MemoryModule x86x64 内存加载DLL
by devseed，此篇教程同时发在论坛和我的博客上，完整源码见我的github。
上篇链接：SimpleDpack
0x0 前言
距离上次我写的SimpleDpack教程（为了理解加壳原理自己写的轻量级压缩壳），已经有一年多了。前几天修复x64的IAT载入序号问题，于是想着可以用这个原理自己重写一个解析PE和内存加载DLL的功能。虽然已有类似实现功能，如MemoryModule，但是这个项目还是有点复杂，模块间耦合度有些高，没法作为shellcode附加到EXE来引导DLL。
因此我决定重写一个低耦合，同时支持x86和x64，可以作为shellcode引导”附加到EXE区段上的DLL”的框架 。
0x1 MemoryModule 设计思路
(1) usage
本框架采取与windows加载DLL相似的名称来设计api，可以实现：
- 将DLL按照内存对齐模式载入到内存中
- 动态重定向DLL中的
reloc地址，并载入IAT
const char *dllpath = "test.dll";
size_t mempesize = 0;
void *memdll = NULL;
// load the pe file in memory and align it to memory align
void *mempe = winpe_memload_file(dllpath, &mempesize, TRUE);
// memory loadlibrary
memdll = winpe_memLoadLibrary(mempe);
winpe_memFreeLibrary(memdll);
// memory loadlibrary at specific address
size_t targetaddr = sizeof(size_t) > 4 ? 0x140030000: 0x90000;
memdll = winpe_memLoadLibraryEx(mempe, targetaddr,
WINPE_LDFLAG_MEMALLOC, (PFN_LoadLibraryA)winpe_findloadlibrarya(),
(PFN_GetProcAddress)winpe_memGetProcAddress);
winpe_memFreeLibrary(memdll);
free(mempe);
- 将DLL注入到EXE的附加区段，并且用
shellcode引导该DLL
win_injectmemdll.exe exepath dllpath [outpath]
(2) libwinpe.c
参考stb用XXX_IMPLEMENTATION宏控制代码实现，使得include单个头文件就能把模块功能全部引入的做法，有着单头文件结成非常容易实用的特点。于是我就用了一周多时间，自己写了个类似的低耦合单头文件winpe.h，用于解析与加载PE。为了便于shellcode引导，系统函数大多用函数指针传入参数，关键函数全部声明为inline。另外，我还用内联汇编写了从ldr中获取kernel32的基地址，从而减少对IAT硬编码的依赖。
libwinpe.c, 与libwinpe.def（引入def是因为stdcall编译后会被修改名称），把winpe.h的声明的导出函数编译为DLL，如下：
#define WINPE_SHARED
#define WINPE_IMPLEMENTATION
#include "winpe.h"
EXPORTS
winpe_appendsecth
winpe_findgetprocaddress
winpe_findkernel32
winpe_findloadlibrarya
winpe_findspace
winpe_findmodulea
winpe_imagebaseval
winpe_imagesizeval
winpe_memFreeLibrary
winpe_memFreeLibraryEx
winpe_memGetProcAddress
winpe_memLoadLibrary
winpe_memLoadLibraryEx
winpe_membindiat
winpe_membindtls
winpe_memfindexp
winpe_memfindiat
winpe_memforwardexp
winpe_memload
winpe_memload_file
winpe_memreloc
winpe_noaslr
winpe_oepval
winpe_overlayload_file
winpe_overlayoffset
(3) win_injectmemdll.c
对于编写shellcode， 这里为了方便调试，采取用pyhton的keystone进行jit生成的机器码写入win_injectmemdll.c的全局变量里的做法， 如下：
def inject_shellcodestubs(srcpath, libwinpepath, targetpath):
pedll = lief.parse(libwinpepath)
pedll_oph = pedll.optional_header
# generate oepint shellcode
if pedll_oph.magic == lief.PE.PE_TYPE.PE32_PLUS:
oepinit_code = gen_oepinit_code64()
pass
elif pedll_oph.magic == lief.PE.PE_TYPE.PE32:
oepinit_code = gen_oepinit_code32()
pass
else:
print("error invalid pe magic!", pedll_oph.magic)
return
# find necessary functions
FUNC_SIZE =0x400
codes = {"winpe_memreloc": 0,
