ctx->n_execution_contexts;
+    if (exec_idx >= MAX_EXECUTION_CONTEXTS)
+        return runtime_error;
+    exec = &ov_ctx->execution_contexts[exec_idx];
+
+    memset(exec, 0, sizeof(*exec));
+    exec->graph = graph;
+
+    CHECK_OV_STATUS(ov_compiled_model_create_infer_request(
+                        graph->compiled_model, &exec->infer_request),
+                    ret);
+
+    *exec_ctx = exec_idx;
+    ov_ctx->n_execution_contexts++;
     return success;
+fail:
+    return ret;
 }
 
 __attribute__((visibility("default"))) wasi_nn_error
@@ -278,10 +372,16 @@ set_input(void *ctx, graph_execution_context exec_ctx, uint32_t index,
           tensor *wasi_nn_tensor)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOExecutionContext *exec;
     wasi_nn_error ret = unsupported_operation;
     ov_shape_t input_shape = { 0 };
+    ov_tensor_t *input_tensor = NULL;
     int64_t *ov_dims = NULL;
 
+    if (exec_ctx >= ov_ctx->n_execution_contexts)
+        return runtime_error;
+    exec = &ov_ctx->execution_contexts[exec_ctx];
+
     /* wasi_nn_tensor -> ov_tensor */
     {
         ret = uint32_array_to_int64_array(wasi_nn_tensor->dimensions->size,
@@ -305,28 +405,21 @@ set_input(void *ctx, graph_execution_context exec_ctx, uint32_t index,
                       shape_info);
 
         CHECK_OV_STATUS(ov_tensor_create_from_host_ptr(input_type, input_shape,
-                                                       wasi_nn_tensor->data,
+                                                       wasi_nn_tensor->data.buf,
-                                                       &ov_ctx->input_tensor),
+                                                       &input_tensor),
                         ret);
     }
 
-    CHECK_OV_STATUS(ov_core_compile_model(ov_ctx->core, ov_ctx->model, "CPU", 0,
-                                          &ov_ctx->compiled_model),
-                    ret);
-
-    CHECK_OV_STATUS(ov_compiled_model_create_infer_request(
-                        ov_ctx->compiled_model, &ov_ctx->infer_request),
-                    ret);
-
     /* install ov_tensor -> infer_request */
     CHECK_OV_STATUS(ov_infer_request_set_input_tensor_by_index(
-                        ov_ctx->infer_request, index, ov_ctx->input_tensor),
+                        exec->infer_request, index, input_tensor),
                     ret);
     ret = success;
-
 fail:
     if (ov_dims)
         os_free(ov_dims);
+    if (input_tensor)
+        ov_tensor_free(input_tensor);
     ov_shape_free(&input_shape);
 
     return ret;
@@ -336,9 +429,14 @@ __attribute__((visibility("default"))) wasi_nn_error
 compute(void *ctx, graph_execution_context exec_ctx)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOExecutionContext *exec;
     wasi_nn_error ret = unsupported_operation;
 
-    CHECK_OV_STATUS(ov_infer_request_infer(ov_ctx->infer_request), ret);
+    if (exec_ctx >= ov_ctx->n_execution_contexts)
+        return runtime_error;
+    exec = &ov_ctx->execution_contexts[exec_ctx];
+
+    CHECK_OV_STATUS(ov_infer_request_infer(exec->infer_request), ret);
     ret = success;
 fail:
     return ret;
@@ -346,28 +444,33 @@ fail:
 
 __attribute__((visibility("default"))) wasi_nn_error
 get_output(void *ctx, graph_execution_context exec_ctx, uint32_t index,
-           tensor_data output_tensor, uint32_t *output_tensor_size)
+           tensor_data *output_tensor, uint32_t *output_tensor_size)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    struct OpenVINOExecutionContext *exec;
     wasi_nn_error ret = unsupported_operation;
     ov_tensor_t *ov_tensor = NULL;
     void *data = NULL;
     size_t byte_size = 0;
 
+    if (exec_ctx >= ov_ctx->n_execution_contexts)
+        return runtime_error;
+    exec = &ov_ctx->execution_contexts[exec_ctx];
+
     CHECK_OV_STATUS(ov_infer_request_get_output_tensor_by_index(
-                        ov_ctx->infer_request, index, &ov_tensor),
+                        exec->infer_request, index, &ov_tensor),
                     ret);
 
     CHECK_OV_STATUS(ov_tensor_get_byte_size(ov_tensor, &byte_size), ret);
 
-    if (byte_size > *output_tensor_size) {
+    if (byte_size > output_tensor->size) {
         ret = too_large;
         goto fail;
     }
 
     CHECK_OV_STATUS(ov_tensor_data(ov_tensor, &data), ret);
 
-    memcpy(output_tensor, data, byte_size);
+    memcpy(output_tensor->buf, data, byte_size);
 
     *output_tensor_size = (uint32_t)byte_size;
 
@@ -421,27 +524,16 @@ __attribute__((visibility("default"))) wasi_nn_error
 deinit_backend(void *ctx)
 {
     OpenVINOContext *ov_ctx = (OpenVINOContext *)ctx;
+    unsigned int i;
 
     if (!ov_ctx)
         return invalid_argument;
 
-    if (ov_ctx->weight_data)
+    for (i = 0; i < ov_ctx->n_execution_contexts; i++)
-        os_free(ov_ctx->weight_data);
+        free_execution_context(&ov_ctx->execution_contexts[i]);
 
-    if (ov_ctx->weights_tensor)
+    for (i = 0; i < ov_ctx->n_graphs; i++)
-        ov_tensor_free(ov_ctx->weights_tensor);
+        free_graph(&ov_ctx->graphs[i]);
-
-    if (ov_ctx->input_tensor)
-        ov_tensor_free(ov_ctx->input_tensor);
-
-    if (ov_ctx->infer_request)
-        ov_infer_request_free(ov_ctx->infer_request);
-
-    if (ov_ctx->compiled_model)
-        ov_compiled_model_free(ov_ctx->compiled_model);
-
-    if (ov_ctx->model)
-        ov_model_free(ov_ctx->model);
 
     if (ov_ctx->core)
         ov_core_free(ov_ctx->core);

core/iwasm/libraries/wasi-nn/src/wasi_nn_private.h

@@ -9,7 +9,11 @@
 #include "wasi_nn_types.h"
 #include "wasm_export.h"
 
+#include "bh_platform.h"
+
 typedef struct {
+    korp_mutex lock;
+    bool busy;
     bool is_backend_ctx_initialized;
     bool is_model_loaded;
     graph_encoding backend;
@@ -28,7 +32,7 @@ typedef wasi_nn_error (*SET_INPUT)(void *, graph_execution_context, uint32_t,
                                    tensor *);
 typedef wasi_nn_error (*COMPUTE)(void *, graph_execution_context);
 typedef wasi_nn_error (*GET_OUTPUT)(void *, graph_execution_context, uint32_t,
-                                    tensor_data, uint32_t *);
+                                    tensor_data *, uint32_t *);
 /* wasi-nn general APIs */
 typedef wasi_nn_error (*BACKEND_INITIALIZE)(void **);
 typedef wasi_nn_error (*BACKEND_DEINITIALIZE)(void *);

core/iwasm/libraries/wasi-nn/src/wasi_nn_tensorflowlite.cpp

@@ -3,13 +3,13 @@
  * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
  */
 
-#include "wasi_nn_tensorflowlite.hpp"
 #include "utils/logger.h"
 
