标签：utf8 build 意义 sed 其他 one dex typename 开启
好久没弄ollvm了，可以继续了，今天给ollvm新增了一个pass，用来加密字符串，这个pass是从别的库里面扒出来的。
本文是基于在Windows 上使用VS2017编译出来的ollvm，在这个基础上来添加。
第一步：
寻找两个pass的代码
头文件
1 #ifndef _STRING_OBFUSCATION_H_ 2 #define _STRING_OBFUSCATION_H_ 3 4 5 // LLVM include 6 #include "llvm/Pass.h" 7 #include "llvm/IR/Function.h" 8 #include "llvm/IR/Instructions.h" 9 #include "llvm/ADT/Statistic.h" 10 #include "llvm/Transforms/IPO.h" 11 #include "llvm/IR/Module.h" 12 #include "llvm/Support/CommandLine.h" 13 #include "llvm/CryptoUtils.h" 14 15 // Namespace 16 using namespace llvm; 17 using namespace std; 18 19 namespace llvm { 20 Pass *createStringObfuscation(bool flag); 21 } 22 23 #endif
源文件
1 #define DEBUG_TYPE "objdiv" 2 #include <string> 3 #include <sstream> 4 5 #include "llvm/ADT/Statistic.h" 6 #include "llvm/IR/Function.h" 7 #include "llvm/IR/Constants.h" 8 #include "llvm/IR/Module.h" 9 #include "llvm/IR/Value.h" 10 #include "llvm/Pass.h" 11 #include "llvm/Support/raw_ostream.h" 12 #include "llvm/CryptoUtils.h" 13 #include "llvm/Transforms/Obfuscation/StringObfuscation.h" 14 #include "llvm/IR/IRBuilder.h" 15 #include "llvm/Transforms/Utils/ModuleUtils.h" 16 17 using namespace llvm; 18 19 STATISTIC(GlobalsEncoded, "Counts number of global variables encoded"); 20 21 #define ZooPrint(_F, ...) fprintf(stdout, "File : [%s](%d) " _F, __FILE__, __LINE__, __VA_ARGS__) 22 23 namespace llvm { 24 25 struct encVar { 26 public: 27 GlobalVariable *var; 28 uint8_t key; 29 }; 30 31 class StringObfuscationPass : public llvm::ModulePass { 32 public: 33 static char ID; // pass identification 34 bool is_flag = false; 35 StringObfuscationPass() : ModulePass(ID) {} 36 StringObfuscationPass(bool flag) : ModulePass(ID) 37 { 38 is_flag = flag; 39 } 40 41 virtual bool runOnModule(Module &M) { 42 ZooPrint(" Run On Module : %d \n", is_flag); 43 if (!is_flag) 44 return false; 45 std::vector<GlobalVariable*> toDelConstGlob; 46 //std::vector<GlobalVariable*> encGlob; 47 std::vector<encVar*> encGlob; 48 ZooPrint(" M.Size : %d \n", M.size()); 49 int i = 0; 50 for (Module::global_iterator gi = M.global_begin(), ge = M.global_end(); gi != ge; ++gi) 51 { 52 53 #if 0 54 // 老式代码，原来的样子 55 @.str = private unnamed_addr constant[13 x i8] c"\E4\BD\A0\E5\A5\BD\E4\B8\96\E7\95\8C\00", align 1 56 @__CFConstantStringClassReference = external global[0 x i32] 57 @.str.1 = private unnamed_addr constant[3 x i16][i16 20320, i16 22909, i16 0], section "__TEXT,__ustring", align 2 58 // 新式字符串的样子 59 @"\01??_C@_07CHPFNFHA@123456?6?$AA@" = linkonce_odr unnamed_addr constant [8 x i8] c"123456\0A\00", comdat, align 1 60 @"\01??_C@_03PMGGPEJJ@?$CFd?6?