InstructionSet.cpp

// Code taken from https://docs.microsoft.com/en-us/cpp/intrinsics/cpuid-cpuidex?redirectedfrom=MSDN&view=msvc-170
// Modifications marked with `BSC]]` initials

// InstructionSet.cpp
// Compile by using: cl /EHsc /W4 InstructionSet.cpp
// processor: x86, x64
// Uses the __cpuid intrinsic to get information about
// CPU extended instruction set support.

#include <iostream>
#include <vector>
#include <bitset>
#include <array>
#include <string>
#include <intrin.h>

// BSC]] BEGIN
// demo from https://revers.engineering/detecting-hypervisor-presence-on-windows-10/
// modified to compile
enum HYPERVStatus {
    STATUS_HYPERV_DETECTED,
    STATUS_HV_NOT_PRESENT
};

//NOTE: this is returning true when you turn on virtualization in bios
// you can check that by pulling up task manager and viewing performance tab
unsigned long UmIsVirtualized(void) {
    unsigned int invalid_leaf = 0x13371337;
    unsigned int valid_leaf = 0x40000000;
    std::array<int, 4> InvalidLeafResponse;
    std::array<int, 4> ValidLeafResponse;
    __cpuid(InvalidLeafResponse.data(), invalid_leaf);
    __cpuid(ValidLeafResponse.data(), valid_leaf);
    if ((InvalidLeafResponse[0] != ValidLeafResponse[0]) ||
        (InvalidLeafResponse[1] != ValidLeafResponse[1]) ||
        (InvalidLeafResponse[2] != ValidLeafResponse[2]) ||
        (InvalidLeafResponse[3] != ValidLeafResponse[3]))
        return STATUS_HYPERV_DETECTED;

    return STATUS_HV_NOT_PRESENT;
}

struct ChromiumBaseCpuFields {
    bool has_mmx_;
    bool has_sse_;
    bool has_sse2_;
    bool has_sse3_;
    bool has_ssse3_;
    bool has_sse41_;
    bool has_sse42_;
    bool has_popcnt_;
    bool is_running_in_vm_;
    // added from https://revers.engineering/detecting-hypervisor-presence-on-windows-10/
    bool UmIsVirtualized_;
};
// BSC]] END

class InstructionSet {
    // forward declarations
    class InstructionSet_Internal;

public:
    // getters
    static std::string Vendor(void) { return CPU_Rep.vendor_; }
    static std::string Brand(void) { return CPU_Rep.brand_; }

    static bool SSE3(void) { return CPU_Rep.f_1_ECX_[0]; }
    static bool PCLMULQDQ(void) { return CPU_Rep.f_1_ECX_[1]; }
    static bool MONITOR(void) { return CPU_Rep.f_1_ECX_[3]; }
    static bool SSSE3(void) { return CPU_Rep.f_1_ECX_[9]; }
    static bool FMA(void) { return CPU_Rep.f_1_ECX_[12]; }
    static bool CMPXCHG16B(void) { return CPU_Rep.f_1_ECX_[13]; }
    static bool SSE41(void) { return CPU_Rep.f_1_ECX_[19]; }
    static bool SSE42(void) { return CPU_Rep.f_1_ECX_[20]; }
    static bool MOVBE(void) { return CPU_Rep.f_1_ECX_[22]; }
    static bool POPCNT(void) { return CPU_Rep.f_1_ECX_[23]; }
    static bool AES(void) { return CPU_Rep.f_1_ECX_[25]; }
    static bool XSAVE(void) { return CPU_Rep.f_1_ECX_[26]; }
    static bool OSXSAVE(void) { return CPU_Rep.f_1_ECX_[27]; }
    static bool AVX(void) { return CPU_Rep.f_1_ECX_[28]; }
    static bool F16C(void) { return CPU_Rep.f_1_ECX_[29]; }
    static bool RDRAND(void) { return CPU_Rep.f_1_ECX_[30]; }

    static bool MSR(void) { return CPU_Rep.f_1_EDX_[5]; }
    static bool CX8(void) { return CPU_Rep.f_1_EDX_[8]; }
    static bool SEP(void) { return CPU_Rep.f_1_EDX_[11]; }
    static bool CMOV(void) { return CPU_Rep.f_1_EDX_[15]; }
    static bool CLFSH(void) { return CPU_Rep.f_1_EDX_[19]; }
    static bool MMX(void) { return CPU_Rep.f_1_EDX_[23]; }
    static bool FXSR(void) { return CPU_Rep.f_1_EDX_[24]; }
    static bool SSE(void) { return CPU_Rep.f_1_EDX_[25]; }
    static bool SSE2(void) { return CPU_Rep.f_1_EDX_[26]; }

