#include "disassembler/llvm/LLVMDisassembler.hxx"
-#include "disassembler/llvm/LLVMBasicBlock.hxx"
-#include "disassembler/llvm/LLVMFunction.hxx"
+#include "core/InformationManager.hxx"
+#include "core/Function.hxx"
+#include "core/BasicBlock.hxx"
#include <stack>
#include <algorithm>
+#include <cassert>
using namespace llvm;
using namespace llvm::object;
using std::error_code;
+/*
+ *
+ */
+Disassembler * createLLVMDisassembler(const std::string& filename, InformationManager* manager) {
+ if (filename == "")
+ return NULL;
+
+ std::unique_ptr<Binary> o;
+ o.reset(createBinary(filename).get());
+ Binary * op = o.release();
+
+ // ELFType<endian, maxalign, 64bit>
+ if (ELF32LEObjectFile * object = dyn_cast<ELF32LEObjectFile>(op)) {
+ return new LLVMDisassembler<ELFType<support::little, 2, false>>(filename, manager, object);
+ }
+ if (ELF64LEObjectFile * object = dyn_cast<ELF64LEObjectFile>(op)) {
+ return new LLVMDisassembler<ELFType<support::little, 2, true>>(filename, manager, object);
+ }
+ if (ELF32BEObjectFile * object = dyn_cast<ELF32BEObjectFile>(op)) {
+ return new LLVMDisassembler<ELFType<support::big, 2, false>>(filename, manager, object);
+ }
+ if (ELF64BEObjectFile * object = dyn_cast<ELF64BEObjectFile>(op)) {
+ return new LLVMDisassembler<ELFType<support::big, 2, true>>(filename, manager, object);
+ }
+
+ return NULL;
+}
+
/*
* 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,
- InformationManager* manager)
+template <typename ELFT>
+LLVMDisassembler<ELFT>::LLVMDisassembler(const std::string& filename,
+ InformationManager* manager,
+ ELFObjectFile<ELFT>* file)
: Disassembler(filename, manager)
, logger(log4cxx::Logger::getLogger("LLVMDisassembler"))
, triple("unknown-unknown-unknown")
, manager(manager)
{
LOG4CXX_DEBUG(logger, "Handling file" << filename);
- auto result = createBinary(filename);
- error_code ec;
- if ((ec = result.getError())) {
- LOG4CXX_ERROR(logger, "Failed to load Binary" << ec.message());
- binary = NULL;
- return;
- }
+ if (!file) {
+ auto result = createBinary(filename);
+
+ error_code ec;
+ if ((ec = result.getError())) {
+ LOG4CXX_ERROR(logger, "Failed to load Binary" << ec.message());
+ binary = NULL;
+ return;
+ }
- binary.reset(result.get());
+ binary.reset(result.get());
- o = dyn_cast<ObjectFile>(binary.get());
+ o = dyn_cast<ELFObjectFile<ELFT>>(binary.get());
+ } else {
+ o = file;
+ binary.reset(file);
+ }
triple.setArch(Triple::ArchType(o->getArch()));
std::string tripleName(triple.getTriple());
Mod.reset(OD->buildModule(false));
}
-void LLVMDisassembler::start() {
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::start() {
readSymbols();
readSections();
disassemble();
+ readDynamicSymbols();
}
-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;
- });
-}
+template <typename ELFT>
+LLVMDisassembler<ELFT>::~LLVMDisassembler() {}
-Function* LLVMDisassembler::disassembleFunctionAt(uint64_t address, const std::string& name) {
+template <typename ELFT>
+Function* LLVMDisassembler<ELFT>::disassembleFunctionAt(uint64_t address, const std::string& name) {
+ Function * function;
SectionRef text_section = sections[".text"];
uint64_t base_address, size;
text_section.getAddress(base_address);
return NULL;
}
- if (functions.find(address) != functions.end()) {
- return functions[address];
- }
+ if (NULL == (function = manager->getFunction(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);
+ if (name == "") {
+ std::stringstream s;
+ s << "<Unnamed 0x" << std::hex << address << ">";
+ function = manager->newFunction(address);
+ function->setName(s.str());
+ } else {
+ function = manager->newFunction(address);
+ function->setName(name);
+ }
+ disassembleFunction(function);
+ manager->finishFunction(function);
}
- functions.insert(std::make_pair(address, function));
-
- disassembleFunction(function);
return function;
}
-void LLVMDisassembler::disassembleFunction(LLVMFunction* function) {
- std::stack<LLVMBasicBlock*> remaining_blocks;
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::disassembleFunction(Function* function) {
+ 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);
LOG4CXX_DEBUG(logger, "Handling function " << function->getName());
- LLVMBasicBlock * block = new LLVMBasicBlock(function->getStartAddress(), this);
+ BasicBlock * block = manager->newBasicBlock(function->getStartAddress());
remaining_blocks.push(block);
- blocks.insert(std::make_pair(block->getStartAddress(), block));
+ new_blocks.insert(std::make_pair(block->getStartAddress(), block));
function->addBasicBlock(block);
while (remaining_blocks.size()) {
- LLVMBasicBlock * current_block = remaining_blocks.top();
+ BasicBlock * current_block = remaining_blocks.top();
remaining_blocks.pop();
- LOG4CXX_DEBUG(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;
jmptarget += base_address;
if (!MIA->isIndirectBranch(inst)) {
if (MIA->isCall(inst)) {
- if (functions.find(jmptarget) == functions.end()) {
+ if (NULL == manager->getFunction(jmptarget))
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));
+ 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 (blocks.find(jmptarget) == blocks.end()) {
- LLVMBasicBlock * block = new LLVMBasicBlock(jmptarget, this);
- blocks.insert(std::make_pair(block->getStartAddress(), block));
