Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Execution of xgetbv instruction and setting up of the XCR register. #1955

Open
Nitr0-G opened this issue May 13, 2024 · 5 comments
Open

Execution of xgetbv instruction and setting up of the XCR register. #1955

Nitr0-G opened this issue May 13, 2024 · 5 comments

Comments

@Nitr0-G
Copy link

Nitr0-G commented May 13, 2024

Hello, guys!
How can I emulate the xgetbv instruction? Judging by the latest updates, there is support for it, but I have not found anywhere examples of configuring the XCR register for it. I wrote the code below, but apparently I still do not have its execution and get: uc_emu_start(uc, ctx.m_ExecuteFromRip, ctx.m_ImageEnd, 0, 0); returned with UC_ERR_INSN_INVALID when xgetbv executing

My code for configuring the XCR register:

	DWORD_PTR XCR0 = _xgetbv(0);

	uc_x86_msr MsrIa32VmxCr0Fixed0;
	MsrIa32VmxCr0Fixed0.rid = (uint32_t)Msr::kIa32VmxCr0Fixed0;
	MsrIa32VmxCr0Fixed0.value = XCR0;

	uc_x86_msr MsrIa32VmxCr0Fixed1;
	MsrIa32VmxCr0Fixed0.rid = (uint32_t)Msr::kIa32VmxCr0Fixed1;
	MsrIa32VmxCr0Fixed0.value = XCR0;

	uc_reg_write(m_uc, UC_X86_REG_MSR, &MsrIa32VmxCr0Fixed0);
	uc_reg_write(m_uc, UC_X86_REG_MSR, &MsrIa32VmxCr0Fixed1);

Verison of UC: I'm using unicorn from the master branch.
@Nitr0-G
Copy link
Author

Nitr0-G commented May 13, 2024 •
edited
Loading

poc

#define X86_CODE64                                                             \
    "\x0F\x01\xD0"
    
// memory address where emulation starts
#define ADDRESS 0x1000000

static void test_x86_64(void)
{
    uc_engine *uc;
    uc_err err;
    uc_hook trace1, trace2, trace3, trace4;

    int64_t rax = 0x71f3029efd49d41d;
    int64_t rbx = 0xd87b45277f133ddb;
    int64_t rcx = 0xab40d1ffd8afc461;
    int64_t rdx = 0x919317b4a733f01;
    int64_t rsi = 0x4c24e753a17ea358;
    int64_t rdi = 0xe509a57d2571ce96;
    int64_t r8 = 0xea5b108cc2b9ab1f;
    int64_t r9 = 0x19ec097c8eb618c1;
    int64_t r10 = 0xec45774f00c5f682;
    int64_t r11 = 0xe17e9dbec8c074aa;
    int64_t r12 = 0x80f86a8dc0f6d457;
    int64_t r13 = 0x48288ca5671c5492;
    int64_t r14 = 0x595f72f6e4017f6e;
    int64_t r15 = 0x1efd97aea331cccc;

    int64_t rsp = ADDRESS + 0x200000;

    printf("Emulate x86_64 code\n");

    // Initialize emulator in X86-64bit mode
    err = uc_open(UC_ARCH_X86, UC_MODE_64, &uc);
    if (err) {
        printf("Failed on uc_open() with error returned: %u\n", err);
        return;
    }

    // map 2MB memory for this emulation
    uc_mem_map(uc, ADDRESS, 2 * 1024 * 1024, UC_PROT_ALL);

    // write machine code to be emulated to memory
    if (uc_mem_write(uc, ADDRESS, X86_CODE64, sizeof(X86_CODE64) - 1)) {
        printf("Failed to write emulation code to memory, quit!\n");
        return;
    }