 #include "bh_platform.h"
-#include "wasi_nn_types.h"
+#include "wasi_nn_backend.h"
 #include "wasm_export.h"
 
+#include <tensorflow/lite/c/c_api.h>
 #include <tensorflow/lite/interpreter.h>
 #include <tensorflow/lite/kernels/register.h>
 #include <tensorflow/lite/model.h>
@@ -280,6 +280,22 @@ set_input(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
           tensor *input_tensor)
 {
     TFLiteContext *tfl_ctx = (TFLiteContext *)tflite_ctx;
+    TfLiteType tfl_type;
+
+    switch (input_tensor->type) {
+        case fp32:
+            tfl_type = TfLiteType::kTfLiteFloat32;
+            break;
+#if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
+        case u8:
+            tfl_type = TfLiteType::kTfLiteUInt8;
+            break;
+#endif
+        default:
+            NN_ERR_PRINTF("unsupported input tensor type %u",
+                          input_tensor->type);
+            return runtime_error;
+    }
 
     if (input_tensor->type != fp32) {
         NN_ERR_PRINTF("unsupported input tensor type %u", input_tensor->type);
@@ -290,19 +306,32 @@ set_input(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
     if (success != (res = is_valid_graph_execution_context(tfl_ctx, ctx)))
         return res;
 
-    uint32_t num_tensors =
+    auto interpreter = tfl_ctx->interpreters[ctx].interpreter.get();
-        tfl_ctx->interpreters[ctx].interpreter->inputs().size();
+
+    uint32_t num_tensors = interpreter->inputs().size();
     NN_DBG_PRINTF("Number of tensors (%d)", num_tensors);
     if (index + 1 > num_tensors) {
         return runtime_error;
     }
 
-    auto tensor = tfl_ctx->interpreters[ctx].interpreter->input_tensor(index);
+    auto tensor = interpreter->input_tensor(index);
     if (tensor == NULL) {
         NN_ERR_PRINTF("Missing memory");
         return too_large;
     }
 
+#if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
+    if (TfLiteTensorType(tensor) != tfl_type) {
+        NN_ERR_PRINTF("Type mismatch");
+        return runtime_error;
+    }
+
+    if (TfLiteTensorCopyFromBuffer(tensor, input_tensor->data.buf,
+                                   input_tensor->data.size)
+        != kTfLiteOk) {
+        return runtime_error;
+    }
+#else
     uint32_t model_tensor_size = 1;
     for (int i = 0; i < tensor->dims->size; ++i)
         model_tensor_size *= (uint32_t)tensor->dims->data[i];
@@ -324,7 +353,7 @@ set_input(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
                 index);
 
         int size = model_tensor_size * sizeof(float);
-        bh_memcpy_s(it, size, input_tensor->data, size);
+        bh_memcpy_s(it, size, input_tensor->data.buf, size);
     }
     else { // TODO: Assuming uint8 quantized networks.
         TfLiteAffineQuantization *quant_info =
@@ -342,11 +371,12 @@ set_input(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
         NN_DBG_PRINTF("input tensor: (scale, offset) = (%f, %f)", scale,
                       zero_point);
 
-        float *input_tensor_f = (float *)input_tensor->data;
+        float *input_tensor_f = (float *)input_tensor->data.buf;
         for (uint32_t i = 0; i < model_tensor_size; ++i) {
             it[i] = (uint8_t)(input_tensor_f[i] / scale + zero_point);
         }
     }
+#endif
 
     return success;
 }
@@ -366,7 +396,7 @@ compute(void *tflite_ctx, graph_execution_context ctx)
 
 __attribute__((visibility("default"))) wasi_nn_error
 get_output(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
-           tensor_data output_tensor, uint32_t *output_tensor_size)
+           tensor_data *output_tensor, uint32_t *output_tensor_size)
 {
     TFLiteContext *tfl_ctx = (TFLiteContext *)tflite_ctx;
 
@@ -389,23 +419,34 @@ get_output(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
         return too_large;
     }
 
-    uint32_t model_tensor_size = 1;
+#if WASM_ENABLE_WASI_EPHEMERAL_NN != 0
-    for (int i = 0; i < (int)tensor->dims->size; ++i)
+    size_t sz = TfLiteTensorByteSize(tensor);
-        model_tensor_size *= (uint32_t)tensor->dims->data[i];
+    if (output_tensor->size < sz) {
-
-    if (*output_tensor_size < model_tensor_size) {
         NN_ERR_PRINTF("Insufficient memory to copy tensor %d", index);
         return too_large;
     }
-
+    if (TfLiteTensorCopyToBuffer(tensor, output_tensor->buf, sz) != kTfLiteOk) {
+        return runtime_error;
+    }
+    *output_tensor_size = sz;
+#else
     if (tensor->quantization.type == kTfLiteNoQuantization) {
         NN_DBG_PRINTF("No quantization information");
-        float *ot =
+        /*
-            tfl_ctx->interpreters[ctx].interpreter->typed_output_tensor<float>(
+         * for now, maintain the bug-to-bug compatibility with the old abi,
-                index);
+         * where the size here is the number of fp32, not bytes.
-
+         */
-        int size = model_tensor_size * sizeof(float);
+        if (output_tensor->size < tensor->bytes / sizeof(float)) {
-        bh_memcpy_s(output_tensor, size, ot, size);
+            NN_ERR_PRINTF("Insufficient memory to copy tensor %d", index);
+            return too_large;
+        }
+        bh_memcpy_s(output_tensor->buf, output_tensor->size, tensor->data.data,
+                    tensor->bytes);
+        /*
+         * for now, maintain the bug-to-bug compatibility with the old abi,
+         * where the size here is the number of fp32, not bytes.
+         */
+        *output_tensor_size = tensor->bytes / sizeof(float);
     }
     else { // TODO: Assuming uint8 quantized networks.
         TfLiteAffineQuantization *quant_info =
@@ -414,6 +455,20 @@ get_output(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
             NN_ERR_PRINTF("Quantization per channel is not supported");
             return runtime_error;
         }
+
+        uint32_t model_tensor_size = 1;
+        for (int i = 0; i < (int)tensor->dims->size; ++i)
+            model_tensor_size *= (uint32_t)tensor->dims->data[i];
+
+        /*
+         * for now, maintain the bug-to-bug compatibility with the old abi,
+         * where the size here is the number of fp32, not bytes.
+         */
+        if (output_tensor->size < model_tensor_size) {
+            NN_ERR_PRINTF("Insufficient memory to copy tensor %d", index);
+            return too_large;
+        }
+
         uint8_t *ot = tfl_ctx->interpreters[ctx]
                           .interpreter->typed_output_tensor<uint8_t>(index);
 
@@ -422,13 +477,19 @@ get_output(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
         NN_DBG_PRINTF("output tensor: (scale, offset) = (%f, %f)", scale,
                       zero_point);
 