    static bool FSGSBASE(void) { return CPU_Rep.f_7_EBX_[0]; }
    static bool BMI1(void) { return CPU_Rep.f_7_EBX_[3]; }
    static bool HLE(void) { return CPU_Rep.isIntel_ && CPU_Rep.f_7_EBX_[4]; }
    static bool AVX2(void) { return CPU_Rep.f_7_EBX_[5]; }
    static bool BMI2(void) { return CPU_Rep.f_7_EBX_[8]; }
    static bool ERMS(void) { return CPU_Rep.f_7_EBX_[9]; }
    static bool INVPCID(void) { return CPU_Rep.f_7_EBX_[10]; }
    static bool RTM(void) { return CPU_Rep.isIntel_ && CPU_Rep.f_7_EBX_[11]; }
    static bool AVX512F(void) { return CPU_Rep.f_7_EBX_[16]; }
    static bool RDSEED(void) { return CPU_Rep.f_7_EBX_[18]; }
    static bool ADX(void) { return CPU_Rep.f_7_EBX_[19]; }
    static bool AVX512PF(void) { return CPU_Rep.f_7_EBX_[26]; }
    static bool AVX512ER(void) { return CPU_Rep.f_7_EBX_[27]; }
    static bool AVX512CD(void) { return CPU_Rep.f_7_EBX_[28]; }
    static bool SHA(void) { return CPU_Rep.f_7_EBX_[29]; }

    static bool PREFETCHWT1(void) { return CPU_Rep.f_7_ECX_[0]; }

    static bool LAHF(void) { return CPU_Rep.f_81_ECX_[0]; }
    static bool LZCNT(void) { return CPU_Rep.isIntel_ && CPU_Rep.f_81_ECX_[5]; }
    static bool ABM(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_ECX_[5]; }
    static bool SSE4a(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_ECX_[6]; }
    static bool XOP(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_ECX_[11]; }
    static bool TBM(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_ECX_[21]; }

    static bool SYSCALL(void) { return CPU_Rep.isIntel_ && CPU_Rep.f_81_EDX_[11]; }
    static bool MMXEXT(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_EDX_[22]; }
    static bool RDTSCP(void) { return CPU_Rep.isIntel_ && CPU_Rep.f_81_EDX_[27]; }
    static bool _3DNOWEXT(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_EDX_[30]; }
    static bool _3DNOW(void) { return CPU_Rep.isAMD_ && CPU_Rep.f_81_EDX_[31]; }

    // BSC]] START
    static ChromiumBaseCpuFields ChromiumValues(void) { return CPU_Rep.chromium_values_; }
    // BSC]] END

private:
    static const InstructionSet_Internal CPU_Rep;

    class InstructionSet_Internal {
    public:
        InstructionSet_Internal()
            : nIds_{ 0 },
            nExIds_{ 0 },
            isIntel_{ false },
            isAMD_{ false },
            f_1_ECX_{ 0 },
            f_1_EDX_{ 0 },
            f_7_EBX_{ 0 },
            f_7_ECX_{ 0 },
            f_81_ECX_{ 0 },
            f_81_EDX_{ 0 },
            data_{},
            extdata_{}
        {
            //int cpuInfo[4] = {-1};
            std::array<int, 4> cpui;

            // Calling __cpuid with 0x0 as the function_id argument
            // gets the number of the highest valid function ID.
            __cpuid(cpui.data(), 0);
            nIds_ = cpui[0];

            for (int i = 0; i <= nIds_; ++i) {
                __cpuidex(cpui.data(), i, 0);
                data_.push_back(cpui);
            }

            // Capture vendor string
            char vendor[0x20];
            memset(vendor, 0, sizeof(vendor));
            *reinterpret_cast<int*>(vendor) = data_[0][1];
            *reinterpret_cast<int*>(vendor + 4) = data_[0][3];
            *reinterpret_cast<int*>(vendor + 8) = data_[0][2];
            vendor_ = vendor;
            if (vendor_ == "GenuineIntel") {
                isIntel_ = true;
            } else if (vendor_ == "AuthenticAMD") {
                isAMD_ = true;
            }

            // load bitset with flags for function 0x00000001
            if (nIds_ >= 1) {
                f_1_ECX_ = data_[1][2];
                f_1_EDX_ = data_[1][3];
            }

            // load bitset with flags for function 0x00000007
            if (nIds_ >= 7) {
                f_7_EBX_ = data_[7][1];
                f_7_ECX_ = data_[7][2];
            }

