+template <typename ELFT>
+void LLVMDisassembler<ELFT>::disassembleFunction(Function* function) {
+ std::vector<uint64_t> called_functions;
+ std::stack<BasicBlock*> remaining_blocks;
+ /* TODO:
+ * Do all blocks get added properly? We should take care to remove
+ * the other ones at the end of the function!
+ */
+ std::map<uint64_t, BasicBlock*> new_blocks;
+ SectionRef text_section = sections[".text"];
+ StringRef bytes;
+ text_section.getContents(bytes);
+ StringRefMemoryObject ref(bytes);
+
+ LOG4CXX_DEBUG(logger, "Handling function " << function->getName());
+
+ BasicBlock * block = manager->newBasicBlock(function->getStartAddress());
+ remaining_blocks.push(block);
+ new_blocks.insert(std::make_pair(block->getStartAddress(), block));
+ function->addBasicBlock(block);
+
+ while (remaining_blocks.size()) {
+ BasicBlock * current_block = remaining_blocks.top();
+ remaining_blocks.pop();
+
+ LOG4CXX_DEBUG(logger, "Handling Block starting at " << std::hex
+ << current_block->getStartAddress());
+
+ uint64_t inst_size;
+ uint64_t base_address;
+ text_section.getAddress(base_address);
+ uint64_t current_address = current_block->getStartAddress() - base_address;
+ while(true) {
+ 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())) {
+ uint64_t jmptarget;
+
+ if (MIA->evaluateBranch(inst, current_address, inst_size, jmptarget)) {
+ jmptarget += base_address;
+ if (!MIA->isIndirectBranch(inst)) {
+ if (MIA->isCall(inst)) {
+ if (NULL == manager->getFunction(jmptarget))
+ called_functions.push_back(jmptarget);
+ } else {
+ current_block->setNextBlock(0, jmptarget);
+ if (new_blocks.find(jmptarget) == new_blocks.end()) {
+ BasicBlock * block = manager->newBasicBlock(jmptarget);
+ assert(block);
+ new_blocks.insert(std::make_pair(block->getStartAddress(), block));
+ function->addBasicBlock(block);
+ remaining_blocks.push(block);
+ } else {
+ LOG4CXX_DEBUG(logger, "Reusing Block starting at " << std::hex
+ << current_block->getStartAddress());
+ function->addBasicBlock(new_blocks.find(jmptarget)->second);
+ }
+ if (MIA->isConditionalBranch(inst)) {
+ jmptarget = base_address + current_address + inst_size;
+ current_block->setNextBlock(1, jmptarget);
+ if (new_blocks.find(jmptarget) == new_blocks.end()) {
+ BasicBlock * block = manager->newBasicBlock(jmptarget);
+ assert(block);
+ new_blocks.insert(std::make_pair(block->getStartAddress(), block));
+ function->addBasicBlock(block);
+ remaining_blocks.push(block);
+ } else {
+ LOG4CXX_DEBUG(logger, "Reusing Block starting at " << std::hex
+ << current_block->getStartAddress());
+ function->addBasicBlock(new_blocks.find(jmptarget)->second);
+ }
+ }
+ }
+ }
+ }
+ } else {
+ inst_size = 0;
+ }
+
+
+ if (inst_size == 0 || MIA->isTerminator(inst) || MIA->isBranch(inst)) {
+ 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;
+ }
+ }
+ splitBlocks(function);
+ LOG4CXX_DEBUG(logger, "Finished function " << function->getName());
+ manager->finishFunction(function);
+ for (uint64_t address : called_functions)
+ disassembleFunctionAt(address);
+}
+
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::disassemble() {
+ SectionRef text_section = sections[".text"];
+ std::vector<Function*> remaining_functions;
+
+ // 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)
+ continue;
+
+ if (x->second.getType(symbol_type)
+ || SymbolRef::ST_Function != symbol_type)
+ continue;
+
+ if (!x->second.getAddress(result)) {
+ Function * fun = manager->newFunction(result);
+ fun->setName(x->first);
+ remaining_functions.push_back(fun);
+ LOG4CXX_DEBUG(logger, "Disasembling " << x->first);
+ }
+ }
+
+ for (Function* function : remaining_functions) {
+ disassembleFunction(function);
+ manager->finishFunction(function);
+ }
+
+ if (binary->isELF()) {
+ uint64_t _entryAddress = entryAddress();
+ LOG4CXX_DEBUG(logger, "Adding entryAddress at: " << std::hex << _entryAddress);
+ std::stringstream s;
+ s << "<_start 0x" << std::hex << _entryAddress << ">";
+
+ disassembleFunctionAt(_entryAddress, s.str());
+ }
+
+ if (!manager->hasFunctions()) {
+ uint64_t text_entry;
+ text_section.getAddress(text_entry);
+ LOG4CXX_INFO(logger, "No Symbols found, starting at the beginning of the text segment");
+ disassembleFunctionAt(text_entry);
+ }
+}
+
+template <>
+uint64_t LLVMDisassembler<COFFT>::entryAddress() {
+ const auto coffobject = dyn_cast<COFFObjectFile>(o);
+ const struct pe32_header* pe32_header;
+ const struct pe32plus_header* pe32plus_header;
+
+ coffobject->getPE32PlusHeader(pe32plus_header);
+
+ if (pe32plus_header) {
+ return pe32plus_header->AddressOfEntryPoint;
+ } else {
+ coffobject->getPE32Header(pe32_header);
+ return pe32_header->AddressOfEntryPoint;
+ }
+}
+
+template <typename ELFT>
+uint64_t LLVMDisassembler<ELFT>::entryAddress() {
+ const auto elffile = dyn_cast<ELFObjectFile<ELFT>>(o)->getELFFile();
+ const auto * header = elffile->getHeader();
+
+ return header->e_entry;