-        float *output_tensor_f = (float *)output_tensor;
+        float *output_tensor_f = (float *)output_tensor->buf;
         for (uint32_t i = 0; i < model_tensor_size; ++i) {
             output_tensor_f[i] = (ot[i] - zero_point) * scale;
         }
-    }
 
+        /*
+         * for now, maintain the bug-to-bug compatibility with the old abi,
+         * where the size here is the number of fp32, not bytes.
+         */
         *output_tensor_size = model_tensor_size;
+    }
+#endif
+
     return success;
 }
 

core/iwasm/libraries/wasi-nn/src/wasi_nn_tensorflowlite.hpp

@@ -1,47 +0,0 @@
-/*
- * Copyright (C) 2019 Intel Corporation.  All rights reserved.
- * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
- */
-
-#ifndef WASI_NN_TENSORFLOWLITE_HPP
-#define WASI_NN_TENSORFLOWLITE_HPP
-
-#include "wasi_nn_types.h"
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-__attribute__((visibility("default"))) wasi_nn_error
-load(void *tflite_ctx, graph_builder_array *builder, graph_encoding encoding,
-     execution_target target, graph *g);
-
-__attribute__((visibility("default"))) wasi_nn_error
-load_by_name(void *tflite_ctx, const char *filename, uint32_t filename_len,
-             graph *g);
-
-__attribute__((visibility("default"))) wasi_nn_error
-init_execution_context(void *tflite_ctx, graph g, graph_execution_context *ctx);
-
-__attribute__((visibility("default"))) wasi_nn_error
-set_input(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
-          tensor *input_tensor);
-
-__attribute__((visibility("default"))) wasi_nn_error
-compute(void *tflite_ctx, graph_execution_context ctx);
-
-__attribute__((visibility("default"))) wasi_nn_error
-get_output(void *tflite_ctx, graph_execution_context ctx, uint32_t index,
-           tensor_data output_tensor, uint32_t *output_tensor_size);
-
-__attribute__((visibility("default"))) wasi_nn_error
-init_backend(void **tflite_ctx);
-
-__attribute__((visibility("default"))) wasi_nn_error
-deinit_backend(void *tflite_ctx);
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif

core/shared/platform/common/posix/posix_file.c

@@ -1034,6 +1034,12 @@ os_realpath(const char *path, char *resolved_path)
     return realpath(path, resolved_path);
 }
 
+os_raw_file_handle
+os_invalid_raw_handle(void)
+{
+    return -1;
+}
+
 // Better to define the function here, as Linux-SGX will
 // use this file to implement the `_os` functions.
 // So we don't need to define them in the Linux-SGX platform.
@@ -1061,9 +1067,3 @@ os_compare_file_handle(os_file_handle handle1, os_file_handle handle2)
 {
     return handle1 == handle2;
 }
-
-os_raw_file_handle
-os_invalid_raw_handle(void)
-{
-    return -1;
-}

core/shared/platform/common/posix/posix_socket.c

@@ -220,12 +220,17 @@ int
 os_socket_accept(bh_socket_t server_sock, bh_socket_t *sock, void *addr,
                  unsigned int *addrlen)
 {
-    *sock = accept(server_sock, addr, addrlen);
+    if (addr == NULL) {
-
+        *sock = accept(server_sock, NULL, NULL);
+    }
+    else {
+        socklen_t len = *addrlen;
+        *sock = accept(server_sock, addr, &len);
+        *addrlen = len;
+    }
     if (*sock < 0) {
         return BHT_ERROR;
     }
-
     return BHT_OK;
 }
 
@@ -373,8 +378,8 @@ is_addrinfo_supported(struct addrinfo *info)
         (info->ai_family == AF_INET || info->ai_family == AF_INET6)
         // Allow only UDP and TCP
         && (info->ai_socktype == SOCK_DGRAM || info->ai_socktype == SOCK_STREAM)
-        && (info->ai_protocol == IPPROTO_TCP
+        && (info->ai_protocol == IPPROTO_TCP || info->ai_protocol == IPPROTO_UDP
-            || info->ai_protocol == IPPROTO_UDP);
+            || info->ai_protocol == 0);
 }
 
 int
@@ -406,12 +411,11 @@ os_socket_addr_resolve(const char *host, const char *service,
 
     res = result;
     while (res) {
-        if (addr_info_size > pos) {
         if (!is_addrinfo_supported(res)) {
             res = res->ai_next;
             continue;
         }
-
+        if (addr_info_size > pos) {
             ret =
                 sockaddr_to_bh_sockaddr(res->ai_addr, &addr_info[pos].sockaddr);
 

core/shared/platform/esp-idf/espidf_file.c

@@ -1034,6 +1034,12 @@ os_realpath(const char *path, char *resolved_path)
     return realpath(path, resolved_path);
 }
 
+os_raw_file_handle
+os_invalid_raw_handle(void)
+{
+    return -1;
+}
+
 int
 os_ioctl(os_file_handle handle, int request, ...)
 {
@@ -1045,9 +1051,3 @@ os_poll(os_poll_file_handle *fds, os_nfds_t nfs, int timeout)
 {
     return BHT_ERROR;
 }
-
-os_raw_file_handle
-os_invalid_raw_handle(void)
-{
-    return -1;
-}

core/shared/platform/esp-idf/espidf_thread.c

@@ -110,6 +110,13 @@ os_thread_create_with_prio(korp_tid *tid, thread_start_routine_t start,
     targ->start = start;
     targ->arg = arg;
 
+#ifdef CONFIG_FREERTOS_TASK_CREATE_ALLOW_EXT_MEM
+    esp_pthread_cfg_t default_config = esp_pthread_get_default_config();
+
+    default_config.stack_alloc_caps = MALLOC_CAP_8BIT | MALLOC_CAP_SPIRAM;
+    ESP_ERROR_CHECK(esp_pthread_set_cfg(&default_config));
+#endif
+
     if (pthread_create(tid, &tattr, os_thread_wrapper, targ) != 0) {
         pthread_attr_destroy(&tattr);
         os_free(targ);

core/shared/platform/windows/win_file.c

@@ -1811,6 +1811,12 @@ os_realpath(const char *path, char *resolved_path)
     return resolved_path;
 }
 
+os_raw_file_handle
+os_invalid_raw_handle(void)
+{
+    return INVALID_HANDLE_VALUE;
+}
+
 bool
 os_compare_file_handle(os_file_handle handle1, os_file_handle handle2)
 {
@@ -1846,9 +1852,3 @@ os_poll(os_poll_file_handle *fds, os_nfds_t nfs, int timeout)
 {
     return BHT_ERROR;
 }
-
-os_raw_file_handle
-os_invalid_raw_handle(void)
-{
-    return INVALID_HANDLE_VALUE;
-}

core/shared/utils/bh_common.c

@@ -59,7 +59,7 @@ b_memcpy_wa(void *s1, unsigned int s1max, const void *s2, unsigned int n)
                 *dest++ = *p_byte_read++;
             }
         }
-        /* read meaning word(s) */
+        /* read remaining word(s) */
         else {
             if ((char *)p + 4 >= src + n) {
                 for (ps = (char *)p; ps < src + n; ps++) {

core/shared/utils/bh_queue.c

@@ -170,7 +170,7 @@ bh_free_msg(bh_queue_node *msg)
         return;
     }
 
-    // note: sometime we just use the payload pointer for a integer value
+    // note: sometimes we just use the payload pointer for an integer value
     //       len!=0 is the only indicator about the body is an allocated buffer.
     if (msg->body && msg->len)
         bh_queue_free(msg->body);

core/shared/utils/bh_vector.c

@@ -35,8 +35,8 @@ extend_vector(Vector *vector, size_t length)
     if (length <= vector->max_elems)
         return true;
 