"winpe_membindiat": 0,
"winpe_membindtls": 0,
"winpe_findloadlibrarya": 0,
"winpe_findgetprocaddress": 0}
for k in codes.keys():
func = next(filter(lambda e : e.name == k,
pedll.exported_functions))
codes[k] = pedll.get_content_from_virtual_address(
func.address, FUNC_SIZE)
codes['winpe_oepinit'] = oepinit_code
# write shellcode to c source file
with open(srcpath, "rb") as fp:
srctext = fp.read().decode('utf8')
for k, v in codes.items():
k = k.replace("winpe_", "")
_codetext = ",".join([hex(x) for x in v])
srctext = re.sub("g_" + k + r"_code(.+?)(\{0x90\})",
"g_" + k + r"_code\1{" + _codetext +"}", srctext)
with open(targetpath, "wb") as fp:
fp.write(srctext.encode('utf8'))
详见win_injectmemdll_shellcodestub.py
0x2 内存加载DLL
(1) windows dll加载机制
参考https://bbs.pediy.com/thread-252260.htm， windows对于加载DLL调用如下：
a) 当调用LoadLibrary时首先打开文件并将按照PE文件格式解析数据，并以IMAGE方式将PE各段内容
逐块映射到内存中,如果需要并修复重定位：
kernel32!LoadlibraryA
->kernel32!LoadLibraryExA
->kernel32!LoadLibraryExW
->ntdll!LdrLoadDll
->ntdll!LdrpLoadDll
->ntdll!LdrpMapDll
->ntdll!LdrpCreateDllSection
->ntdll!NtOpenFile //打开目标文件*
->ntdll!NtCreateSection //创建映射*
->ntdll!NtMapViewOfSection //映射*
->LdrRelocateImageWithBias //修复重定位
b) 接着处理调入表等，然后调用TLS回调函数和DLLMAN入口函数；
kernel32!LoadlibraryA
->kernel32!LoadLibraryExA
->kernel32!LoadLibraryExW
->ntdll!LdrLoadDll
->ntdll!LdrpLoadDll
->ntdll!LdrpMapDll
->ntdll!LdrpWalkImportDescriptor //处理导入表等
->LdrpRunInitializeRoutines //调用DLLMAN
->LdrpCallTlsInitializers
->LdrpCallInitRoutine
总结为如下流程：
- 载入DLL到内存上，将DLL进行
memory align对齐
- 通过
reloc段重定向硬编码地址
- 逐一扫码
IAT表OFT的模块和函数名，加载对于模块，将地址填入FT中
- 调用
DllMain函数（即DLL的OEP），fdwReason参数为DLL_PROCESS_ATTACH
- 初始化
TLS, reason 为DLL_PROCESS_ATTACH
(2) 模拟Windows加载DLL
.1 载入DLL到内存上，将DLL进行memory align对齐
这里为了降低耦合，没有用系统的memset和memcpy，自己用inline实现的。
inline size_t STDCALL winpe_memload(
const void *rawpe, size_t rawsize,
void *mempe, size_t memsize,
bool_t same_align)
{
// load rawpe to memalign
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)rawpe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)rawpe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_SECTION_HEADER pSectHeader = (PIMAGE_SECTION_HEADER)
((void *)pOptHeader + pFileHeader->SizeOfOptionalHeader);
WORD sectNum = pFileHeader->NumberOfSections;
size_t imagesize = pOptHeader->SizeOfImage;
if(!mempe) return imagesize;
else if(memsize!=0 && memsize<imagesize) return 0;
_inl_memset(mempe, 0, imagesize);
_inl_memcpy(mempe, rawpe, pOptHeader->SizeOfHeaders);
for(WORD i=0;i<sectNum;i++)
{
_inl_memcpy(mempe+pSectHeader[i].VirtualAddress,
rawpe+pSectHeader[i].PointerToRawData,