$AA@" = linkonce_odr unnamed_addr constant [4 x i8] c"%d\0A\00", comdat, align 1 61 @__local_stdio_printf_options._OptionsStorage = internal global i64 0, align 8 62 #endif 63 // Loop over all global variables 64 GlobalVariable* gv = &(*gi); 65 //errs() << "Global var " << gv->getName(); 66 //std::string::size_type str_idx = gv->getName().str().find(".str."); 67 std::string section(gv->getSection()); 68 69 ZooPrint(" %d : String : \"%s\" ， section : \"%s\" ， isConstant : %d ， hasInitializer : %d ， isa : %d ， r : %d \n", i++, gv->getName().str().c_str(), section.c_str(), gv->isConstant(), gv->hasInitializer(), isa<ConstantDataSequential>(gv->getInitializer()), gv->getName().str().substr(0, 8) == "\"\x01??_C@_"); 70 // ZooPrint(" 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X 0x%02X \n", gv->getName()[0] & 0xFF, gv->getName()[1] & 0xFF, gv->getName()[2] & 0xFF, gv->getName()[3] & 0xFF, gv->getName()[4] & 0xFF, gv->getName()[5] & 0xFF, gv->getName()[6] & 0xFF, gv->getName()[7] & 0xFF); 71 72 // Let‘s encode the static ones 73 //if (gv->getName().str().substr(0, 4) == ".str"&& 74 if (gv->getName().str().substr(0, 7) == "\x01??_C@_" && 75 gv->isConstant() && 76 gv->hasInitializer() && 77 isa<ConstantDataSequential>(gv->getInitializer()) && 78 section != "llvm.metadata" && 79 section.find("__objc_methname") == std::string::npos 80 /*&&gv->getType()->getArrayElementType()->getArrayElementType()->isIntegerTy()*/) 81 { 82 ZooPrint(" In Global Encode \n"); 83 ++GlobalsEncoded; 84 //errs() << " is constant"; 85 86 // Duplicate global variable 87 GlobalVariable *dynGV = new GlobalVariable(M, 88 gv->getType()->getElementType(), 89 !(gv->isConstant()), gv->getLinkage(), 90 (Constant*)0, gv->getName(), 91 (GlobalVariable*)0, 92 gv->getThreadLocalMode(), 93 gv->getType()->getAddressSpace()); 94 // dynGV->copyAttributesFrom(gv); 95 dynGV->setInitializer(gv->getInitializer()); 96 97 std::string tmp = gv->getName().str(); 98 // errs()<<"GV: "<<*gv<<"\n"; 99 100 Constant *initializer = gv->getInitializer(); 101 ConstantDataSequential *cdata = dyn_cast<ConstantDataSequential>(initializer); 102 if (cdata) { 103 const char *orig = cdata->getRawDataValues().data(); 104 unsigned len = cdata->getNumElements()*cdata->getElementByteSize(); 105 106 encVar *cur = new encVar(); 107 cur->var = dynGV; 108 cur->key = llvm::cryptoutils->get_uint8_t(); 109 // casting away const is undef. behavior in C++ 110 // TODO a clean implementation would retrieve the data, generate a new constant 111 // set the correct type, and copy the data over. 112 //char *encr = new char[len]; 113 //Constant *initnew = ConstantDataArray::getString(M.getContext(), encr, true); 114 char *encr = const_cast<char *>(orig); 115 // Simple xor encoding 116 for (unsigned i = 0; i != len; ++i) { 117 encr[i] = orig[i] ^ cur->key; 118 } 119 120 // FIXME Second part of the unclean hack. 121 dynGV->setInitializer(initializer); 122 123 // Prepare to add decode function for this variable 124 encGlob.push_back(cur); 125 } 126 else { 127 // just copying default initializer for now 128 dynGV->setInitializer(initializer); 129 } 130 131 // redirect references to new GV and remove old one 132 