    // initialize machine registers
    uc_reg_write(uc, UC_X86_REG_RSP, &rsp);

    uc_reg_write(uc, UC_X86_REG_RAX, &rax);
    uc_reg_write(uc, UC_X86_REG_RBX, &rbx);
    uc_reg_write(uc, UC_X86_REG_RCX, &rcx);
    uc_reg_write(uc, UC_X86_REG_RDX, &rdx);
    uc_reg_write(uc, UC_X86_REG_RSI, &rsi);
    uc_reg_write(uc, UC_X86_REG_RDI, &rdi);
    uc_reg_write(uc, UC_X86_REG_R8, &r8);
    uc_reg_write(uc, UC_X86_REG_R9, &r9);
    uc_reg_write(uc, UC_X86_REG_R10, &r10);
    uc_reg_write(uc, UC_X86_REG_R11, &r11);
    uc_reg_write(uc, UC_X86_REG_R12, &r12);
    uc_reg_write(uc, UC_X86_REG_R13, &r13);
    uc_reg_write(uc, UC_X86_REG_R14, &r14);
    uc_reg_write(uc, UC_X86_REG_R15, &r15);

    // tracing all basic blocks with customized callback
    uc_hook_add(uc, &trace1, UC_HOOK_BLOCK, hook_block, NULL, 1, 0);

    // tracing all instructions in the range [ADDRESS, ADDRESS+20]
    uc_hook_add(uc, &trace2, UC_HOOK_CODE, hook_code64, NULL, ADDRESS,
                ADDRESS + 20);

    // tracing all memory WRITE access (with @begin > @end)
    uc_hook_add(uc, &trace3, UC_HOOK_MEM_WRITE, hook_mem64, NULL, 1, 0);

    // tracing all memory READ access (with @begin > @end)
    uc_hook_add(uc, &trace4, UC_HOOK_MEM_READ, hook_mem64, NULL, 1, 0);

    // emulate machine code in infinite time (last param = 0), or when
    // finishing all the code.
    err = uc_emu_start(uc, ADDRESS, ADDRESS + sizeof(X86_CODE64) - 1, 0, 0);
    if (err) {
        printf("Failed on uc_emu_start() with error returned %u: %s\n", err,
               uc_strerror(err));
    }

    // now print out some registers
    printf(">>> Emulation done. Below is the CPU context\n");

    uc_reg_read(uc, UC_X86_REG_RAX, &rax);
    uc_reg_read(uc, UC_X86_REG_RBX, &rbx);
    uc_reg_read(uc, UC_X86_REG_RCX, &rcx);
    uc_reg_read(uc, UC_X86_REG_RDX, &rdx);
    uc_reg_read(uc, UC_X86_REG_RSI, &rsi);
    uc_reg_read(uc, UC_X86_REG_RDI, &rdi);
    uc_reg_read(uc, UC_X86_REG_R8, &r8);
    uc_reg_read(uc, UC_X86_REG_R9, &r9);
    uc_reg_read(uc, UC_X86_REG_R10, &r10);
    uc_reg_read(uc, UC_X86_REG_R11, &r11);
    uc_reg_read(uc, UC_X86_REG_R12, &r12);
    uc_reg_read(uc, UC_X86_REG_R13, &r13);
    uc_reg_read(uc, UC_X86_REG_R14, &r14);
    uc_reg_read(uc, UC_X86_REG_R15, &r15);

    printf(">>> RAX = 0x%" PRIx64 "\n", rax);
    printf(">>> RBX = 0x%" PRIx64 "\n", rbx);
    printf(">>> RCX = 0x%" PRIx64 "\n", rcx);
    printf(">>> RDX = 0x%" PRIx64 "\n", rdx);
    printf(">>> RSI = 0x%" PRIx64 "\n", rsi);
    printf(">>> RDI = 0x%" PRIx64 "\n", rdi);
    printf(">>> R8 = 0x%" PRIx64 "\n", r8);
    printf(">>> R9 = 0x%" PRIx64 "\n", r9);
    printf(">>> R10 = 0x%" PRIx64 "\n", r10);
    printf(">>> R11 = 0x%" PRIx64 "\n", r11);
    printf(">>> R12 = 0x%" PRIx64 "\n", r12);
    printf(">>> R13 = 0x%" PRIx64 "\n", r13);
    printf(">>> R14 = 0x%" PRIx64 "\n", r14);
    printf(">>> R15 = 0x%" PRIx64 "\n", r15);

    uc_close(uc);
}

@Nitr0-G
Copy link
Author

Nitr0-G commented May 13, 2024

I looked through all the showcases that are presented on the site and did not find the initialization of the XCR register there, as I understand it, it is initialized automatically. I have already tried to study the problem and for some reason, when translating instructions, the check does not pass(https://github.com/unicorn-engine/unicorn/blob/master/qemu/target/i386/translate.c#L8088)

