/llvm-msvc-ex

Forked LLVM focused on MSVC Compatibility. This version is designed for windows users

Primary LanguageLLVMGNU Affero General Public License v3.0AGPL-3.0

llvm-msvc

add Something for ollvm

Original obfuscation passes including bogus control flow (-bcf), split basic block (-split), control flow flattening (-fla) and substitution (-sub) and MBA-substitution(-mba-subs) and Indirect-Call(-ind-call) and StringXor(-string-obfus) and ConstVarXor(-const-obfus) and VM-Flattening (-vm-fla).

感谢

https://github.com/gmh5225, 日天同学的llvm-msvc让人心情激动

参考

https://github.com/gmh5225/awesome-llvm-security#ollvm

更新

放弃TAG自动Build了，自己build吧

计划

  • 在vm-fla-sym添加反符号执行和反内存追踪
  • vm-fla-enc 对vm-fla的部分数据加密
  • [ ] MBA-subs的bug
  • [ ] 移植xVMP
  • 在vm-fla-enc中使用间接全局变量访问
  • vm-fla-level 0~7 8个处理等级 7最强，0最弱，默认7
  • 弱鸡vmp加入
  • 添加combine功能
  • 添加fla強化 x-fla-enh
  • x-full 功能，在function上使用vm-fla-level=7
  • 字符串加密等相似加了combine
  • 自定義分割合併 combine_func[tag_number] 模式
  • [ ] x-var-rot 待处理
  • 新功能
  • new functions

windows-llvm-msvc-build android-llvm-msvc-build macos-arm64-llvm-msvc-build

image Github All Releases GitHub release

llvm-msvc is a compiler based on LLVM that isn't limited by MSVC. The aim is to provide the same experience as MSVC on Windows. You can use naked functions anywhere and also add custom support like obfuscation.