pSectHeader[i].SizeOfRawData);
}
// adjust all to mem align
if(same_align)
{
pDosHeader = (PIMAGE_DOS_HEADER)mempe;
pNtHeader = (PIMAGE_NT_HEADERS)((void*)mempe + pDosHeader->e_lfanew);
pFileHeader = &pNtHeader->FileHeader;
pOptHeader = &pNtHeader->OptionalHeader;
pSectHeader = (PIMAGE_SECTION_HEADER)
((void *)pOptHeader + pFileHeader->SizeOfOptionalHeader);
sectNum = pFileHeader->NumberOfSections;
pOptHeader->FileAlignment = pOptHeader->SectionAlignment;
for(WORD i=0;i<sectNum;i++)
{
pSectHeader[i].PointerToRawData = pSectHeader[i].VirtualAddress;
}
}
return imagesize;
}
.2 通过reloc段重定向硬编码地址
手动将以imagebase为基准的va重定向到目标imagebase的va。
inline size_t STDCALL winpe_memreloc(
void *mempe, size_t newimagebase)
{
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)mempe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)mempe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_DATA_DIRECTORY pDataDirectory = pOptHeader->DataDirectory;
PIMAGE_DATA_DIRECTORY pRelocEntry = &pDataDirectory[IMAGE_DIRECTORY_ENTRY_BASERELOC];
DWORD reloc_count = 0;
DWORD reloc_offset = 0;
int64_t shift = (int64_t)newimagebase -
(int64_t)pOptHeader->ImageBase;
while (reloc_offset < pRelocEntry->Size)
{
PIMAGE_BASE_RELOCATION pBaseReloc = (PIMAGE_BASE_RELOCATION)
((void*)mempe + pRelocEntry->VirtualAddress + reloc_offset);
PRELOCOFFSET pRelocOffset = (PRELOCOFFSET)((void*)pBaseReloc
+ sizeof(IMAGE_BASE_RELOCATION));
// RELOCOFFSET block num
DWORD item_num = (pBaseReloc->SizeOfBlock -
sizeof(IMAGE_BASE_RELOCATION)) / sizeof(RELOCOFFSET);
for (size_t i = 0; i < item_num; i++)
{
if (!pRelocOffset[i].type &&
!pRelocOffset[i].offset) continue;
DWORD targetoffset = pBaseReloc->VirtualAddress +
pRelocOffset[i].offset;
size_t *paddr = (size_t *)((void*)mempe + targetoffset);
size_t relocaddr = (size_t)((int64_t)*paddr + shift);
*paddr = relocaddr;
}
reloc_offset += sizeof(IMAGE_BASE_RELOCATION) +
sizeof(RELOCOFFSET) * item_num;
reloc_count += item_num;
}
pOptHeader->ImageBase = newimagebase;
return reloc_count;
}
.3 逐一扫码IAT表OFT的模块和函数名，加载对于模块，将地址填入FT中
用LoadLibraryA和GetProcAddress实现，考虑到shellcode，不能直接调用，应该通过传入函数指针调用。
另外还要考虑OFT中会出现用序号的情况，x86和x64判断表中不同，用sizeof(size_t)区分。
WINPEDEF WINPE_EXPORT
inline size_t STDCALL winpe_membindiat(void *mempe,
PFN_LoadLibraryA pfnLoadLibraryA,
PFN_GetProcAddress pfnGetProcAddress)
{
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)mempe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)mempe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_DATA_DIRECTORY pDataDirectory = pOptHeader->DataDirectory;
PIMAGE_DATA_DIRECTORY pImpEntry =
