X-Git-Url: https://git.siccegge.de//index.cgi?p=frida%2Ffrida.git;a=blobdiff_plain;f=src%2Fdisassembler%2Fllvm%2FLLVMDisassembler.cxx;h=1ae024a5aea39f9ef6dd786e3933768e8cc6cf5c;hp=e60e0c28bf635e0f0ac21f6fd34232f2ec4d3251;hb=70d986e452c08aea59d947bfd5700f6c60b3f820;hpb=d84d4109b1f5bd5f87732b7e49860639d963a553 diff --git a/src/disassembler/llvm/LLVMDisassembler.cxx b/src/disassembler/llvm/LLVMDisassembler.cxx index e60e0c2..1ae024a 100644 --- a/src/disassembler/llvm/LLVMDisassembler.cxx +++ b/src/disassembler/llvm/LLVMDisassembler.cxx @@ -7,13 +7,13 @@ using namespace llvm; using namespace llvm::object; +using std::error_code; /* * TODO: fallback code falls die Datei kein ELF/PE/COFF/MacO/.. binary * ist sondern z.B. einfach nur Instruktionen oder ein Bootsektor oder * foo */ - LLVMDisassembler::LLVMDisassembler(const std::string& filename) : Disassembler(filename) , logger(log4cxx::Logger::getLogger("LLVMDisassembler")) @@ -73,24 +73,30 @@ LLVMDisassembler::LLVMDisassembler(const std::string& filename) return; } - DisAsm.reset(target->createMCDisassembler(*STI)); + MOFI.reset(new MCObjectFileInfo); + MCContext Ctx(AsmInfo.get(), MRI.get(), MOFI.get()); + + DisAsm.reset(target->createMCDisassembler(*STI, Ctx)); if (!DisAsm) { LOG4CXX_ERROR(logger, "no disassembler for target " << tripleName); return; } - - MOFI.reset(new MCObjectFileInfo); - Ctx.reset(new MCContext(AsmInfo.get(), MRI.get(), MOFI.get())); RelInfo.reset( - target->createMCRelocationInfo(tripleName, *Ctx.get())); + target->createMCRelocationInfo(tripleName, Ctx)); if (RelInfo) { Symzer.reset( - MCObjectSymbolizer::createObjectSymbolizer(*Ctx.get(), RelInfo, o)); + MCObjectSymbolizer::createObjectSymbolizer(Ctx, std::move(RelInfo), o)); if (Symzer) - DisAsm->setSymbolizer(Symzer); + DisAsm->setSymbolizer(std::move(Symzer)); } + RelInfo.release(); + Symzer.release(); MIA.reset(target->createMCInstrAnalysis(MII.get())); + if (!MIA) { + LOG4CXX_ERROR(logger, "no instruction analysis for target " << tripleName); + return; + } int AsmPrinterVariant = AsmInfo->getAssemblerDialect(); IP.reset(target->createMCInstPrinter(AsmPrinterVariant, *AsmInfo, *MII, *MRI, *STI)); @@ -102,7 +108,7 @@ LLVMDisassembler::LLVMDisassembler(const std::string& filename) IP->setPrintImmHex(llvm::HexStyle::C); IP->setPrintImmHex(true); - OwningPtr OD( + std::unique_ptr OD( new MCObjectDisassembler(*o, *DisAsm, *MIA)); Mod.reset(OD->buildModule(false)); @@ -111,30 +117,77 @@ LLVMDisassembler::LLVMDisassembler(const std::string& filename) disassemble(); } +LLVMDisassembler::~LLVMDisassembler() { + std::for_each(functions.begin(), functions.end(), + [](std::pair it) { + delete it.second; + }); + std::for_each(blocks.begin(), blocks.end(), + [](std::pair it) { + delete it.second; + }); +} void LLVMDisassembler::disassemble() { std::stack remaining_functions; std::stack remaining_blocks; SectionRef text_section = sections[".text"]; - std::for_each(symbols.begin(), symbols.end(), - [&](std::pair x) { - uint64_t result; - bool contains; - SymbolRef::Type symbol_type; + // Assume all function symbols actually start a real function + for (auto x = symbols.begin(); x != symbols.end(); ++x) { + uint64_t result; + bool contains; + SymbolRef::Type symbol_type; - if (text_section.containsSymbol(x.second, contains) || !contains) - return; - if (x.second.getType(symbol_type) - || SymbolRef::ST_Function != symbol_type) - return; + if (text_section.containsSymbol(x->second, contains) || !contains) + continue; - if (!x.second.getAddress(result)) { - remaining_functions.push(new LLVMFunction(x.first, result)); - LOG4CXX_DEBUG(logger, "Disasembling " << x.first); - } - }); + if (x->second.getType(symbol_type) + || SymbolRef::ST_Function != symbol_type) + continue; + + if (!x->second.getAddress(result)) { + LLVMFunction * fun = new