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣀⣀⣀⣀⣀⣠⣼⠂⠀⠀⠀⠀⠙⣦⢀⠀⠀⠀⠀⠀⢶⣤⣀⣀⣀⣀⣀⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣠⣴⣶⣿⣿⣿⣿⣿⣿⣿⣿⠷⢦⠀⣹⣶⣿⣦⣿⡘⣇⠀⠀⠀⢰⠾⣿⣿⣿⣟⣻⣿⣿⣿⣷⣦⣄⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢀⣤⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿⡟⠀⠀⠀⠀⢺⣿⣿⣿⣿⣿⣿⣿⣆⠀⠀⠀⠀⠀⠀⢹⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⣦⡀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⢀⣴⢟⣥⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡇⠀⠀⠀⠀⢻⣿⣿⡏⢹⣿⣿⣿⣿⠀⠀⠀⠀⠀⠀⠀⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣮⣝⢷⣄⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⣴⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⢛⣿⣿⣿⡇⠀⠀⠀⠀⠛⣿⣿⣷⡀⠘⢿⣧⣻⡷⠀⠀⠀⠀⠀⠀⣿⣿⣿⣟⢿⣿⣿⣿⣿⣿⣿⣿⣿⣝⢧⡀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⢠⣾⣿⠟⣡⣾⣿⣿⣧⣿⡿⣋⣴⣿⣿⣿⣿⣧⠀⠀⠀⠀⠀⢻⣿⣿⣿⣶⡄⠙⠛⠁⠀⠀⠀⠀⠀⢸⣿⣿⣿⣿⣷⣝⢻⣿⣟⣿⣿⣷⣮⡙⢿⣽⣆⠀⠀⠀⠀⠀
⠀⠀⠀⠀⢀⡿⢋⣴⣿⣿⣿⣿⣿⣼⣯⣾⣿⣿⡿⣻⣿⣿⣿⣦⠀⠀⠀⠀⢀⣹⣿⣿⣿⣿⣶⣤⠀⠀⠀⠀⠀⣰⣿⣿⣿⣿⠻⣿⣿⣿⣮⣿⣿⣿⣿⣿⣿⣦⡙⢿⣇⠀⠀⠀⠀
⠀⠀⠀⣠⡏⣰⣿⣿⡿⢿⣿⣿⣿⣿⣿⣿⡿⢋⣼⣿⣿⣿⣿⣿⣷⡤⠀⣠⣿⣿⣿⣿⣿⣿⣿⣿⣷⣄⠀⢠⣾⣿⣿⣿⣿⣿⣷⡜⢿⣿⣿⣿⣿⣿⣿⡿⠿⣿⣿⣦⡙⣦⠀⠀⠀
⠀⠀⣰⢿⣿⣿⠟⠋⣠⣾⣿⣿⣿⣿⣿⠛⢡⣾⡿⢻⣿⣿⣿⣿⣿⣿⣿⣿⡿⠋⠻⣿⡟⣿⣿⣿⠻⢿⣿⣿⣿⣿⣿⣿⣿⣟⠻⣿⣆⠙⢿⣿⣿⣿⣿⣿⣦⡈⠻⣿⣿⣟⣧⠀⠀
⠀⣰⢣⣿⡿⠃⣠⡾⠟⠁⠀⣸⣿⡟⠁⢀⣿⠋⢠⣿⡏⣿⣿⣿⣿⣿⢿⠁⢀⣠⣴⢿⣷⣿⣿⣿⠀⠀⠽⢻⣿⣿⣿⣿⡼⣿⡇⠈⢿⡆⠀⠻⣿⣧⠀⠈⠙⢿⣆⠈⠻⣿⣎⢧⠀
⠀⢣⣿⠟⢀⡼⠋⠀⠀⢀⣴⠿⠋⠀⠀⣾⡟⠀⢸⣿⠙⣿⠃⠘⢿⡟⠀⣰⢻⠟⠻⣿⣿⣿⣿⣿⣀⠀⠀⠘⣿⠋⠀⣿⡇⣿⡇⠀⠸⣿⡄⠀⠈⠻⣷⣄⠀⠀⠙⢷⡀⠙⣿⣆⠁
⢀⣿⡏⠀⡞⠁⢀⡠⠞⠋⠁⠀⠀⠀⠈⠉⠀⠀⠀⠿⠀⠈⠀⠀⠀⠀⠀⣿⣿⣰⣾⣿⣿⣿⣿⣿⣿⣤⠀⠀⠀⠀⠀⠉⠀⠸⠃⠀⠀⠈⠋⠀⠀⠀⠀⠙⠳⢤⣀⠀⠹⡄⠘⣿⡄
⣸⡟⠀⣰⣿⠟⠋⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠛⠿⠿⠿⠟⠁⠀⠹⣿⣷⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣧⠀⢹⣷
⣿⠃⢠⡿⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢠⣄⣤⣀⠀⠀⣿⣿⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢻⡇⠀⣿
⣿⠀⢸⠅⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣰⡿⠋⠉⢻⣧⢀⣿⡇⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣿⠀⢸
⡇⠀⠈⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢿⣧⡀⠀⠀⣿⣾⡟⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠀⢸
⢸⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⠿⣿⣿⠟⠋⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⡾
⠈⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣰⡿⠋⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠃
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢰⡏⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⣧⢀⣾⣤⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀
⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣀⡼⣿⣿⣾⣤⣠⡼⠀⠀⠀

Features:

  • Compatible with MSVC syntax as much as possible.
  • Improved SEH stability.
  • Added some special Intrinsic functions(__vmx_vmread/__vmx_write).
  • Supports x64/ARM64 windows drivers.
  • Supports AArch64 android GKI drivers.
  • Allows naked X64 inline asm.
  • Enables multiple cores compilation.
  • Supports /MP when precompiled headers are present.
  • Supports /GL (LTO optimization).

FAQ

Why do we make this project?

  • Clang follows the GCC standard, while MSVC has its own unique syntax.
  • Some of the code is pretty hacky. Can't submit it officially.
  • Waiting for the official fix takes too long.

How to use llvm-msvc?

  • Install Visual Studio 2015-2022 (I recommend using 2022) along with WDK11.
  • Download and install llvm-msvc from the following repository llvm-msvc_X86_64_installer-PDB.exe.
  • Create your Windows driver project and choose "LLVM-MSVC_v143_KernelMode" as your platform toolset.
  • Compile your project.

How to compile?