@Nitr0-G
Copy link
Author

Nitr0-G commented May 13, 2024

uc_mem_map => uc_init => x86_reg_reset = reset of all values in spite of uc_ctl_set_cpu_model(uc, UC_CPU_X86_EPYC);
изображение

@Nitr0-G
Copy link
Author

Nitr0-G commented May 13, 2024 •
edited
Loading

so for xgetbv, cr4 configuration is still necessary(https://github.com/unicorn-engine/unicorn/blob/master/qemu/target/i386/fpu_helper.c#L1514), but despite this, I do not get access to xgetbv due to the fact that the entire CPUID is reset on uc_mem_map => uc_init and the flag(CPUID_EXT_XSAVE) is removed from s->cpuid_ext_features

@Nitr0-G
Copy link
Author

Nitr0-G commented May 13, 2024 •
edited
Loading

All code is work without cpuid reset(https://github.com/unicorn-engine/unicorn/blob/master/qemu/target/i386/unicorn.c#L68C1-L182C2)

void x86_reg_reset(struct uc_struct *uc)
{
    CPUArchState *env = uc->cpu->env_ptr;

    //env->features[FEAT_1_EDX] = CPUID_CX8 | CPUID_CMOV | CPUID_SSE2 |
    //                            CPUID_FXSR | CPUID_SSE | CPUID_CLFLUSH;
    //env->features[FEAT_1_ECX] = CPUID_EXT_SSSE3 | CPUID_EXT_SSE41 |
    //                            CPUID_EXT_SSE42 | CPUID_EXT_AES |
    //                            CPUID_EXT_CX16;
    //env->features[FEAT_8000_0001_EDX] = CPUID_EXT2_3DNOW | CPUID_EXT2_RDTSCP;
    //env->features[FEAT_8000_0001_ECX] = CPUID_EXT3_LAHF_LM | CPUID_EXT3_ABM |
    //                                    CPUID_EXT3_SKINIT | CPUID_EXT3_CR8LEG;
    //env->features[FEAT_7_0_EBX] = CPUID_7_0_EBX_BMI1 | CPUID_7_0_EBX_BMI2 |
    //                              CPUID_7_0_EBX_ADX | CPUID_7_0_EBX_SMAP;

    memset(env->regs, 0, sizeof(env->regs));
    memset(env->segs, 0, sizeof(env->segs));
    memset(env->cr, 0, sizeof(env->cr));

    memset(&env->ldt, 0, sizeof(env->ldt));
    memset(&env->gdt, 0, sizeof(env->gdt));
    memset(&env->tr, 0, sizeof(env->tr));
    memset(&env->idt, 0, sizeof(env->idt));

    env->eip = 0;
    cpu_load_eflags(env, 0, -1);
    env->cc_op = CC_OP_EFLAGS;

    env->fpstt = 0; /* top of stack index */
    env->fpus = 0;
    env->fpuc = 0;
    memset(env->fptags, 0, sizeof(env->fptags)); /* 0 = valid, 1 = empty */

    env->mxcsr = 0;
    memset(env->xmm_regs, 0, sizeof(env->xmm_regs));
    memset(&env->xmm_t0, 0, sizeof(env->xmm_t0));
    memset(&env->mmx_t0, 0, sizeof(env->mmx_t0));

    memset(env->ymmh_regs, 0, sizeof(env->ymmh_regs));

    memset(env->opmask_regs, 0, sizeof(env->opmask_regs));
    memset(env->zmmh_regs, 0, sizeof(env->zmmh_regs));