-    if (length < vector->size_elem * 3 / 2)
+    if (length < vector->max_elems * 3 / 2)
-        length = vector->size_elem * 3 / 2;
+        length = vector->max_elems * 3 / 2;
 
     if (!(data = alloc_vector_data(length, vector->size_elem))) {
         return false;
@@ -194,12 +194,12 @@ bh_vector_append(Vector *vector, const void *elem_buf)
         goto just_return;
     }
 
-    /* make sure one more slot is used by the thread who allocas it */
+    /* make sure one more slot is used by the thread who allocates it */
     if (vector->lock)
         os_mutex_lock(vector->lock);
 
     if (!extend_vector(vector, vector->num_elems + 1)) {
-        LOG_ERROR("Append ector elem failed: extend vector failed.\n");
+        LOG_ERROR("Append vector elem failed: extend vector failed.\n");
         goto unlock_return;
     }
 

core/shared/utils/runtime_timer.c

@@ -44,22 +44,22 @@ bh_get_tick_ms()
 uint32
 bh_get_elpased_ms(uint32 *last_system_clock)
 {
-    uint32 elpased_ms;
+    uint32 elapsed_ms;
-    /* attention: the bh_get_tick_ms() return 64 bits integer, but
+    /* attention: the bh_get_tick_ms() returns a 64-bit integer, but
-       the bh_get_elpased_ms() is designed to use 32 bits clock count */
+       bh_get_elpased_ms() is designed to use a 32-bit clock count */
     uint32 now = (uint32)bh_get_tick_ms();
 
     /* system clock overrun */
     if (now < *last_system_clock) {
         PRINT("system clock overrun!\n");
-        elpased_ms = now + (UINT32_MAX - *last_system_clock) + 1;
+        elapsed_ms = now + (UINT32_MAX - *last_system_clock) + 1;
     }
     else {
-        elpased_ms = now - *last_system_clock;
+        elapsed_ms = now - *last_system_clock;
     }
 
     *last_system_clock = now;
-    return elpased_ms;
+    return elapsed_ms;
 }
 
 static app_timer_t *
@@ -162,7 +162,7 @@ reschedule_timer(timer_ctx_t ctx, app_timer_t *timer)
               prev->id);
     }
     else {
-        /* insert at the begin */
+        /* insert at the beginning */
         bh_assert(ctx->app_timers == NULL);
         ctx->app_timers = timer;
         PRINT("rescheduled timer [%d] as first\n", timer->id);
@@ -213,7 +213,7 @@ release_timer_list(app_timer_t **p_list)
 
 timer_ctx_t
 create_timer_ctx(timer_callback_f timer_handler,
-                 check_timer_expiry_f expiery_checker, int prealloc_num,
+                 check_timer_expiry_f expiry_checker, int prealloc_num,
                  unsigned int owner)
 {
     timer_ctx_t ctx = (timer_ctx_t)BH_MALLOC(sizeof(struct _timer_ctx));
@@ -225,7 +225,7 @@ create_timer_ctx(timer_callback_f timer_handler,
 
     ctx->timer_callback = timer_handler;
     ctx->pre_allocated = prealloc_num;
-    ctx->refresh_checker = expiery_checker;
+    ctx->refresh_checker = expiry_checker;
     ctx->owner = owner;
 
     while (prealloc_num > 0) {

core/version.h

@@ -17,7 +17,7 @@
 
 /* clang-format off */
 #define WAMR_VERSION_MAJOR 2
-#define WAMR_VERSION_MINOR 3
+#define WAMR_VERSION_MINOR 4
 #define WAMR_VERSION_PATCH 1
 /* clang-format on */
 

doc/build_wamr.md

@@ -102,6 +102,7 @@ cmake -DWAMR_BUILD_PLATFORM=linux -DWAMR_BUILD_TARGET=ARM
 
 ### **Enable lib wasi-nn**
 - **WAMR_BUILD_WASI_NN**=1/0, default to disable if not set
+> Note: WAMR_BUILD_WASI_NN without WAMR_BUILD_WASI_EPHEMERAL_NN is deprecated and will likely be removed in future versions of WAMR. Please consider to enable WAMR_BUILD_WASI_EPHEMERAL_NN as well.
 > Note: See [WASI-NN](../core/iwasm/libraries/wasi-nn) for more details.
 
 ### **Enable lib wasi-nn GPU mode**
@@ -113,7 +114,7 @@ cmake -DWAMR_BUILD_PLATFORM=linux -DWAMR_BUILD_TARGET=ARM
 - **WAMR_BUILD_WASI_NN_EXTERNAL_DELEGATE_PATH**=Path to the external delegate shared library (e.g. `libedgetpu.so.1.0` for Coral USB)
 
 ### **Enable lib wasi-nn with `wasi_ephemeral_nn` module support**
-- **WAMR_BUILD_WASI_EPHEMERAL_NN**=1/0, default to disable if not set
+- **WAMR_BUILD_WASI_EPHEMERAL_NN**=1/0, default to enable if not set
 
 ### **Disable boundary check with hardware trap**
 - **WAMR_DISABLE_HW_BOUND_CHECK**=1/0, default to enable if not set and supported by platform
@@ -292,6 +293,10 @@ Currently we only profile the memory consumption of module, module_instance and
 - **WAMR_BUILD_AOT_INTRINSICS**=1/0, enable the AOT intrinsic functions, default to enable if not set. These functions can be called from the AOT code when `--disable-llvm-intrinsics` flag or `--enable-builtin-intrinsics=<intr1,intr2,...>` flag is used by wamrc to generate the AOT file.
 > Note: See [Tuning the XIP intrinsic functions](./xip.md#tuning-the-xip-intrinsic-functions) for more details.
 
+### **Enable extended constant expression**
+- **WAMR_BUILD_EXTENDED_CONST_EXPR**=1/0, default to disable if not set.
+> Note: See [Extended Constant Expressions](https://github.com/WebAssembly/extended-const/blob/main/proposals/extended-const/Overview.md) for more details.
+
 ### **Configurable memory access boundary check**
 - **WAMR_CONFIGURABLE_BOUNDS_CHECKS**=1/0, default to disable if not set
 > Note: If it is enabled, allow to run `iwasm --disable-bounds-checks` to disable the memory access boundary checks for interpreter mode.

doc/build_wasm_app.md

@@ -382,17 +382,26 @@ a non-compatible`AOT_CURRENT_VERSION`.
 We try our best to maintain our runtime ABI for AoT-compiled modules
 compatible among WAMR versions with compatible `AOT_CURRENT_VERSION`
 so that combinations of older wamrc and newer runtime usually work.
-However, there might be minor incompatibilities time to time.
-For productions, we recommend to use compatible versions of
-wamrc and the runtime.
 