+ 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);
}
}
}
manager->signal_new_function(function);
}
-void LLVMDisassembler::disassemble() {
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::disassemble() {
SectionRef text_section = sections[".text"];
- std::vector<LLVMFunction*> remaining_functions;
+ std::vector<Function*> remaining_functions;
// Assume all function symbols actually start a real function
for (auto x = symbols.begin(); x != symbols.end(); ++x) {
continue;
if (!x->second.getAddress(result)) {
- LLVMFunction * fun = new LLVMFunction(x->first, result);
+ Function * fun = manager->newFunction(result);
+ fun->setName(x->first);
remaining_functions.push_back(fun);
- functions.insert(std::make_pair(result, fun));
LOG4CXX_DEBUG(logger, "Disasembling " << x->first);
}
}
- for (LLVMFunction* function : remaining_functions) {
+ for (Function* function : remaining_functions) {
disassembleFunction(function);
+ manager->finishFunction(function);
}
if (binary->isELF()) {
- bool is64bit = (binary->getData()[4] == 0x02);
+ const ELFO * elffile = o->getELFFile();
+ const typename ELFO::Elf_Ehdr * header = elffile->getHeader();
- 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);
+ _entryAddress = header->e_entry;
+ LOG4CXX_DEBUG(logger, "Adding entryAddress at: " << std::hex << _entryAddress);
std::stringstream s;
- s << "<_start 0x" << std::hex << entry << ">";
+ s << "<_start 0x" << std::hex << _entryAddress << ">";
- disassembleFunctionAt(entry, s.str());
+ disassembleFunctionAt(_entryAddress, 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);
- }
+ 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);
}
-void LLVMDisassembler::splitBlocks(LLVMFunction* function) {
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::splitBlocks(Function* function) {
SectionRef text_section = sections[".text"];
StringRef bytes;
text_section.getContents(bytes);
if(llvm::MCDisassembler::Success ==
DisAsm->getInstruction(inst, inst_size, ref, current_address, nulls(), nulls())) {
// See if some other block starts here
- auto other = blocks.find(current_address + inst_size + base_address);
+ BasicBlock* other = manager->getBasicBlock(current_address
+ + inst_size
+ + base_address);
// Special case, other block starts here but we are at the end anyway
- if (other != blocks.end()) {
+ if (other != NULL) {
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);
- function->addBasicBlock(other->second);
+ << other->getStartAddress());
+ function->addBasicBlock(other);
current_block->setEndAddress(endaddress);
- current_block->setNextBlock(0, other->first);
+ current_block->setNextBlock(0, other->getStartAddress());
current_block->setNextBlock(1, 0);
}
}
}
}
-void LLVMDisassembler::readSymbols() {
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::readDynamicSymbols() {
+ const ELFO * elffile = o->getELFFile();
+ for (typename ELFO::Elf_Sym_Iter
+ it = elffile->begin_dynamic_symbols(),
+ end = elffile->end_dynamic_symbols();
+ it != end;
+ ++it) {
+ if (it->getType() == 2) { // Function
+ bool is_default;
+ // TODO: Error handling
+ std::string symbolname = *(elffile->getSymbolName(it));
+ std::string symbolversion = *(elffile->getSymbolVersion(nullptr, &*it, is_default));
+ manager->signal_new_dyn_symbol(symbolname + (is_default? "@@" : "@") + symbolversion);
+ LOG4CXX_DEBUG(logger, "Adding dynamic Symbol " << symbolname << (is_default? "@@" : "@") << symbolversion);
+ }
+ }
+}
+
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::readSymbols() {
error_code ec;
symbol_iterator si(o->symbol_begin()), se(o->symbol_end());
for (; si != se; ++si) {
}
}
-void LLVMDisassembler::readSections() {
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::readSections() {
error_code ec;
section_iterator i(o->section_begin()), e(o->section_end());
for (; i != e; ++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) {
+// template <typename ELFT>
+// void LLVMDisassembler<ELFT>::forEachFunction(std::function<void (uint64_t, Function*)> callback) {
+// // std::for_each(functions.begin(), functions.end(),
+// // [&](std::pair<uint64_t, Function*> x) {
+// // callback(x.first, x.second);
+// // });
+// }
+
+template <typename ELFT>
+void LLVMDisassembler<ELFT>::printEachInstruction(uint64_t start, uint64_t end,
+ std::function<void (uint8_t*, size_t,
+ const std::string&,
+ const std::string&)> fun) {
SectionRef text_section = sections[".text"];
uint64_t base_address;
text_section.getAddress(base_address);
ref.readBytes(current_address, inst_size, bytes);
uint64_t jmptarget;
+ std::string ref("");
+ IP->printInst(&inst, s, "");
if (MIA->evaluateBranch(inst, current_address, inst_size, jmptarget)) {
std::stringstream stream;
+ if (MIA->isCall(inst))
+ stream << "function:";
+ else
+ stream << "block:";
+
stream << std::hex << (base_address + jmptarget);
- IP->printInst(&inst, s, stream.str());
- } else
- IP->printInst(&inst, s, "");
+ ref = stream.str();
+ }
+
- fun(bytes, inst_size, s.str());
+ fun(bytes, inst_size, s.str(), ref);
} else {
LOG4CXX_WARN(logger, "Invalid byte at" << std::hex << current_address + base_address);
- fun(NULL, 0, "Invalid Byte");
+ fun(NULL, 0, "Invalid Byte", "");
inst_size = 1;
}
projects / frida / frida.git / blobdiff