    /* sysenter registers */
    env->sysenter_cs = 0;
    env->sysenter_esp = 0;
    env->sysenter_eip = 0;
    env->efer = 0;
    env->star = 0;

    env->vm_hsave = 0;

    env->tsc = 0;
    env->tsc_adjust = 0;
    env->tsc_deadline = 0;

    env->mcg_status = 0;
    env->msr_ia32_misc_enable = 0;
    env->msr_ia32_feature_control = 0;

    env->msr_fixed_ctr_ctrl = 0;
    env->msr_global_ctrl = 0;
    env->msr_global_status = 0;
    env->msr_global_ovf_ctrl = 0;
    memset(env->msr_fixed_counters, 0, sizeof(env->msr_fixed_counters));
    memset(env->msr_gp_counters, 0, sizeof(env->msr_gp_counters));
    memset(env->msr_gp_evtsel, 0, sizeof(env->msr_gp_evtsel));

#ifdef TARGET_X86_64
    memset(env->hi16_zmm_regs, 0, sizeof(env->hi16_zmm_regs));
    env->lstar = 0;
    env->cstar = 0;
    env->fmask = 0;
    env->kernelgsbase = 0;
#endif

    // TODO: reset other registers in CPUX86State qemu/target-i386/cpu.h

    // properly initialize internal setup for each mode
    switch (uc->mode) {
    default:
        break;
    case UC_MODE_16:
        env->hflags = 0;
        env->cr[0] = 0;
        // undo the damage done by the memset of env->segs above
        // for R_CS, not quite the same as x86_cpu_reset
        cpu_x86_load_seg_cache(env, R_CS, 0, 0, 0xffff,
                               DESC_P_MASK | DESC_S_MASK | DESC_CS_MASK |
                                   DESC_R_MASK | DESC_A_MASK);
        // remainder yields same state as x86_cpu_reset
        load_seg_16_helper(env, R_DS, 0);
        load_seg_16_helper(env, R_ES, 0);
        load_seg_16_helper(env, R_SS, 0);
        load_seg_16_helper(env, R_FS, 0);
        load_seg_16_helper(env, R_GS, 0);

        break;
    case UC_MODE_32:
        env->hflags |= HF_CS32_MASK | HF_SS32_MASK | HF_OSFXSR_MASK;
        cpu_x86_update_cr0(env, CR0_PE_MASK); // protected mode
        break;
    case UC_MODE_64:
        env->hflags |= HF_CS32_MASK | HF_SS32_MASK | HF_CS64_MASK |
                       HF_LMA_MASK | HF_OSFXSR_MASK;
        env->hflags &= ~(HF_ADDSEG_MASK);
        env->efer |= MSR_EFER_LMA | MSR_EFER_LME; // extended mode activated
        cpu_x86_update_cr0(env, CR0_PE_MASK);     // protected mode
        /* If we are operating in 64bit mode then add the Long Mode flag
         * to the CPUID feature flag
         */
        env->features[FEAT_8000_0001_EDX] |= CPUID_EXT2_LM;
        break;
    }
}

Work code:

Emulate x86_64 code
>>> Tracing basic block at 0x1000000, block size = 0x3
>>> Tracing instruction at 0x1000000, instruction size = 0x3
>>> RIP is 0x1000000
>>> Emulation done. Below is the CPU context
>>> RAX = 0x3
>>> RBX = 0xd87b45277f133ddb
>>> RCX = 0x0
>>> RDX = 0x0
>>> RSI = 0x4c24e753a17ea358
>>> RDI = 0xe509a57d2571ce96
>>> R8 = 0xea5b108cc2b9ab1f
>>> R9 = 0x19ec097c8eb618c1
>>> R10 = 0xec45774f00c5f682
>>> R11 = 0xe17e9dbec8c074aa
>>> R12 = 0x80f86a8dc0f6d457
>>> R13 = 0x48288ca5671c5492
>>> R14 = 0x595f72f6e4017f6e
>>> R15 = 0x1efd97aea331cccc
===================================