&pDataDirectory[IMAGE_DIRECTORY_ENTRY_IMPORT];
PIMAGE_IMPORT_DESCRIPTOR pImpDescriptor =
(PIMAGE_IMPORT_DESCRIPTOR)(mempe + pImpEntry->VirtualAddress);
PIMAGE_THUNK_DATA pFtThunk = NULL;
PIMAGE_THUNK_DATA pOftThunk = NULL;
LPCSTR pDllName = NULL;
PIMAGE_IMPORT_BY_NAME pImpByName = NULL;
size_t funcva = 0;
char *funcname = NULL;
// origin GetProcAddress will crash at InitializeSListHead
if(!pfnLoadLibraryA) pfnLoadLibraryA =
(PFN_LoadLibraryA)winpe_findloadlibrarya();
if(!pfnGetProcAddress) pfnGetProcAddress =
(PFN_GetProcAddress)winpe_findgetprocaddress();
DWORD iat_count = 0;
for (; pImpDescriptor->Name; pImpDescriptor++)
{
pDllName = (LPCSTR)(mempe + pImpDescriptor->Name);
pFtThunk = (PIMAGE_THUNK_DATA)
(mempe + pImpDescriptor->FirstThunk);
pOftThunk = (PIMAGE_THUNK_DATA)
(mempe + pImpDescriptor->OriginalFirstThunk);
size_t dllbase = (size_t)pfnLoadLibraryA(pDllName);
if(!dllbase) return 0;
for (int j=0; pFtThunk[j].u1.Function
&& pOftThunk[j].u1.Function; j++)
{
size_t _addr = (size_t)(mempe + pOftThunk[j].u1.AddressOfData);
if(sizeof(size_t)>4) // x64
{
if((_addr>>63) == 1)
{
funcname = (char *)(_addr & 0xffff);
}
else
{
pImpByName=(PIMAGE_IMPORT_BY_NAME)_addr;
funcname = pImpByName->Name;
}
}
else
{
if(((size_t)pImpByName>>31) == 1)
{
funcname = (char *)(_addr & 0xffff);
}
else
{
pImpByName=(PIMAGE_IMPORT_BY_NAME)_addr;
funcname = pImpByName->Name;
}
}
funcva = (size_t)pfnGetProcAddress(
(HMODULE)dllbase, funcname);
if(!funcva) continue;
pFtThunk[j].u1.Function = funcva;
assert(funcva == (size_t)GetProcAddress(
(HMODULE)dllbase, funcname));
iat_count++;
}
}
return iat_count;
}
.4 调用DllMain函数（即DLL的OEP），fdwReason参数为DLL_PROCESS_ATTACH
前三步都完成后可以来通过DllMain来引导了。
// initial memory module
if(!winpe_memreloc((void*)imagebase, imagebase))
return NULL;
if(!winpe_membindiat((void*)imagebase,
pfnLoadLibraryA, pfnGetProcAddress)) return NULL;
winpe_membindtls(mempe, DLL_PROCESS_ATTACH);
PFN_DllMain pfnDllMain = (PFN_DllMain)
(imagebase + winpe_oepval((void*)imagebase, 0));
pfnDllMain((HINSTANCE)imagebase, DLL_PROCESS_ATTACH, NULL);
return (void*)imagebase;
.5 初始化TLS, reason 为DLL_PROCESS_ATTACH
这部分我没有找到有TLS的DLL，所以没法进一步测试了。这部分代码直接借鉴的MemoryModule的相关函数。
inline size_t STDCALL winpe_membindtls(void *mempe, DWORD resaon)
{
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)mempe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)mempe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_DATA_DIRECTORY pDataDirectory = pOptHeader->DataDirectory;
PIMAGE_DATA_DIRECTORY pTlsDirectory =
&pDataDirectory[IMAGE_DIRECTORY_ENTRY_TLS];
if(!pTlsDirectory->VirtualAddress) return 0;