-| WAMR version | AOT_CURRENT_VERSION | Compatible AOT version |
+However, there might be minor incompatibilities from time to time. For
-| ------------ | ------------------- | ---------------------- |
+example, we usually avoid bumping the version when making a change which
-| 1.x          | 3                   | 3                      |
+affects only a small fraction of users. For productions, we recommend
-| 2.0.0        | 3                   | 3                      |
+using exactly same versions of wamrc and the runtime.
-| 2.1.x        | 3                   | 3                      |
+
-| 2.2.0        | 3                   | 3                      |
+| WAMR version | AOT_CURRENT_VERSION | Compatible AOT version |                        |
-| next         | 4                   | 3,4                    |
+| ------------ | ------------------- | ---------------------- | ---------------------- |
+| 1.x          | 3                   | 3                      |                        |
+| 2.0.0        | 3                   | 3                      |                        |
+| 2.1.x        | 3                   | 3                      |                        |
+| 2.2.0        | 3                   | 3                      |                        |
+| 2.3.0        | 4                   | 3,4                    |                        |
+| 2.4.0        | 4                   | 3,4                    | See the following note |
+| 2.4.1        | 5                   | 5                      |                        |
+
+Note: 2.4.0 had a broken AoT versioning. See [issue 4504] for details.
+We recommend all 2.4.0 users to migrate to 2.4.1.
+
+[issue 4504]: https://github.com/bytecodealliance/wasm-micro-runtime/issues/4504
 
 ## AoT compilation with 3rd-party toolchains
 

doc/security_issue_runbook.md

@@ -0,0 +1,46 @@
+# Security Issue Runbook
+
+This runbook provides step-by-step guidance on handling a security advisory. Typically, it begins with a draft security advisory when we initiate the process outlined in this runbook. The draft security advisory is created by a contributor or a maintainer.
+
+For information on what types of issues are considered security vulnerabilities and require a security advisory for resolution, please refer to [identifying a security issue](./security_need_to_know.md#identifying-a-security-issue).
+
+## Step 1: Initial Response to Security Advisory
+
+- Receive Security Advisory: When a new security advisory is received, the Incident Manager, typically the maintainer who opened the advisory, becomes the first responder. If the advisory was opened by someone else, a maintainer should take on the role of Incident Manager. The Incident Manager can hand off this role to another maintainer if necessary.
+- Acknowledge Receipt: The Incident Manager should promptly acknowledge receipt of the advisory and communicate that the investigation will begin immediately. Security issues are the highest priority.
+
+## Step 2: Investigating the Vulnerability
+
+- Identify the Vulnerability: Reproduce the issue to understand the vulnerability. Determine which versions and platforms are affected. Fill out the advisory details with this information.
+- Accept the Report: Accept the security report and create a temporary private fork to collaborate on a fix. Invite necessary helpers and stakeholders to this fork, as their input can be valuable.
+
+## Step 3: Communication and Collaboration
+
+- Use Non-Public Channels: Communicate through non-public channels, preferably email, during the resolution process. Avoid filing issues or pull requests on third-party repositories if they are involved.
+- Workaround for Third-Party Dependencies: If third-party dependencies are involved, consider a workaround to patch the issue quickly unless the third party can release a fix promptly.
+
+## Step 4: Finalizing and Preparing for Release
+
+- Finalize Details: Once a fix is developed and the vulnerability is fully understood, finalize the advisory details and prepare for public release. Ensure the security issues are resolved in the private fork.
+- Request CVE: Use the Big Green Button on the advisory to request a CVE number from GitHub staff.
+- Advanced Disclosure Email: Decide on a disclosure date, typically within a week, and send an email to sec-announce@bytecodealliance.org about the upcoming security release. Other ways are also available to communicate the disclosure date.
+
+## Step 5: Preparing and Testing Patch Releases
+
+- Prepare PRs for Patch Releases: Create pull requests in the private fork for each version being patched. Ensure each PR is ready to apply cleanly and includes release notes for each release branch.
+- Run Full Test Suite: Run the full test suite locally for the main branch. Attempt to run as much of the CI matrix locally as possible.
+
+## Step 6: Public Release and Communication
+
+- Open Version Bump PRs: Open version bump pull requests on the public repository without including patch notes or release notes for the fix.
+- Manually Make PRs from Private Fork: Transfer the necessary pull requests from the private fork to the public repository.
+- Merge and Trigger Releases: Merge the version bump PRs and trigger the release process.
+- Publish GitHub Advisories: Delete the private forks and use the Big Green Button to publish the advisory.
+- Send Security Release Email: Send a follow-up email to sec-announce@bytecodealliance.org describing the security release. Other communication channels can also be used to inform users about the security release.
+
+By following these steps, you can effectively manage and resolve security issues for your open source project, ensuring timely communication and collaboration while maintaining the integrity and security of your software.
+
+## References
+
+- [Vulnerability Response Runbook](https://github.com/bytecodealliance/rfcs/blob/main/accepted/vulnerability-response-runbook.md)
+- [Wasmtime Security Vulnerability Runbook](https://docs.wasmtime.dev/security-vulnerability-runbook.html)

doc/security_need_to_know.md

@@ -30,4 +30,4 @@ Before reporting an issue, particularly one related to crashing, consult [the ch
 
 Upon receiving an issue, thoroughly review [the cheat sheet](https://github.com/bytecodealliance/rfcs/blob/main/accepted/what-is-considered-a-security-bug.md#cheat-sheet-is-this-bug-considered-a-security-vulnerability) to assess and _Report a security vulnerability_ if the issue is indeed a security vulnerability.
 
-Once a security issue is confirmed, please refer to [the runbook](https://github.com/bytecodealliance/rfcs/blob/main/accepted/vulnerability-response-runbook.md) for the subsequent steps to take.
+Once a security issue is confirmed, please refer to [the runbook](./security_issue_runbook.md) for the subsequent steps to take.

doc/socket_api.md

@@ -58,7 +58,8 @@ enabled.
 
 _iwasm_ accepts address ranges via an option, `--addr-pool`, to implement
 the capability control. All IP address the WebAssembly application may need to `bind()` or `connect()`
-should be announced first. Every IP address should be in CIDR notation.
+should be announced first. Every IP address should be in CIDR notation. If not, _iwasm_ will return
+an error.
 
 ```bash
 $ iwasm --addr-pool=1.2.3.4/15,2.3.4.6/16 socket_example.wasm
@@ -87,3 +88,38 @@ $ iwasm --addr-pool=1.2.3.4/15,2.3.4.6/16 socket_example.wasm
 ```
 
 Refer to [socket api sample](../samples/socket-api) for the compilation of the Wasm applications and [_iwasm_ for Intel SGX](../product-mini/platforms/linux-sgx) for the Wasm runtime.
+
+## The background and compatibility notes
+
+### WASIp1
+
+The WASIp1 provides a subset of the socket API.
+Namely,
+
+* send()
+* recv()
+* shutdown()
+* accept()
+
+Functionalities like connect() and listen() are intentionally omitted
+there to maintain the capability-based security model, inherited from
+cloudabi. The common practice for applications is to make the host code
+pass already connected/listening sockets to wasm module.
+
+### WAMR extensions
+
+WAMR extends the WASIp1 with the rest of socket API functionalities
+for convenience.
+
+* socket()
+* connect()
+* bind()
+* listen()
+* some of getsockopt/setsockopt options
+* name resolution (a subset of getaddrinfo)
+
+### Compatibilities
+
+Many of runtimes (eg. Wasmer and WasmEdge) provide similar extensions.
+Unfortunately, they are all incompatible. Thus, portable applications
+should not rely on these extensions.