LLVMFunction(x->first, result); + remaining_functions.push(fun); + functions.insert(std::make_pair(result, fun)); + LOG4CXX_DEBUG(logger, "Disasembling " << x->first); + } + } + + if (binary->isELF()) { + bool is64bit = (binary->getData()[4] == 0x02); + + uint64_t entry(0); + for (int i(0); i < (is64bit? 8 : 4); ++i) { + if (binary->isLittleEndian()) { + entry |= (unsigned int)((unsigned char)binary->getData()[0x18 + i]) << 8*i; + } else { + entry = entry << 8; + entry |= (unsigned char)binary->getData()[0x18 + i]; + } + } + if (functions.find(entry) == functions.end()) { + LOG4CXX_DEBUG(logger, "Adding entry at: " << std::hex << entry); + std::stringstream s; + s << "<_start 0x" << std::hex << entry << ">"; + LLVMFunction * fun = new LLVMFunction(s.str(), entry); + functions.insert(std::make_pair(entry, fun)); + remaining_functions.push(fun); + } + } + + if (functions.empty()) { + uint64_t text_entry; + text_section.getAddress(text_entry); + LOG4CXX_INFO(logger, "No Symbols found, starting at the beginning of the text segment"); + + std::stringstream s; + s << ""; + LLVMFunction * fun = new LLVMFunction(s.str(), text_entry); + functions.insert(std::make_pair(text_entry, fun)); + remaining_functions.push(fun); + } StringRef bytes; text_section.getContents(bytes); @@ -144,18 +197,17 @@ void LLVMDisassembler::disassemble() { LLVMFunction * current_function = remaining_functions.top(); remaining_functions.pop(); - LOG4CXX_INFO(logger, "Handling function " << current_function->getName()); + LOG4CXX_DEBUG(logger, "Handling function " << current_function->getName()); - // if ("_start" != current_function->getName()) - // continue; - - remaining_blocks.push(new LLVMBasicBlock(current_function->getStartAddress())); + LLVMBasicBlock * block = new LLVMBasicBlock(current_function->getStartAddress(), this); + remaining_blocks.push(block); + blocks.insert(std::make_pair(block->getStartAddress(), block)); while (remaining_blocks.size()) { LLVMBasicBlock * current_block = remaining_blocks.top(); remaining_blocks.pop(); - LOG4CXX_INFO(logger, "Handling Block starting at " << std::hex << current_block->getStartAddress()); + LOG4CXX_DEBUG(logger, "Handling Block starting at " << std::hex << current_block->getStartAddress()); uint64_t inst_size; uint64_t base_address; @@ -168,35 +220,34 @@ void LLVMDisassembler::disassemble() { if(llvm::MCDisassembler::Success == DisAsm->getInstruction(inst, inst_size, ref, current_address, nulls(), nulls())) { - LOG4CXX_DEBUG(logger, "Inst Size " << inst_size); - - uint8_t bytes[inst_size+2]; - ref.readBytes(current_address, inst_size, bytes); - s << '\t'; - for(uint8_t* cur = bytes; cur < bytes + inst_size; ++cur) { - s.write_hex(*cur); - s << ' '; - } - s << '\t'; - - IP->printInst(&inst, s, ""); - - LOG4CXX_DEBUG(logger, std::hex << current_address + base_address << s.str()); uint64_t jmptarget; if (MIA->evaluateBranch(inst, current_address, inst_size, jmptarget)) { - jmptarget += base_address; + jmptarget += base_address; if (!MIA->isIndirectBranch(inst)) { if (MIA->isCall(inst)) { - if (blocks.find(jmptarget) == blocks.end()) - remaining_functions.push(new LLVMFunction("", jmptarget)); + if (functions.find(jmptarget) == functions.end()) { + std::stringstream s; + s << ""; + LLVMFunction * fun = new LLVMFunction(s.str(), jmptarget); + functions.insert(std::make_pair(jmptarget, fun)); + remaining_functions.push(fun); + } } else { - if (blocks.find(jmptarget) == blocks.end()) - remaining_blocks.push(new LLVMBasicBlock(jmptarget)); + current_block->setNextBlock(0, jmptarget); + if (blocks.find(jmptarget) == blocks.end()) { + LLVMBasicBlock * block = new LLVMBasicBlock(jmptarget, this); + blocks.insert(std::make_pair(block->getStartAddress(), block)); + remaining_blocks.push(block); + } if (MIA->isConditionalBranch(inst)) { - jmptarget = base_address + current_address + inst_size; - if (blocks.find(jmptarget) == blocks.end()) - remaining_blocks.push(new LLVMBasicBlock(jmptarget)); + jmptarget = base_address + current_address + inst_size; + current_block->setNextBlock(1, jmptarget); + if (blocks.find(jmptarget) == blocks.end()) { + LLVMBasicBlock * block = new LLVMBasicBlock(jmptarget, this); + blocks.insert(std::make_pair(block->getStartAddress(), block)); + remaining_blocks.push(block); + } } } } @@ -207,16 +258,52 @@ void LLVMDisassembler::disassemble() { if (inst_size == 0 || MIA->isTerminator(inst) || MIA->isBranch(inst)) { - current_block->setEndAddress(current_address + base_address); - blocks.insert(std::make_pair(current_block->getStartAddress(), current_block)); - LOG4CXX_INFO(logger, "Finished Block at " << current_block->getEndAddress()); + current_block->setEndAddress(current_address + base_address + inst_size); + LOG4CXX_DEBUG(logger, "Finished Block at " << std::hex << + current_block->getEndAddress()); break; } current_address += inst_size; } } - LOG4CXX_INFO(logger, "Finished function " << current_function->getName()); + LOG4CXX_DEBUG(logger, "Finished function " << current_function->getName()); } + + // Split blocks where jumps are going inside the block + for (auto it = blocks.begin(); it != blocks.end(); ++it) { + LLVMBasicBlock * current_block = it->second; + uint64_t inst_size; + uint64_t base_address; + text_section.getAddress(base_address); + uint64_t current_address = current_block->getStartAddress() - base_address; + while(current_block->getEndAddress() - base_address > current_address) { + MCInst inst; + std::string buf; + llvm::raw_string_ostream s(buf); + + if(llvm::MCDisassembler::Success == + DisAsm->getInstruction(inst, inst_size, ref, current_address, nulls(), nulls())) { + auto other = blocks.find(current_address + inst_size + base_address); + + if (other != blocks.end()) { + uint64_t endaddress = current_address + inst_size + base_address; + if (endaddress != current_block->getEndAddress()) { + LOG4CXX_DEBUG(logger, "Shortening block starting at " + << std::hex + << current_block->getStartAddress() + << " now ending at " + << other->first); + current_block->setEndAddress(endaddress); + current_block->setNextBlock(0, other->first); + current_block->setNextBlock(1, 0); + } + } + } else { + inst_size = 1; + } + current_address += inst_size; + } + } } void LLVMDisassembler::readSymbols() { @@ -248,6 +335,51 @@ void LLVMDisassembler::readSections() { } -BasicBlock * LLVMDisassembler::generateControlFlowGraph(uint64_t address) { +void LLVMDisassembler::forEachFunction(std::function callback) { + std::for_each(functions.begin(), functions.end(), + [&](std::pair x) { + callback(x.first, x.second); + }); +} +void LLVMDisassembler::printEachInstruction(uint64_t start, uint64_t end, + std::function fun) { + SectionRef text_section = sections[".text"]; + uint64_t base_address; + text_section.getAddress(base_address); + uint64_t current_address = start - base_address; + + StringRef bytes; + text_section.getContents(bytes); + StringRefMemoryObject ref(bytes); + + while (current_address < end - base_address) { + uint64_t inst_size; + MCInst inst; + std::string buf; + llvm::raw_string_ostream s(buf); + + if(llvm::MCDisassembler::Success == + DisAsm->getInstruction(inst, inst_size, ref, current_address, nulls(), nulls())) { + + uint8_t bytes[inst_size+2]; + ref.readBytes(current_address, inst_size, bytes); + + uint64_t jmptarget; + if (MIA->evaluateBranch(inst, current_address, inst_size, jmptarget)) { + std::stringstream stream; + stream << std::hex << (base_address + jmptarget); + IP->printInst(&inst, s, stream.str()); + } else + IP->printInst(&inst, s, ""); + + fun(bytes, inst_size, s.str()); + } else { + LOG4CXX_WARN(logger, "Invalid byte at" << std::hex << current_address + base_address); + fun(NULL, 0, "Invalid Byte"); + inst_size = 1; + } + + current_address += inst_size; + } }