size_t tls_count = 0;
PIMAGE_TLS_DIRECTORY pTlsEntry = (PIMAGE_TLS_DIRECTORY)
(mempe + pTlsDirectory->VirtualAddress);
PIMAGE_TLS_CALLBACK *tlscb= (PIMAGE_TLS_CALLBACK*)
pTlsEntry->AddressOfCallBacks;
if(tlscb)
{
while(*tlscb)
{
(*tlscb)(mempe, reason, NULL);
tlscb++;
tls_count++;
}
}
return tls_count;
}
0x3 DLL附加到EXE区段与shellcode编写
(1) EXE区段扩容
首先需要修改exe的区段头，在后面增加一个区段。这里要特别注意区段的对齐，之后还要修正SizeOfImage。
inline size_t STDCALL winpe_appendsecth(void *pe,
PIMAGE_SECTION_HEADER psecth)
{
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)pe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)pe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_SECTION_HEADER pSectHeader = (PIMAGE_SECTION_HEADER)
((void*)pOptHeader + pFileHeader->SizeOfOptionalHeader);
WORD sectNum = pFileHeader->NumberOfSections;
PIMAGE_SECTION_HEADER pLastSectHeader = &pSectHeader[sectNum-1];
DWORD addr, align;
// check the space to append section
if(pFileHeader->SizeOfOptionalHeader
+ sizeof(IMAGE_SECTION_HEADER)
> pSectHeader[0].PointerToRawData) return 0;
// fill rva addr
align = pOptHeader->SectionAlignment;
addr = pLastSectHeader->VirtualAddress + pLastSectHeader->Misc.VirtualSize;
if(addr % align) addr += align - addr%align;
psecth->VirtualAddress = addr;
// fill file offset
align = pOptHeader->FileAlignment;
addr = pLastSectHeader->PointerToRawData+ pLastSectHeader->SizeOfRawData;
if(addr % align) addr += align - addr%align;
psecth->PointerToRawData = addr;
// adjust the section and imagesize
pFileHeader->NumberOfSections++;
_inl_memcpy(&pSectHeader[sectNum], psecth, sizeof(IMAGE_SECTION_HEADER));
align = pOptHeader->SectionAlignment;
addr = psecth->VirtualAddress + psecth->Misc.VirtualSize;
if(addr % align) addr += align - addr%align;
pOptHeader->SizeOfImage = addr;
return pOptHeader->SizeOfImage;
}
(2) 引导DLL的shellcode编写
这部分内容和winpe_memLoadLibrary函数实现的方式类似，只不过要用汇编来写shellcode，引导完成后跳转到原来exe的oep。 关键地址可以先留空，在注入时候填入对位置。对于相应的函数的shellcode，由win_injectmemdll_shellcodestub.py读取对应的libpe32.dll函数，填入对应的全局变量。下面代码以x86为例：
def gen_oepinit_code32():
ks = Ks(KS_ARCH_X86, KS_MODE_32)
code_str = f"""
// for relative address, get the base of addr
call getip;
lea ebx, [eax-5];
// get the imagebase
mov eax, 0x30; // to avoid relative addressing
mov edi, dword ptr fs:[eax]; //peb
mov edi, [edi + 0ch]; //ldr
mov edi, [edi + 14h]; //InMemoryOrderLoadList, this
mov edi, [edi -8h + 18h]; //this.DllBase
// get loadlibrarya, getprocaddress
mov eax, [ebx + findloadlibrarya];
add eax, edi;
call eax;
mov [ebx + findloadlibrarya], eax;
mov eax, [ebx + findgetprocaddress];
add eax, edi;
call eax;