product-mini/README.md

@@ -111,7 +111,7 @@ The Fast JIT is a lightweight JIT engine with quick startup, small footprint and
 (6) To enable the `Multi-tier JIT` mode:
 ``` Bash
 mkdir build && cd build
-cmake .. -DWAMR_BUILD_FAST_JTI=1 -DWAMR_BUILD_JIT=1
+cmake .. -DWAMR_BUILD_FAST_JIT=1 -DWAMR_BUILD_JIT=1
 make
 ```
 The Multi-tier JIT is a two level JIT tier-up engine, which launches Fast JIT to run the wasm module as soon as possible and creates backend threads to compile the LLVM JIT functions at the same time, and when the LLVM JIT functions are compiled, the runtime will switch the extecution from the Fast JIT jitted code to LLVM JIT jitted code gradually, so as to gain the best performance.

product-mini/platforms/common/libc_wasi.c

@@ -45,7 +45,7 @@ libc_wasi_print_help(void)
            "path, for example:\n");
     printf("                             --map-dir=<guest-path1::host-path1> "
            "--map-dir=<guest-path2::host-path2>\n");
-    printf("  --addr-pool=<addrs>      Grant wasi access to the given network "
+    printf("  --addr-pool=<addr/mask>  Grant wasi access to the given network "
            "addresses in\n");
     printf("                           CIDR notation to the program, separated "
            "with ',',\n");

product-mini/platforms/posix/main.c

@@ -57,6 +57,10 @@ print_help(void)
 #else
     printf("  --heap-size=n            Set maximum heap size in bytes, default is 16 KB when libc wasi is diabled\n");
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+    printf("  --shared-heap-size=n     Create shared heap of n bytes and attach to the wasm app.\n");
+    printf("                           The size n will be adjusted to a minumum number aligned to page size\n");
+#endif
 #if WASM_ENABLE_FAST_JIT != 0
     printf("  --jit-codecache-size=n   Set fast jit maximum code cache size in bytes,\n");
     printf("                           default is %u KB\n", FAST_JIT_DEFAULT_CODE_CACHE_SIZE / 1024);
@@ -578,6 +582,11 @@ main(int argc, char *argv[])
 #else
     uint32 heap_size = 16 * 1024;
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+    SharedHeapInitArgs heap_init_args;
+    uint32 shared_heap_size = 0;
+    void *shared_heap = NULL;
+#endif
 #if WASM_ENABLE_FAST_JIT != 0
     uint32 jit_code_cache_size = FAST_JIT_DEFAULT_CODE_CACHE_SIZE;
 #endif
@@ -596,6 +605,7 @@ main(int argc, char *argv[])
     wasm_module_inst_t wasm_module_inst = NULL;
     RunningMode running_mode = 0;
     RuntimeInitArgs init_args;
+    struct InstantiationArgs2 *inst_args;
     char error_buf[128] = { 0 };
 #if WASM_ENABLE_LOG != 0
     int log_verbose_level = 2;
@@ -684,6 +694,13 @@ main(int argc, char *argv[])
                 return print_help();
             heap_size = atoi(argv[0] + 12);
         }
+#if WASM_ENABLE_SHARED_HEAP != 0
+        else if (!strncmp(argv[0], "--shared-heap-size=", 19)) {
+            if (argv[0][19] == '\0')
+                return print_help();
+            shared_heap_size = atoi(argv[0] + 19);
+        }
+#endif
 #if WASM_ENABLE_FAST_JIT != 0
         else if (!strncmp(argv[0], "--jit-codecache-size=", 21)) {
             if (argv[0][21] == '\0')
@@ -949,10 +966,20 @@ main(int argc, char *argv[])
     libc_wasi_init(wasm_module, argc, argv, &wasi_parse_ctx);
 #endif
 
+    if (!wasm_runtime_instantiation_args_create(&inst_args)) {
+        printf("failed to create instantiate args\n");
+        goto fail3;
+    }
+    wasm_runtime_instantiation_args_set_default_stack_size(inst_args,
+                                                           stack_size);
+    wasm_runtime_instantiation_args_set_host_managed_heap_size(inst_args,
+                                                               heap_size);
+
     /* instantiate the module */
-    if (!(wasm_module_inst =
+    wasm_module_inst = wasm_runtime_instantiate_ex2(
-              wasm_runtime_instantiate(wasm_module, stack_size, heap_size,
+        wasm_module, inst_args, error_buf, sizeof(error_buf));
-                                       error_buf, sizeof(error_buf)))) {
+    wasm_runtime_instantiation_args_destroy(inst_args);
+    if (!wasm_module_inst) {
         printf("%s\n", error_buf);
         goto fail3;
     }
@@ -996,6 +1023,24 @@ main(int argc, char *argv[])
     }
 #endif
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+    if (shared_heap_size > 0) {
+        memset(&heap_init_args, 0, sizeof(heap_init_args));
+        heap_init_args.size = shared_heap_size;
+        shared_heap = wasm_runtime_create_shared_heap(&heap_init_args);
+        if (!shared_heap) {
+            printf("Create preallocated shared heap failed\n");
+            goto fail6;
+        }
+
+        /* attach module instance 2 to the shared heap */
+        if (!wasm_runtime_attach_shared_heap(wasm_module_inst, shared_heap)) {
+            printf("Attach shared heap failed.\n");
+            goto fail6;
+        }
+    }
+#endif
+
     ret = 0;
     const char *exception = NULL;
     if (is_repl_mode) {
@@ -1039,6 +1084,9 @@ main(int argc, char *argv[])
     }
 #endif
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+fail6:
+#endif
 #if WASM_ENABLE_THREAD_MGR != 0
 fail5:
 #endif