POC

#define X86_CODE64                                                             \
    "\x0F\x01\xD0"
    
// memory address where emulation starts
#define ADDRESS 0x1000000

static void test_x86_64(void)
{
    uc_engine *uc;
    uc_err err;
    uc_hook trace1, trace2, trace3, trace4;

    int64_t rax = 0x71f3029efd49d41d;
    int64_t rbx = 0xd87b45277f133ddb;
    int64_t rcx = 0x0;
    int64_t rdx = 0x919317b4a733f01;
    int64_t rsi = 0x4c24e753a17ea358;
    int64_t rdi = 0xe509a57d2571ce96;
    int64_t r8 = 0xea5b108cc2b9ab1f;
    int64_t r9 = 0x19ec097c8eb618c1;
    int64_t r10 = 0xec45774f00c5f682;
    int64_t r11 = 0xe17e9dbec8c074aa;
    int64_t r12 = 0x80f86a8dc0f6d457;
    int64_t r13 = 0x48288ca5671c5492;
    int64_t r14 = 0x595f72f6e4017f6e;
    int64_t r15 = 0x1efd97aea331cccc;

    int64_t rsp = ADDRESS + 0x200000;

    printf("Emulate x86_64 code\n");

    // Initialize emulator in X86-64bit mode
    err = uc_open(UC_ARCH_X86, UC_MODE_64, &uc);
    uc_ctl_set_cpu_model(uc, UC_CPU_X86_EPYC);
    if (err) {
        printf("Failed on uc_open() with error returned: %u\n", err);
        return;
    }

    // map 2MB memory for this emulation
    uc_mem_map(uc, ADDRESS, 2 * 1024 * 1024, UC_PROT_ALL);

    // write machine code to be emulated to memory
    if (uc_mem_write(uc, ADDRESS, X86_CODE64, sizeof(X86_CODE64) - 1)) {
        printf("Failed to write emulation code to memory, quit!\n");
        return;
    }

    // initialize machine registers
    uc_reg_write(uc, UC_X86_REG_RSP, &rsp);

    /*	CR4 cr4{};
	cr4.x64.Bitmap.VME = true;
	cr4.x64.Bitmap.PVI = true;
	cr4.x64.Bitmap.TSD = false;
	cr4.x64.Bitmap.DE = false;
	cr4.x64.Bitmap.PSE = false;
	cr4.x64.Bitmap.PAE = false;
	cr4.x64.Bitmap.MCE = true;
	cr4.x64.Bitmap.PGE = false;
	cr4.x64.Bitmap.PCE = true;
	cr4.x64.Bitmap.OSFXSR = true;
	cr4.x64.Bitmap.OSXMMEXCPT = true;
	cr4.x64.Bitmap.UMIP = true;
	cr4.x64.Bitmap.LA57 = false;
	cr4.x64.Bitmap.VMXE = false;
	cr4.x64.Bitmap.SMXE = true;
	cr4.x64.Bitmap.FSGSBASE = true;
	cr4.x64.Bitmap.PCIDE = false;
	cr4.x64.Bitmap.OSXSAVE = true;
	cr4.x64.Bitmap.KL = true;
	cr4.x64.Bitmap.SMEP = false;
	cr4.x64.Bitmap.SMAP = false;
	cr4.x64.Bitmap.PKE = false;
	cr4.x64.Bitmap.CET = true;
	cr4.x64.Bitmap.PKS = false;
	cr4.x64.Bitmap.UINTR = false;*/
    DWORD_PTR Cr4 = 0x85cf43;
    uc_reg_write(uc, UC_X86_REG_CR4, &Cr4);