mov [ebx + findgetprocaddress], eax;
// reloc
mov eax, [ebx + dllrva];
add eax, edi;
push eax;
push eax;
mov eax, [ebx + memrelocrva];
add eax, edi;
call eax;
// bind iat
mov eax, [ebx + findgetprocaddress];
push eax; // arg3, getprocaddress
mov eax, [ebx + findloadlibrarya];
push eax; // arg2, loadlibraryas
mov eax, [ebx + dllrva];
add eax, edi;
push eax; // arg1, dllbase value
mov eax, [ebx + membindiatrva];
add eax, edi
call eax;
// bind tls
xor eax, eax;
inc eax;
push eax; // arg2, reason for tls
mov eax, [ebx + dllrva]
add eax, edi;
push eax; // arg1, dllbase
mov eax, [ebx + membindtlsrva];
add eax, edi;
call eax;
// call dll oep, for dll entry
xor eax, eax;
push eax; // lpvReserved
inc eax;
push eax; // fdwReason, DLL_PROCESS_ATTACH
mov eax, [ebx + dllrva];
add eax, edi;
push eax; // hinstDLL
mov eax, [ebx + dlloeprva];
add eax, edi;
call eax;
// jmp to origin oep
mov eax, [ebx+exeoeprva];
add eax, edi;
jmp eax;
getip:
mov eax, [esp]
ret
exeoeprva: nop;nop;nop;nop;
dllrva: nop;nop;nop;nop;
dlloeprva: nop;nop;nop;nop;
memrelocrva: nop;nop;nop;nop;
membindiatrva: nop;nop;nop;nop;
membindtlsrva: nop;nop;nop;nop;
findloadlibrarya: nop;nop;nop;nop;
findgetprocaddress: nop;nop;nop;nop;
// these functions are stub function, will be filled by python
#define FUNC_SIZE 0x400
#define SHELLCODE_SIZE 0X2000
unsigned char g_oepinit_code[] = {0x90};
unsigned char g_memreloc_code[] = {0x90};
unsigned char g_membindiat_code[] = {0x90};
unsigned char g_membindtls_code[] = {0x90};
unsigned char g_findloadlibrarya_code[] = {0x90};
unsigned char g_findgetprocaddress_code[] = {0x90};
void _makeoepcode(void *shellcode,
size_t shellcoderva, size_t dllrva,
DWORD orgexeoeprva, DWORD orgdlloeprva)
{
// bind the pointer to buffer
size_t oepinit_end = sizeof(g_oepinit_code);
size_t memreloc_start = FUNC_SIZE;
size_t membindiat_start = memreloc_start + FUNC_SIZE;
size_t membindtls_start = membindiat_start + FUNC_SIZE;
size_t findloadlibrarya_start = membindtls_start + FUNC_SIZE;
size_t findgetprocaddress_start = findloadlibrarya_start + FUNC_SIZE;
// fill the address table
size_t *pexeoeprva = (size_t*)(g_oepinit_code + oepinit_end - 8*sizeof(size_t));
size_t *pdllbrva = (size_t*)(g_oepinit_code + oepinit_end - 7*sizeof(size_t));
size_t *pdlloeprva = (size_t*)(g_oepinit_code + oepinit_end - 6*sizeof(size_t));
size_t *pmemrelocrva = (size_t*)(g_oepinit_code + oepinit_end - 5*sizeof(size_t));
size_t *pmembindiatrva = (size_t*)(g_oepinit_code + oepinit_end - 4*sizeof(size_t));
size_t *pmembindtlsrva = (size_t*)(g_oepinit_code + oepinit_end - 3*sizeof(size_t));