product-mini/platforms/windows/main.c

@@ -53,6 +53,10 @@ print_help()
     printf("  --gc-heap-size=n         Set maximum gc heap size in bytes,\n");
     printf("                           default is %u KB\n", GC_HEAP_SIZE_DEFAULT / 1024);
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+    printf("  --shared-heap-size=n     Create shared heap of n bytes and attach to the wasm app.\n");
+    printf("                           The size n will be adjusted to a minumum number aligned to page size\n");
+#endif
 #if WASM_ENABLE_JIT != 0
     printf("  --llvm-jit-size-level=n  Set LLVM JIT size level, default is 3\n");
     printf("  --llvm-jit-opt-level=n   Set LLVM JIT optimization level, default is 3\n");
@@ -364,6 +368,13 @@ main(int argc, char *argv[])
             gc_heap_size = atoi(argv[0] + 15);
         }
 #endif
+#if WASM_ENABLE_SHARED_HEAP != 0
+        else if (!strncmp(argv[0], "--shared-heap-size=", 19)) {
+            if (argv[0][19] == '\0')
+                return print_help();
+            shared_heap_size = atoi(argv[0] + 19);
+        }
+#endif
 #if WASM_ENABLE_JIT != 0
         else if (!strncmp(argv[0], "--llvm-jit-size-level=", 22)) {
             if (argv[0][22] == '\0')
@@ -551,6 +562,23 @@ main(int argc, char *argv[])
         goto fail3;
     }
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+    if (shared_heap_size > 0) {
+        memset(&shared_heap_init_args, 0, sizeof(shared_heap_init_args));
+        shared_heap_init_args.size = shared_heap_size;
+        shared_heap = wasm_runtime_create_shared_heap(&shared_heap_init_args);
+
+        if (!shared_heap) {
+            printf("Create shared heap failed.\n");
+            goto fail5;
+        }
+        if (!wasm_runtime_attach_shared_heap(wasm_module_inst, shared_heap)) {
+            printf("Attach shared heap failed.\n");
+            goto fail5;
+        }
+    }
+#endif
+
 #if WASM_ENABLE_DEBUG_INTERP != 0
     if (ip_addr != NULL) {
         wasm_exec_env_t exec_env =
@@ -598,6 +626,9 @@ main(int argc, char *argv[])
     if (exception)
         printf("%s\n", exception);
 
+#if WASM_ENABLE_SHARED_HEAP != 0
+fail5:
+#endif
 #if WASM_ENABLE_DEBUG_INTERP != 0
 fail4:
 #endif

product-mini/platforms/zephyr/simple/Dockerfile

@@ -3,7 +3,23 @@
 
 # Refer to https://docs.zephyrproject.org/3.7.0/develop/getting_started/index.html
 # for more information on how to set up the Zephyr development environment.
-FROM ubuntu:22.04
+
+# https://docs.zephyrproject.org/latest/develop/application/index.html#zephyr-workspace-application
+# zephyrproject/       --> CI ROOT
+# ├─── .west/
+# │    └─── config
+# ├─── bootloader/
+# ├─── zephyr/         --> Zephyr source code
+# ├─── zephyr-sdk/
+# ├─── modules/
+# │    |─── wasm-micro-runtime   --> WAMR source code
+# ├─── tools/
+# ├─── vendor/
+# └─── application/    --> DUMMY. keep west_lite.yml here
+
+# If you modify this file, you may need to sync the modifications to the
+# .github/actions/setup-zephyr/action.yml
+FROM ghcr.io/zephyrproject-rtos/ci-base:v0.26-branch
 
 ARG DEBIAN_FRONTEND=noninteractive
 ENV TZ=Asian/Shanghai
@@ -11,44 +27,31 @@ ARG ZEPHYR_SDK_VERSION=0.16.9
 # In west_lite.yml, the Zephyr version is set to v3.7.0
 #ARG ZEPHYR_VERSION=3.7.0
 
-# Install dependencies for Zephyr
+# Install the Zephyr Software Development Kit (SDK) minimal version
-# hadolint ignore=DL3008
+WORKDIR /root/zephyrproject/zephyr-sdk
-RUN apt-get update && apt-get install -y --no-install-recommends git cmake ninja-build gperf \
-    ccache dfu-util device-tree-compiler wget \
-    python3-dev python3-pip python3-setuptools python3-tk python3-wheel xz-utils file \
-    make gcc gcc-multilib g++-multilib libsdl2-dev libmagic1 \ 
-    && apt-get clean -y && rm -rf /var/lib/apt/lists/*
-
-# Install the Zephyr Software Development Kit (SDK)
-WORKDIR /opt
 # hadolint ignore=DL4006
-RUN wget --progress=dot:giga https://github.com/zephyrproject-rtos/sdk-ng/releases/download/v${ZEPHYR_SDK_VERSION}/zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64.tar.xz \
+RUN wget --progress=dot:giga https://github.com/zephyrproject-rtos/sdk-ng/releases/download/v${ZEPHYR_SDK_VERSION}/zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64_minimal.tar.xz \
     && wget --progress=dot:giga -O - https://github.com/zephyrproject-rtos/sdk-ng/releases/download/v${ZEPHYR_SDK_VERSION}/sha256.sum | shasum --check --ignore-missing \
-    && tar xf zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64.tar.xz && rm zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64.tar.xz
+    && tar --strip-components=1 -xf zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64_minimal.tar.xz && rm zephyr-sdk-${ZEPHYR_SDK_VERSION}_linux-x86_64_minimal.tar.xz
-
-WORKDIR /opt/zephyr-sdk-${ZEPHYR_SDK_VERSION}
 # hadolint ignore=DL4006
-# Install host tools and Register Zephyr SDK CMake package
+# Install arc tools, host tools and Register Zephyr SDK CMake package
-RUN ./setup.sh -h -c
+# If you want to use other toolchains, please change the -t option
+RUN ./setup.sh -t arc-zephyr-elf -h -c
 
-# Get Zephyr 
+# Install west
-WORKDIR /root/zephyrproject/smoke-test
+# hadolint ignore=DL3013,DL3059
-
-# hadolint ignore=DL3013
 RUN pip3 install --no-cache-dir west
-COPY ./west_lite.yml ./west.yml
+
+# Setup a T2(Star topology) workspace
+WORKDIR /root/zephyrproject/application
+COPY ./west_lite.yml ./west_lite.yml
 
 # init the west workspace with a minimal manifest
-RUN west init -l
+RUN west init -l --mf west_lite.yml .
 
 WORKDIR /root/zephyrproject
 RUN west update --stats
 
-WORKDIR /root/zephyrproject/modules/zephyr
-RUN west zephyr-export && pip install --no-cache-dir -r ./scripts/requirements.txt
-
-ENV ZEPHYR_BASE="/root/zephyrproject/modules/zephyr"
-
 # Git clone wamr
 WORKDIR /root/zephyrproject/modules/
 RUN git clone https://github.com/bytecodealliance/wasm-micro-runtime.git wasm-micro-runtime

product-mini/platforms/zephyr/simple/west_lite.yml

@@ -8,8 +8,8 @@ manifest:
       url: https://github.com/zephyrproject-rtos/zephyr
       revision: v3.7.0
       clone-depth: 1
-      path: modules/zephyr
+      path: zephyr
       west-commands: scripts/west-commands.yml
 
   self:
-    path: smoke-test
+    path: application

samples/shared-heap/CMakeLists.txt

@@ -83,17 +83,21 @@ target_link_libraries(vmlib ${LLVM_AVAILABLE_LIBS} ${UV_A_LIBS} -lm -ldl -lpthre
 include_directories(${CMAKE_CURRENT_LIST_DIR}/src)
 include (${SHARED_DIR}/utils/uncommon/shared_uncommon.cmake)
 
+add_executable (shared_heap_chain_test src/shared_heap_chain.c ${UNCOMMON_SHARED_SOURCE})
 add_executable (shared_heap_test src/main.c ${UNCOMMON_SHARED_SOURCE})
 
 check_pie_supported()
 set_target_properties (shared_heap_test PROPERTIES POSITION_INDEPENDENT_CODE ON)
 
 if (APPLE)
-  target_link_libraries (shared_heap_test vmlib -lm -ldl -lpthread)
+  set (LIBS vmlib -lm -ldl -lpthread)
 else ()
-  target_link_libraries (shared_heap_test vmlib -lm -ldl -lpthread -lrt)
+  set (LIBS vmlib -lm -ldl -lpthread -lrt)
 endif ()
 