            // BSC]] START
            // demo using Chromium code
            // base/cpu.cc => CPU::Initialize()
            // see https://source.chromium.org/chromium/chromium/src/+/main:base/cpu.cc;l=226;drc=f45f47606083b408b7d08c68b1167e8ffcd38e76
            std::array<int, 4> cpui2;
            __cpuid(cpui2.data(), 1);
            chromium_values_.has_mmx_ = (cpui2[3] & 0x00800000) != 0;
            chromium_values_.has_sse_ = (cpui2[3] & 0x02000000) != 0;
            chromium_values_.has_sse2_ = (cpui2[3] & 0x04000000) != 0;
            chromium_values_.has_sse3_ = (cpui2[2] & 0x00000001) != 0;
            chromium_values_.has_ssse3_ = (cpui2[2] & 0x00000200) != 0;
            chromium_values_.has_sse41_ = (cpui2[2] & 0x00080000) != 0;
            chromium_values_.has_sse42_ = (cpui2[2] & 0x00100000) != 0;
            chromium_values_.has_popcnt_ = (cpui2[2] & 0x00800000) != 0;
            // this is the value causing a problem for our testing. 
            // https://lwn.net/Articles/301888/ was cited in Chromium code but it may be irrelevant / outdated.
            chromium_values_.is_running_in_vm_ = (cpui2[2] & 0x80000000) != 0;
            chromium_values_.UmIsVirtualized_ = UmIsVirtualized() == STATUS_HYPERV_DETECTED;
            // BSC]] END

            // Calling __cpuid with 0x80000000 as the function_id argument
            // gets the number of the highest valid extended ID.
            __cpuid(cpui.data(), 0x80000000);
            nExIds_ = cpui[0];

            char brand[0x40];
            memset(brand, 0, sizeof(brand));

            for (int i = 0x80000000; i <= nExIds_; ++i) {
                __cpuidex(cpui.data(), i, 0);
                extdata_.push_back(cpui);
            }

            // load bitset with flags for function 0x80000001
            if (nExIds_ >= 0x80000001) {
                f_81_ECX_ = extdata_[1][2];
                f_81_EDX_ = extdata_[1][3];
            }

            // Interpret CPU brand string if reported
            if (nExIds_ >= 0x80000004) {
                memcpy(brand, extdata_[2].data(), sizeof(cpui));
                memcpy(brand + 16, extdata_[3].data(), sizeof(cpui));
                memcpy(brand + 32, extdata_[4].data(), sizeof(cpui));
                brand_ = brand;
            }
        };

        int nIds_;
        int nExIds_;
        std::string vendor_;
        std::string brand_;
        bool isIntel_;
        bool isAMD_;
        std::bitset<32> f_1_ECX_;
        std::bitset<32> f_1_EDX_;
        std::bitset<32> f_7_EBX_;
        std::bitset<32> f_7_ECX_;
        std::bitset<32> f_81_ECX_;
        std::bitset<32> f_81_EDX_;
        std::vector<std::array<int, 4>> data_;
        std::vector<std::array<int, 4>> extdata_;
        // BSC]] START
        ChromiumBaseCpuFields chromium_values_;
        // BSC]] END
    };
};

// Initialize static member data
const InstructionSet::InstructionSet_Internal InstructionSet::CPU_Rep;

// Print out supported instruction set extensions
// BSC]] START
// EXPERIMENTAL CODE TO CALL invvpid instruction
//typedef uint16_t u16;
typedef uint8_t u8;
typedef uint64_t u64;
//typedef unsigned long gva_t;
//#define VMX_NR_VPIDS				(1 << 16)
#define VMX_VPID_EXTENT_INDIVIDUAL_ADDR		0
#define VMX_VPID_EXTENT_SINGLE_CONTEXT		1
#define VMX_VPID_EXTENT_ALL_CONTEXT		    2
//#define VMX_VPID_EXTENT_SINGLE_NON_GLOBAL	3
// see https://www.felixcloutier.com/x86/invvpid for opcode info
//#define ASM_VMX_INVVPID		  ".byte 0x66, 0x0f, 0x38, 0x81, 0x08"

//shout out to https://github.com/wbenny/hvpp
// and https://github.com/asamy/ksm/blob/6ef52b87903b25c62560d80e8fbec0867e5c2a8b/vmx.h
//enum class invvpid_t : uint32_t {
//    individual_address = 0x00000000,
//    single_context = 0x00000001,
//    all_contexts = 0x00000002,
//    single_context_retaining_globals = 0x00000003,
//};
//struct invvpid_desc_t {
//    uint64_t vpid : 16;
//    uint64_t reserved : 48;
//    uint64_t linear_address;
//};
typedef struct {
    u64 vpid : 16;
    u64 rsvd : 48;
    u64 gva;
} invvpid_t;

extern "C" {
    u8 __invvpid(u64 type, const invvpid_t* i);
}

inline u8 invvpid(u64 type) {
    static invvpid_t demo = { 0, 0, 0 };
    return __invvpid(type, &demo);
}

// rename main() => InstructionSet_main() and change return type to void
void InstructionSet_main() {
    std::cout << std::endl << __FILE__ << std::endl;
// BSC]] END
    auto& outstream = std::cout;

    auto support_message = [&outstream](std::string isa_feature, bool is_supported) {
        outstream << isa_feature << (is_supported ? " supported" : " not supported") << std::endl;
    };