size_t *pfindloadlibrarya = (size_t*)(g_oepinit_code + oepinit_end - 2*sizeof(size_t));
size_t *pfindgetprocaddress = (size_t*)(g_oepinit_code + oepinit_end - 1*sizeof(size_t));
*pexeoeprva = orgexeoeprva;
*pdllbrva = dllrva;
*pdlloeprva = dllrva + orgdlloeprva;
*pmemrelocrva = shellcoderva + memreloc_start;
*pmembindiatrva = shellcoderva + membindiat_start;
*pmembindtlsrva = shellcoderva + membindtls_start;
*pfindloadlibrarya = shellcoderva + findloadlibrarya_start;
*pfindgetprocaddress = shellcoderva + findgetprocaddress_start;
// copy to the target
memcpy(shellcode ,
g_oepinit_code, sizeof(g_oepinit_code));
memcpy(shellcode + memreloc_start,
g_memreloc_code, sizeof(g_memreloc_code));
memcpy(shellcode + membindiat_start,
g_membindiat_code, sizeof(g_membindiat_code));
memcpy(shellcode + membindtls_start,
g_membindtls_code, sizeof(g_membindtls_code));
memcpy(shellcode + findloadlibrarya_start,
g_findloadlibrarya_code, sizeof(g_findloadlibrarya_code));
memcpy(shellcode + findgetprocaddress_start,
g_findgetprocaddress_code, sizeof(g_findgetprocaddress_code));
}
0x4 技巧tips
(1) x64适配
由于大部分都用的size_t或void*等与cpu字长相关的变量，大部分代码不用修改就能编译。
其余的差异部分，可以用sizeof(size_t)>4区分，或者_WIN64宏。如下所示
// append section header to exe
size_t align = sizeof(size_t) > 4 ? 0x10000: 0x1000;
size_t padding = _sectpaddingsize(mempe_exe, mempe_dll, align);
secth.Characteristics = IMAGE_SCN_MEM_READ |
IMAGE_SCN_MEM_WRITE | IMAGE_SCN_MEM_EXECUTE;
secth.Misc.VirtualSize = SHELLCODE_SIZE + padding + mempe_dllsize;
secth.SizeOfRawData = SHELLCODE_SIZE + padding + mempe_dllsize;
strcpy((char*)secth.Name, ".module");
winpe_noaslr(mempe_exe);
winpe_appendsecth(mempe_exe, §h);
inline void* winpe_findkernel32()
{
// return (void*)LoadLibrary("kernel32.dll");
// TEB->PEB->Ldr->InMemoryOrderLoadList->curProgram->ntdll->kernel32
void *kerenl32 = NULL;
#ifdef _WIN64
__asm{
mov rax, gs:[60h]; peb
mov rax, [rax+18h]; ldr
mov rax, [rax+20h]; InMemoryOrderLoadList, currentProgramEntry
mov rax, [rax]; ntdllEntry, currentProgramEntry->->Flink
mov rax, [rax]; kernel32Entry, ntdllEntry->Flink
mov rax, [rax-10h+30h]; kernel32.DllBase
mov kerenl32, rax;
}
#else
__asm{
mov eax, fs:[30h]; peb
mov eax, [eax+0ch]; ldr
mov eax, [eax+14h]; InMemoryOrderLoadList, currentProgramEntry
mov eax, [eax]; ntdllEntry, currentProgramEntry->->Flink
mov eax, [eax]; kernel32Entry, ntdllEntry->Flink
mov eax, [eax - 8h +18h]; kernel32.DllBase
mov kerenl32, eax;
}
#endif
return kerenl32;
}
(2) 扫描export找到LoadLibrary和GetProcAddress
上面的代码我们可以获取kernel32的基址了，之后可以自己通过扫描export表来查找对应的导出函数了。