    uc_reg_write(uc, UC_X86_REG_RAX, &rax);
    uc_reg_write(uc, UC_X86_REG_RBX, &rbx);
    uc_reg_write(uc, UC_X86_REG_RCX, &rcx);
    uc_reg_write(uc, UC_X86_REG_RDX, &rdx);
    uc_reg_write(uc, UC_X86_REG_RSI, &rsi);
    uc_reg_write(uc, UC_X86_REG_RDI, &rdi);
    uc_reg_write(uc, UC_X86_REG_R8, &r8);
    uc_reg_write(uc, UC_X86_REG_R9, &r9);
    uc_reg_write(uc, UC_X86_REG_R10, &r10);
    uc_reg_write(uc, UC_X86_REG_R11, &r11);
    uc_reg_write(uc, UC_X86_REG_R12, &r12);
    uc_reg_write(uc, UC_X86_REG_R13, &r13);
    uc_reg_write(uc, UC_X86_REG_R14, &r14);
    uc_reg_write(uc, UC_X86_REG_R15, &r15);

    // tracing all basic blocks with customized callback
    uc_hook_add(uc, &trace1, UC_HOOK_BLOCK, hook_block, NULL, 1, 0);

    // tracing all instructions in the range [ADDRESS, ADDRESS+20]
    uc_hook_add(uc, &trace2, UC_HOOK_CODE, hook_code64, NULL, ADDRESS,
                ADDRESS + 20);

    // tracing all memory WRITE access (with @begin > @end)
    uc_hook_add(uc, &trace3, UC_HOOK_MEM_WRITE, hook_mem64, NULL, 1, 0);

    // tracing all memory READ access (with @begin > @end)
    uc_hook_add(uc, &trace4, UC_HOOK_MEM_READ, hook_mem64, NULL, 1, 0);

    // emulate machine code in infinite time (last param = 0), or when
    // finishing all the code.
    err = uc_emu_start(uc, ADDRESS, ADDRESS + sizeof(X86_CODE64) - 1, 0, 0);
    if (err) {
        printf("Failed on uc_emu_start() with error returned %u: %s\n", err,
               uc_strerror(err));
    }

    // now print out some registers
    printf(">>> Emulation done. Below is the CPU context\n");

    uc_reg_read(uc, UC_X86_REG_RAX, &rax);
    uc_reg_read(uc, UC_X86_REG_RBX, &rbx);
    uc_reg_read(uc, UC_X86_REG_RCX, &rcx);
    uc_reg_read(uc, UC_X86_REG_RDX, &rdx);
    uc_reg_read(uc, UC_X86_REG_RSI, &rsi);
    uc_reg_read(uc, UC_X86_REG_RDI, &rdi);
    uc_reg_read(uc, UC_X86_REG_R8, &r8);
    uc_reg_read(uc, UC_X86_REG_R9, &r9);
    uc_reg_read(uc, UC_X86_REG_R10, &r10);
    uc_reg_read(uc, UC_X86_REG_R11, &r11);
    uc_reg_read(uc, UC_X86_REG_R12, &r12);
    uc_reg_read(uc, UC_X86_REG_R13, &r13);
    uc_reg_read(uc, UC_X86_REG_R14, &r14);
    uc_reg_read(uc, UC_X86_REG_R15, &r15);

    printf(">>> RAX = 0x%" PRIx64 "\n", rax);
    printf(">>> RBX = 0x%" PRIx64 "\n", rbx);
    printf(">>> RCX = 0x%" PRIx64 "\n", rcx);
    printf(">>> RDX = 0x%" PRIx64 "\n", rdx);
    printf(">>> RSI = 0x%" PRIx64 "\n", rsi);
    printf(">>> RDI = 0x%" PRIx64 "\n", rdi);
    printf(">>> R8 = 0x%" PRIx64 "\n", r8);
    printf(">>> R9 = 0x%" PRIx64 "\n", r9);
    printf(">>> R10 = 0x%" PRIx64 "\n", r10);
    printf(">>> R11 = 0x%" PRIx64 "\n", r11);
    printf(">>> R12 = 0x%" PRIx64 "\n", r12);
    printf(">>> R13 = 0x%" PRIx64 "\n", r13);
    printf(">>> R14 = 0x%" PRIx64 "\n", r14);
    printf(">>> R15 = 0x%" PRIx64 "\n", r15);

    uc_close(uc);
}

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
1 participant
@Nitr0-G