// BSC]] START
    auto dump_chromium = [&outstream](ChromiumBaseCpuFields values) {
        outstream << "BEGIN - Chromium Base::CPU fields" << std::endl;
        outstream << "has_mmx_ = " << values.has_mmx_ << std::endl;
        outstream << "has_sse_ = " << values.has_sse_ << std::endl;
        outstream << "has_sse2_ = " << values.has_sse2_ << std::endl;
        outstream << "has_sse3_ = " << values.has_sse3_ << std::endl;
        outstream << "has_ssse3_ = " << values.has_ssse3_ << std::endl;
        outstream << "has_sse41_ = " << values.has_sse41_ << std::endl;
        outstream << "has_sse42_ = " << values.has_sse42_ << std::endl;
        outstream << "has_popcnt_ = " << values.has_popcnt_ << std::endl;
        outstream << "is_running_in_vm_ = " << values.is_running_in_vm_ << std::endl;
        outstream << "UmIsVirtualized = " << values.UmIsVirtualized_ << std::endl;
        outstream << "END - Chromium Base::CPU fields" << std::endl;
    };
    // experimental code to call invvpid instruction (clearing TLB)
    //u8 val = invvpid(VMX_VPID_EXTENT_ALL_CONTEXT);
    //outstream << "invvpid returned " << val << std::endl;
    dump_chromium(InstructionSet::ChromiumValues());


// BSC]] END

    std::cout << InstructionSet::Vendor() << std::endl;
    std::cout << InstructionSet::Brand() << std::endl;

    support_message("3DNOW", InstructionSet::_3DNOW());
    support_message("3DNOWEXT", InstructionSet::_3DNOWEXT());
    support_message("ABM", InstructionSet::ABM());
    support_message("ADX", InstructionSet::ADX());
    support_message("AES", InstructionSet::AES());
    support_message("AVX", InstructionSet::AVX());
    support_message("AVX2", InstructionSet::AVX2());
    support_message("AVX512CD", InstructionSet::AVX512CD());
    support_message("AVX512ER", InstructionSet::AVX512ER());
    support_message("AVX512F", InstructionSet::AVX512F());
    support_message("AVX512PF", InstructionSet::AVX512PF());
    support_message("BMI1", InstructionSet::BMI1());
    support_message("BMI2", InstructionSet::BMI2());
    support_message("CLFSH", InstructionSet::CLFSH());
    support_message("CMPXCHG16B", InstructionSet::CMPXCHG16B());
    support_message("CX8", InstructionSet::CX8());
    support_message("ERMS", InstructionSet::ERMS());
    support_message("F16C", InstructionSet::F16C());
    support_message("FMA", InstructionSet::FMA());
    support_message("FSGSBASE", InstructionSet::FSGSBASE());
    support_message("FXSR", InstructionSet::FXSR());
    support_message("HLE", InstructionSet::HLE());
    support_message("INVPCID", InstructionSet::INVPCID());
    support_message("LAHF", InstructionSet::LAHF());
    support_message("LZCNT", InstructionSet::LZCNT());
    support_message("MMX", InstructionSet::MMX());
    support_message("MMXEXT", InstructionSet::MMXEXT());
    support_message("MONITOR", InstructionSet::MONITOR());
    support_message("MOVBE", InstructionSet::MOVBE());
    support_message("MSR", InstructionSet::MSR());
    support_message("OSXSAVE", InstructionSet::OSXSAVE());
    support_message("PCLMULQDQ", InstructionSet::PCLMULQDQ());
    support_message("POPCNT", InstructionSet::POPCNT());
    support_message("PREFETCHWT1", InstructionSet::PREFETCHWT1());
    support_message("RDRAND", InstructionSet::RDRAND());
    support_message("RDSEED", InstructionSet::RDSEED());
    support_message("RDTSCP", InstructionSet::RDTSCP());
    support_message("RTM", InstructionSet::RTM());
    support_message("SEP", InstructionSet::SEP());
    support_message("SHA", InstructionSet::SHA());
    support_message("SSE", InstructionSet::SSE());
    support_message("SSE2", InstructionSet::SSE2());
    support_message("SSE3", InstructionSet::SSE3());
    support_message("SSE4.1", InstructionSet::SSE41());
    support_message("SSE4.2", InstructionSet::SSE42());
    support_message("SSE4a", InstructionSet::SSE4a());
    support_message("SSSE3", InstructionSet::SSSE3());
    support_message("SYSCALL", InstructionSet::SYSCALL());
    support_message("TBM", InstructionSet::TBM());
    support_message("XOP", InstructionSet::XOP());
    support_message("XSAVE", InstructionSet::XSAVE());
}