这里注意有个坑点，有的导出表项会forward到其他dll上，如kernel32!InitializeSListHead -> NTDLL!RtlInitializeSListHead，详见crashing-kernel32-initializeslisthead。通过DirRVA < FuncRVA < DirRVA + ExportSize来判断是否forward。
此处为了找到LoadLibraryA和GetProcAddress函数，不考虑forward情况，否则会引入递归问题。
inline void* STDCALL winpe_memfindexp(
void *mempe, LPCSTR funcname)
{
PIMAGE_DOS_HEADER pDosHeader = (PIMAGE_DOS_HEADER)mempe;
PIMAGE_NT_HEADERS pNtHeader = (PIMAGE_NT_HEADERS)
((void*)mempe + pDosHeader->e_lfanew);
PIMAGE_FILE_HEADER pFileHeader = &pNtHeader->FileHeader;
PIMAGE_OPTIONAL_HEADER pOptHeader = &pNtHeader->OptionalHeader;
PIMAGE_DATA_DIRECTORY pDataDirectory = pOptHeader->DataDirectory;
PIMAGE_DATA_DIRECTORY pExpEntry =
&pDataDirectory[IMAGE_DIRECTORY_ENTRY_EXPORT];
PIMAGE_EXPORT_DIRECTORY pExpDescriptor =
(PIMAGE_EXPORT_DIRECTORY)(mempe + pExpEntry->VirtualAddress);
WORD *ordrva = mempe + pExpDescriptor->AddressOfNameOrdinals;
DWORD *namerva = mempe + pExpDescriptor->AddressOfNames;
DWORD *funcrva = mempe + pExpDescriptor->AddressOfFunctions;
if((size_t)funcname <= MAXWORD) // find by ordnial
{
WORD ordbase = LOWORD(pExpDescriptor->Base) - 1;
WORD funcord = LOWORD(funcname);
return (void*)(mempe + funcrva[ordrva[funcord-ordbase]]);
}
else
{
for(int i=0;i<pExpDescriptor->NumberOfNames;i++)
{
LPCSTR curname = (LPCSTR)(mempe+namerva[i]);
if(_inl_stricmp(curname, funcname)==0)
{
return (void*)(mempe + funcrva[ordrva[i]]);
}
}
}
return NULL;
}
inline PROC winpe_findloadlibrarya()
{
// return (PROC)LoadLibraryA;
HMODULE hmod_kernel32 = (HMODULE)winpe_findkernel32();
char name_LoadLibraryA[] = {'L', 'o', 'a', 'd', 'L', 'i', 'b', 'r', 'a', 'r', 'y', 'A', '\0'};
return (PROC)winpe_memfindexp( // suppose exp no forward, to avoid recursive
(void*)hmod_kernel32, name_LoadLibraryA);
}
inline PROC winpe_findgetprocaddress()
{
// return (PROC)GetProcAddress;
HMODULE hmod_kernel32 = (HMODULE)winpe_findkernel32();
char name_GetProcAddress[] = {'G', 'e', 't', 'P', 'r', 'o', 'c', 'A', 'd', 'd', 'r', 'e', 's', 's', '\0'};
return (PROC)winpe_memfindexp(hmod_kernel32, name_GetProcAddress);
}
0x5 后记
本项目是我实现不加壳vfs（即通过hook nt函数将文件和文件夹等读取重定向到zip文件中）的衍生产物，采取的是尽量单头文件法和内联函数减小耦合，方便集成到代码中的做法。与其他的MemoryModule最大的区别是支持从shellcode引导DLL。由于以前我有了自己写压缩壳的经验，整体上处理起来还算是顺利。强迫症，感觉好多时间都花在了怎么统一化命名api和代码风格规范上了。至于实现内存加载Resource方面，暂时不考虑实现，汉化游戏一般也用不到这方面内容。
另外，这里还有个坑卡了我好长时间，就是x64的堆栈必须要0x10对齐（即调用函数时要sub rsp, 0x28预留空间与对齐），才能在movaps指令上不报错。之前一直不知道这条指令需要对齐，一直以为是少了哪些初始化，走偏了好久。特别感谢@YeLikERs指出了这一点，同时在PEB，LDR相关内容上帮助了很多。
经测试兼容性还算不错，windows xp, windows 7, windows 10, windows 11测试example文件都通过了，甚至我连linux上的wine也测试了（即用我汉化的游戏测试shellcode引导附加到exe区段上的DLL）。最后，附上几张测试图：
[复制链接]
发表于 2022-3-29 18:36
