#include "disassembler/llvm/LLVMDisassembler.hxx"
+#include "disassembler/llvm/LLVMBasicBlock.hxx"
+#include "disassembler/llvm/LLVMFunction.hxx"
+
+#include <stack>
+#include <algorithm>
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"))
triple.setArch(Triple::ArchType(o->getArch()));
std::string tripleName(triple.getTriple());
- LOG4CXX_INFO(logger, "Architecture " << tripleName);
+ LOG4CXX_INFO(logger, "Architecture " << tripleName);
- std::string es;
+ std::string es;
target = TargetRegistry::lookupTarget("", triple, es);
if (!target) {
- LOG4CXX_ERROR(logger, es);
+ LOG4CXX_ERROR(logger, es);
return;
}
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));
return;
}
- OwningPtr<MCObjectDisassembler> OD(
+ IP->setPrintImmHex(llvm::HexStyle::C);
+ IP->setPrintImmHex(true);
+
+ std::unique_ptr<MCObjectDisassembler> OD(
new MCObjectDisassembler(*o, *DisAsm, *MIA));
Mod.reset(OD->buildModule(false));
+
+ readSymbols();
+ readSections();
+ disassemble();
+}
+
+LLVMDisassembler::~LLVMDisassembler() {
+ std::for_each(functions.begin(), functions.end(),
+ [](std::pair<uint64_t,LLVMFunction*> it) {
+ delete it.second;
+ });
+ std::for_each(blocks.begin(), blocks.end(),
+ [](std::pair<uint64_t, LLVMBasicBlock*> it) {
+ delete it.second;
+ });
+}
+
+Function* LLVMDisassembler::disassembleFunctionAt(uint64_t address, const std::string& name) {
+ SectionRef text_section = sections[".text"];
+ uint64_t base_address, size;
+ text_section.getAddress(base_address);
+ text_section.getSize(size);
+
+ if (address < base_address ||
+ address >= base_address + size) {
+ return NULL;
+ }
+
+ if (functions.find(address) != functions.end()) {
+ return functions[address];
+ }
+
+ LLVMFunction * function;
+ if (name == "") {
+ std::stringstream s;
+ s << "<Unnamed 0x" << std::hex << address << ">";
+ function = new LLVMFunction(s.str(), address);
+ } else {
+ function = new LLVMFunction(name, address);
+ }
+ functions.insert(std::make_pair(address, function));
+
+ disassembleFunction(function);
+
+ return function;
+}
+
+void LLVMDisassembler::disassembleFunction(LLVMFunction* function) {
+ std::stack<LLVMBasicBlock*> remaining_blocks;
+ SectionRef text_section = sections[".text"];
+ StringRef bytes;
+ text_section.getContents(bytes);
+ StringRefMemoryObject ref(bytes);
+
+ LOG4CXX_DEBUG(logger, "Handling function " << function->getName());
+
+ LLVMBasicBlock * block = new LLVMBasicBlock(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_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 (functions.find(jmptarget) == functions.end()) {
+ disassembleFunctionAt(jmptarget);
+ }
+ } else {
+ 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;
+ 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);
+ }
+ }
+ }
+ }
+ }
+ } 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;
+ }
+ }
+ LOG4CXX_DEBUG(logger, "Finished function " << function->getName());
+}
+
+void LLVMDisassembler::disassemble() {
+ SectionRef text_section = sections[".text"];
+ std::vector<LLVMFunction*> 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)) {
+ LLVMFunction * fun = new LLVMFunction(x->first, result);
+ remaining_functions.push_back(fun);
+ functions.insert(std::make_pair(result, fun));
+ LOG4CXX_DEBUG(logger, "Disasembling " << x->first);
+ }
+ }
+
+ for (LLVMFunction* function : remaining_functions) {
+ disassembleFunction(function);
+ }
+
+ 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];
+ }
+ }
+ LOG4CXX_DEBUG(logger, "Adding entry at: " << std::hex << entry);
+ std::stringstream s;
+ s << "<_start 0x" << std::hex << entry << ">";
+
+ disassembleFunctionAt(entry, s.str());
+ }
+
+ 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");
+ disassembleFunctionAt(text_entry);
+ }
+
+ splitBlocks();
+}
+
+void LLVMDisassembler::splitBlocks() {
+ SectionRef text_section = sections[".text"];
+ StringRef bytes;
+ text_section.getContents(bytes);
+ StringRefMemoryObject ref(bytes);
+
+ // 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() {
+ error_code ec;
+ symbol_iterator si(o->symbol_begin()), se(o->symbol_end());
+ for (; si != se; ++si) {
+ StringRef name;
+ if ((ec = si->getName(name))) {
+ LOG4CXX_ERROR(logger, ec.message());
+ break;
+ }
+ LOG4CXX_DEBUG(logger, "Added symbol " << name.str());
+ symbols.insert(make_pair(name.str(), *si));
+ }
+}
+
+void LLVMDisassembler::readSections() {
+ error_code ec;
+ section_iterator i(o->section_begin()), e(o->section_end());
+ for (; i != e; ++i) {
+ StringRef name;
+ if ((ec = i->getName(name))) {
+ LOG4CXX_ERROR(logger, ec.message());
+ break;
+ }
+ LOG4CXX_DEBUG(logger, "Added section " << name.str());
+ sections.insert(make_pair(name.str(), *i));
+ }
+
+}
+
+void LLVMDisassembler::forEachFunction(std::function<void (uint64_t, Function*)> callback) {
+ std::for_each(functions.begin(), functions.end(),
+ [&](std::pair<uint64_t, LLVMFunction*> x) {
+ callback(x.first, x.second);
+ });
+}
+
+void LLVMDisassembler::printEachInstruction(uint64_t start, uint64_t end,
+ std::function<void (uint8_t*, size_t, const std::string&)> 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;
+ }
}
projects / frida / frida.git / blobdiff