gv->replaceAllUsesWith(dynGV); 133 toDelConstGlob.push_back(gv); 134 135 } 136 } 137 138 // actuallte delete marked globals 139 for (unsigned i = 0, e = toDelConstGlob.size(); i != e; ++i) 140 toDelConstGlob[i]->eraseFromParent(); 141 142 addDecodeFunction(&M, &encGlob); 143 144 145 return true; 146 } 147 148 private: 149 void addDecodeFunction(Module *mod, std::vector<encVar*> *gvars) { 150 ZooPrint(" Add Decode Function \n"); 151 // Declare and add the function definition 152 //errs()<<"Successful enter decode function"<<"\n"; 153 std::vector<Type*>FuncTy_args; 154 FunctionType* FuncTy = FunctionType::get( 155 /*Result=*/Type::getVoidTy(mod->getContext()), // returning void 156 /*Params=*/FuncTy_args, // taking no args 157 /*isVarArg=*/false); 158 uint64_t StringObfDecodeRandomName = cryptoutils->get_uint64_t(); 159 std::string random_str; 160 std::stringstream random_stream; 161 random_stream << StringObfDecodeRandomName; 162 random_stream >> random_str; 163 StringObfDecodeRandomName++; 164 Constant* c = mod->getOrInsertFunction(".datadiv_decode" + random_str, FuncTy); 165 Function* fdecode = cast<Function>(c); 166 fdecode->setCallingConv(CallingConv::C); 167 168 169 BasicBlock* entry = BasicBlock::Create(mod->getContext(), "entry", fdecode); 170 171 IRBuilder<> builder(mod->getContext()); 172 builder.SetInsertPoint(entry); 173 174 175 for (unsigned i = 0, e = gvars->size(); i != e; ++i) { 176 GlobalVariable *gvar = (*gvars)[i]->var; 177 uint8_t key = (*gvars)[i]->key; 178 179 Constant *init = gvar->getInitializer(); 180 ConstantDataSequential *cdata = dyn_cast<ConstantDataSequential>(init); 181 182 unsigned len = cdata->getNumElements()*cdata->getElementByteSize(); 183 --len; 184 185 BasicBlock *preHeaderBB = builder.GetInsertBlock(); 186 BasicBlock* for_body = BasicBlock::Create(mod->getContext(), "for-body", fdecode); 187 BasicBlock* for_end = BasicBlock::Create(mod->getContext(), "for-end", fdecode); 188 builder.CreateBr(for_body); 189 builder.SetInsertPoint(for_body); 190 PHINode *variable = builder.CreatePHI(Type::getInt32Ty(mod->getContext()), 2, "i"); 191 Value *startValue = builder.getInt32(0); 192 Value *endValue = builder.getInt32(len); 193 variable->addIncoming(startValue, preHeaderBB); 194 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// 195 196 //LoadInst *Load=builder.CreateLoad(gvar); 197 //errs()<<"Load: "<<*(Load->getPointerOperand())<<"\n"; 198 Value* indexList[2] = { ConstantInt::get(variable->getType(), 0), variable }; 199 Value *const_key = builder.getInt8(key); 200 Value *GEP = builder.CreateGEP(gvar, ArrayRef<Value*>(indexList, 2), "arrayIdx"); 201 LoadInst *loadElement = builder.CreateLoad(GEP, false); 202 loadElement->setAlignment(1); 203 //errs()<<"Type: "<<*loadElement<<"\n"; 204 //CastInst* extended = new ZExtInst(const_key, loadElement->getType(), "extended", for_body); 205 //Value* extended = builder.CreateZExtOrBitCast(const_key, loadElement->getType(),"extended"); 206 Value *Xor = builder.CreateXor(loadElement, const_key, "xor"); 207 StoreInst *Store = builder.CreateStore(Xor, GEP, false); 208 Store->setAlignment(1); 209 210 /////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// 211 Value *stepValue = builder.getInt32(1); 212 Value *nextValue = builder.CreateAdd(variable, stepValue, "next-value"); 213 Value *endCondition = builder.CreateICmpULT(variable, endValue, "end-condition"); 214 endCondition = builder.CreateICmpNE(endCondition, builder.getInt1(0), "loop-condition"); 215 BasicBlock *loopEndBB = builder.GetInsertBlock(); 216 builder.CreateCondBr(endCondition, loopEndBB, for_end); 217 builder.SetInsertPoint(for_end); 218 variable->addIncoming(nextValue, loopEndBB); 219 220 } 221 builder.CreateRetVoid(); 222 appendToGlobalCtors(*mod, fdecode, 0); 223 224 225 } 226 227 }; 228 229 Pass *createStringObfuscation(bool flag); 230 } 231 232 #if 0 233 RegisterPass(const char *PassArg, const char *Name, bool CFGOnly = false, bool is_analysis = false) 234 235 上面这个是RegisterPass的构造函数。 236 参数说明： 237 238 template<typename passName> ：YourPassName； 239 PassArg ：opt调用时所用的命行参数； 240 Name ：此pass的简要说明； 241 CFGOnly ：如果一个遍历CFG而不修改它，那么这个参数被设置为true； 242 is_analysis ：如果一个Pass是一个分析Pass，例如dominator tree pass，那么这个参数被设置为true。 243 例子： 244 245 static RegisterPass<Hello> X("hello", "Hello World Pass", false, false); 246 #endif 247 248 char StringObfuscationPass::ID = 0; 249 static RegisterPass<StringObfuscationPass> X("GVDiv", "Global variable (i.e., const char*) diversification pass", false, true); 250 251 Pass * llvm::createStringObfuscation(bool flag) { 252 ZooPrint("new my pass \n"); 253 return new StringObfuscationPass(flag); 254 }
第二步：
将头文件放在如下位置：ollvm\obfuscator-llvm-4.0\include\llvm\Transforms\Obfuscation\StringObfuscation.h
将源文件放在如下位置：ollvm\obfuscator-llvm-4.0\lib\Transforms\Obfuscation\StringEncode.cpp
第三步：
将源文件放到如下工程中
第四步：
在此文件中新增代码：ollvm\obfuscator-llvm-4.0\lib\Transforms\IPO\PassManagerBuilder.cpp
新增导入头文件
1 #include "llvm/Transforms/Obfuscation/StringObfuscation.h"
新增全局变量代码如下
1 static cl::opt<std::string> Seed("seed", cl::init(""), 2 cl::desc("seed for the random")); 3 4 // 全局开关，根据参数判断是否设置 5 static cl::opt<bool> StringObf("sobf", cl::init(false), 6 cl::desc("Enable the string obfuscation"));
在：PassManagerBuilder::populateModulePassManager 函数中，新增挂载新的pass代码，如下
1 MPM.add(createStringObfuscation(StringObf));
意义为根据全局开关来判断是否启用当前pass
经过以上四步，问题全部解决了，直接重新编译ollvm即可。
后续可以修改pass代码，可以修改解密函数。
新增其他pass新增步骤也如上。
使用方式如下
1 G:\ollvm\Test>G:\ollvm\build\RelWithDebInfo\bin\clang.exe -mllvm -sobf -mllvm -fla main.c
含义是，开启字符串加密，并且启动代码扁平化
效果：
源代码如下
编译后如下
已经开启了代码扁平化，原始字符串也已经不一样了，具体情况，
看data段就好了：
已经完全没个人样了
重点在最后，忘了，补充一句，由于字符串在ollvm里面是以UTF8的格式保存的，所以中文字符串天然就是乱码，
有时间想办法来解决一下中文字符串的乱码问题。
这个pass挺简单的，我也没怎么改就拿来用了，后续有时间再改一下吧。
标签：utf8 build 意义 sed 其他 one dex typename 开启
原文地址：https://www.cnblogs.com/suanguade/p/11386337.html