+target_link_libraries (shared_heap_chain_test ${LIBS})
+target_link_libraries (shared_heap_test ${LIBS})
+
 add_subdirectory(wasm-apps)
 
 if (WAMR_BUILD_AOT EQUAL 1)
@@ -107,21 +111,31 @@ if (WAMR_BUILD_AOT EQUAL 1)
   )
   if (WAMR_COMPILER)
     message (CHECK_PASS "found")
-  else()
+  else ()
     message (CHECK_FAIL "not found")
-  endif()
+  endif ()
   if (NOT EXISTS ${WAMR_COMPILER})
     message (FATAL_ERROR "Please build wamrc under ${WAMR_ROOT_DIR}/wamr-compiler")
-  else()
+  else ()
     message (STATUS "WAMR_COMPILER is ${WAMR_COMPILER}")
-  endif()
+  endif ()
+
+  if (WAMR_BUILD_TARGET STREQUAL "X86_32")
+    set (WAMR_COMPILER_FLAGS --enable-shared-heap --target=i386)
+    set (WAMR_COMPILER_CHAIN_FLAGS --enable-shared-chain --target=i386)
+  else ()
+    set (WAMR_COMPILER_FLAGS --enable-shared-heap)
+    set (WAMR_COMPILER_CHAIN_FLAGS --enable-shared-chain)
+  endif ()
 
   add_custom_target(
     wasm_to_aot
     ALL
     DEPENDS wasm-apps/test1.wasm wasm-apps/test2.wasm ${WAMR_COMPILER}
-    COMMAND ${WAMR_COMPILER} --enable-shared-heap -o wasm-apps/test1.aot wasm-apps/test1.wasm
+    COMMAND ${WAMR_COMPILER} ${WAMR_COMPILER_FLAGS} -o wasm-apps/test1.aot wasm-apps/test1.wasm
-    COMMAND ${WAMR_COMPILER} --enable-shared-heap -o wasm-apps/test2.aot wasm-apps/test2.wasm
+    COMMAND ${WAMR_COMPILER} ${WAMR_COMPILER_FLAGS} -o wasm-apps/test2.aot wasm-apps/test2.wasm
+    COMMAND ${WAMR_COMPILER} ${WAMR_COMPILER_CHAIN_FLAGS} -o wasm-apps/test1_chain.aot wasm-apps/test1.wasm
+    COMMAND ${WAMR_COMPILER} ${WAMR_COMPILER_CHAIN_FLAGS} -o wasm-apps/test2_chain.aot wasm-apps/test2.wasm
     WORKING_DIRECTORY ${CMAKE_BINARY_DIR}
   )
 endif()

samples/shared-heap/README.md

@@ -0,0 +1,50 @@
+# Shared heap Sample introduction
+
+This is a sample to show how to use the shared heap feature in WAMR. The shared heap feature allows multiple WASM instances to share the same memory space. This feature is useful when you want to run multiple WASM instances in the same process and share data between them. The sandbox nature of WASM is still maintained in the shared heap by WAMR. But the data management and correct data synchronization in shared heap is relied on the user's implementation.
+
+> Note: The shared heap feature is experimental feature, it should be used with caution. It's optional and only available when building WAMR with the CMake cache variable `WAMR_BUILD_SHARED_HEAP` set to 1.
+
+## Build and run the sample
+
+To build the shared heap used in multi thread sample and the shared heap chain sample with following commands:
+
+```bash
+cmake -S . -B build
+cmake --build build
+```
+
+For the shared heap sample, it demonstrates how to create a shared heap and use it shares data between two WASM instances, which would satisfy most of the use cases. Use the following commands to run the sample:
+
+```bash
+cd build
+./shared_heap_test
+```
+
+For the shared heap chain sample. It chains a pre-allocated heap and a normal shared heap to one chain(linked list) as a whole and attaches/detaches all together, and pass the WASM address directly between two WASM instances. Use the following commands to run the sample:
+
+```bash
+cd build
+./shared_heap_chain_test
+```
+
+## How to use shared heap
+
+The shared heap is an advanced feature in WAMR that gives the user flexibility to share data between multiple WASM instances(it will be the same address mapping for different WASM instance) or between WebAssembly and the host without incurring any copy overhead. The shared heap can be regarded as an extension of the WebAssembly linear memory. But it also heavily relies on the user's implementation to manage the shared data correctly. The following are some takeaway points to help the user use the shared heap correctly.
+
+### Create and manage shared heap
+
+You can create a shared heap by calling the `wasm_runtime_create_shared_heap(SharedHeapInitArgs *init_args)` API. And based on the `init_args`, you can create a shared heap in two ways:
+
+1. WAMR managed shared heap: when only `init_args.size` is given and `init_args.pre_allocated_addr` stays as NULL, WAMR will allocate a shared heap(not from the linear memory) with the given size. The shared heap will be managed by WAMR, the wasm app or host(WAMR users) can dynamically manage memory from it by calling `wasm_runtime_shared_heap_malloc()` and `wasm_runtime_shared_heap_free()` on demand. Only the memory allocated from the shared heap is valid and can be shared, not the unallocated part of shared heap memory. And it will be automatically freed when runtime is destroyed(when `wasm_runtime_destroy()` is called).
+
+2. Preallocated shared heap: the user can also use a pre-allocated memory(it can be allocated from the system heap, or is a static global buffer, the correctness of its accessibility and size needs to be ensured by the user) as a shared heap by giving `init_args.pre_allocated_addr` and `init_args.size`. This kind of shared heap serves as an area for data exchange, primarily between the host and WebAssembly. Any data within this area can be directly accessed by both sides (assuming the layout of the data structure is known). For instance, the host can store large structured variables in this space, allowing the WebAssembly application to operate on them without the need for copying. And the pre-allocated memory will relies on user to manage its life cycle.
+
+After creation, the shared heap can be attached to a WASM instance(an additional segment appended to the end of the linear memory) by calling `wasm_runtime_attach_shared_heap(wasm_module_inst_t module_inst, wasm_shared_heap_t shared_heap)`. And it can be detached by calling `wasm_runtime_detach_shared_heap(wasm_module_inst_t module_inst)`. So that the data sharing can only happen between the WASM instances that have the same shared heap attached, complete by user's choice.
+
+#### Shared heap chain
+
+Sometimes you may want to use multiple shared heaps to attach together as a chain(linked list) and to share data more flexibly. You can call `wasm_runtime_chain_shared_heaps(wasm_shared_heap_t head, wasm_shared_heap_t body)` to chain two shared heaps together. The shared heap list remains one continuous shared heap in wasm app's point of view. To create a shared heap chain, the shared heaps can't be currently attached to any WASM instance.
+
+> PS: At most one shared heap in shared heap list can be  WAMR managed shared heap, the rest have to be the pre-allocated shared heap.
+
+![shared-heap-chain](./images/shared_heap_chain.png)