X86：clang+lld+RelWithDebInfo

mkdir build-RelWithDebInfo-64
pushd build-RelWithDebInfo-64
cmake .. -G "Visual Studio 17 2022" -A X64 -DCMAKE_CXX_FLAGS="/utf-8" -DCMAKE_C_FLAGS="/utf-8" -DLLDB_ENABLE_PYTHON=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_ENABLE_PROJECTS="clang;lld;lldb" -DCMAKE_INSTALL_PREFIX=E:\llvm\install-RelWithDebInfo-64 -DLLVM_ENABLE_LIBXML2=OFF -DLLVM_ENABLE_ZLIB=OFF -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_OBFUSCATION_LINK_INTO_TOOLS=ON -DCMAKE_BUILD_TYPE=RelWithDebInfo -DLLVM_USE_CRT_RELEASE=MT ../llvm
msbuild /m -p:Configuration=RelWithDebInfo INSTALL.vcxproj 

X86：clang+lld+release

mkdir build-release-64
pushd build-release-64
cmake .. -G "Visual Studio 17 2022" -A X64 -DCMAKE_CXX_FLAGS="/utf-8" -DCMAKE_C_FLAGS="/utf-8" -DLLDB_ENABLE_PYTHON=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_ENABLE_PROJECTS="clang;lld;lldb" -DCMAKE_INSTALL_PREFIX=E:\llvm\install-release-64 -DLLVM_ENABLE_LIBXML2=OFF -DLLVM_ENABLE_ZLIB=OFF -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_OBFUSCATION_LINK_INTO_TOOLS=ON -DCMAKE_BUILD_TYPE=release -DLLVM_USE_CRT_RELEASE=MT ../llvm
msbuild /m -p:Configuration=release INSTALL.vcxproj

混淆例子

Add To VS Project Compiler Cmdline

set /GL off

set /O2 on

最大保护（文件将超过100MB）

-mllvm -data-obfus -mllvm -const-obfus -mllvm -string-obfus -mllvm -ind-call -mllvm -vm-fla -mllvm -fla -mllvm -sub -mllvm -sub_loop=1 -mllvm -split -mllvm -split_num=3 -mllvm -bcf -mllvm -bcf_loop=1 -mllvm -bcf_prob=40 -mllvm -vm-fla-level=7 -mllvm -x-fla-enh -mllvm -x-var-rot -mllvm -x-combine

单纯使用特色部分（轻量模式 不加Light会导致文件飞升到10MB）

-mllvm -data-obfus -mllvm -const-obfus -mllvm -string-obfus -mllvm -ind-call -mllvm -vm-fla -mllvm -vm-fla-level=0 -mllvm -x-fla-enh -mllvm -x-combine -mllvm -x-linear

需要修改载研究的部分

vm sample and x-full sample

__attribute((__annotate__(("x-vm,x-full,x-cfg")))) void crypt_func1(uint8_t *var,uint8_t*key,size_t var_size,size_t key_size){
    for(auto i=0;i<var_size;i++){
        var[i]^=key[i%key_size];
    }
}
__attribute((__annotate__(("x-cfg,ind-br,alias-access")))) void crypt_func2(uint8_t *var,uint8_t*key,size_t var_size,size_t key_size){
    for(auto i=0;i<var_size;i++){
        var[i]^=key[i%key_size];
    }
}
__attribute((__annotate__(("x-cfg,x-vm,ind-br,alias-access")))) void crypt_func3(uint8_t *var,uint8_t*key,size_t var_size,size_t key_size){
    for(auto i=0;i<var_size;i++){
        var[i]^=key[i%key_size];
    }
}

combine sample

__attribute((__annotate__(("combine_func[tag1]")))) int a1(int a, int b)
{
    printf("%d , %d\r\n", a, b);
    printf("%x\r\n", a ^ b);
    return a + b;
}


__attribute((__annotate__(("combine_func[tag1]")))) int a2(int a, int b)
{
    std::cout << "hello1" << std::endl;
    for (auto i = std::min(a, b);i < std::max(a, b);i++)
    {
        printf("%x,", i);
    }
    printf("\n");
   
    return a * b+ a1(a, b);
}


__attribute((__annotate__(("combine_func[tag2]")))) int a3(int a,int b)
{
    printf("%d , %d\r\n", a+1, b+2);
    printf("%x\r\n", a ^ b);
    return a + b+a^b+ a2(a, b);
}

How to contribute?

How can I learn LLVM?

If you don't know how to learn LLVM, you can check out this repository of mine.

Can it run on linux?

Yes.

Can it run on macos?

Yes.

Credits

  • LLVM
  • Some anonymous people