static bool LiftFunctionsIntoModule(NativeModulePtr natMod, llvm::Module *M) {
// populate functions
for (auto &func_info : natMod->get_funcs()) {
NativeFunctionPtr f = func_info.second;
if (!InsertFunctionIntoModule(natMod, f, M)) {
std::string fname = f->get_name();
std::cerr << "Could not insert function: " << fname
<< " into the LLVM module" << std::endl;
return false;
}
}
return true;
}
void PrintCFGFunctionList(const NativeModulePtr native_module, const std::string &architecture) noexcept {
std::ios::fmtflags original_stream_flags(std::cout.flags());
int address_digit_count = (architecture == "amd64" ? 16 : 8);
std::cout << "\nCFG Function List:\n";
const auto &function_map = native_module->get_funcs();
for (const auto &function_descriptor : function_map) {
VA virtual_address = function_descriptor.first;
const NativeFunctionPtr function = function_descriptor.second;
std::cout << " " << std::hex << std::setw(address_digit_count) << std::setfill('0') << virtual_address << " ";
std::cout << function->get_name() << std::endl;
}
std::cout.flags(original_stream_flags);
}
NativeFunctionPtr getFunc(ExecutableContainer *c,
LLVMByteDecoder &d,
stack<VA> &funcs,
ExternalFunctionMap &f,
VA e,
raw_ostream &out)
{
NativeFunctionPtr F = NativeFunctionPtr(new NativeFunction(e));
stack<VA> toVisit;
set<VA> visited;
toVisit.push(e);
out << "getFunc: Starting at 0x" << to_string<VA>(e,hex) << "\n";
out << "getFunc: toVisit size is: " << toVisit.size() << "\n";
while(toVisit.size() > 0) {
VA curBlockHeader = toVisit.top();
toVisit.pop();
if(visited.find(curBlockHeader) != visited.end()) {
continue;
} else {
visited.insert(curBlockHeader);
}
// funcs is new functions to visit later
// toVisit is basic blocks of *this* function to visit now
NativeBlockPtr B = decodeBlock(c,
f,
d,
toVisit,
curBlockHeader,
funcs,
out);
F->add_block(B);
}
out << "getFunc: Function recovery complete for ";
out << " func at " << to_string<VA>(e,hex) << "\n";
F->compute_graph();
return F;
}
void doPrintModule(NativeModulePtr m) {
string pathBase = "./";
list<NativeFunctionPtr> mod_funcs = m->get_funcs();
list<NativeFunctionPtr>::iterator it = mod_funcs.begin();
for(; it != mod_funcs.end(); ++it) {
NativeFunctionPtr f = *it;
string n =
pathBase+to_string<uint64_t>(f->get_start(), hex) + ".dot";
ofstream out(n.c_str());
block_label_writer bgl(f);
CFG g = f->get_cfg();
write_graphviz(out, g, bgl);
}
return;
}
static bool InsertFunctionIntoModule(NativeModulePtr mod,
NativeFunctionPtr func, llvm::Module *M) {
auto &C = M->getContext();
auto F = M->getFunction(func->get_name());
if (!F) {
throw TErr(__LINE__, __FILE__, "Could not get func " + func->get_name());
}
if (!F->empty()) {
std::cout << "WARNING: Asking to re-insert function: " << func->get_name()
<< std::endl << "\tReturning current function instead"
<< std::endl;
return true;
}
auto entryBlock = llvm::BasicBlock::Create(F->getContext(), "entry", F);
ArchAllocRegisterVars(entryBlock);
TranslationContext ctx;
ctx.natM = mod;
ctx.natF = func;
ctx.natI = nullptr;
ctx.M = M;
ctx.F = F;
// Create basic blocks for each basic block in the original function.
for (auto block_info : func->get_blocks()) {
ctx.va_to_bb[block_info.first] = llvm::BasicBlock::Create(
C, block_info.second->get_name(), F);
}
// Create a branch from the end of the entry block to the first block
llvm::BranchInst::Create(ctx.va_to_bb[func->get_start()], entryBlock);
try {
// Lift every basic block into the functions.
auto error = false;
for (auto block_info : func->get_blocks()) {
ctx.natB = block_info.second;
error = LiftBlockIntoFunction(ctx) || error;
ctx.natB = nullptr;
}
// For ease of debugging generated code, don't allow lifted functions to
// be inlined. This will make lifted and native call graphs one-to-one.
F->addFnAttr(llvm::Attribute::NoInline);
//we should be done, having inserted every block into the module
return !error;
} catch (std::exception &e) {
std::cerr << "error: " << std::endl << e.what() << std::endl
<< "in function " << std::hex
<< ctx.natF->get_start() << std::endl;
if (ctx.natB) {
std::cerr << "in block " << std::hex << ctx.natB->get_base() << std::endl;
if (ctx.natI) {
std::cerr << "in inst " << std::hex << ctx.natI->get_loc() << std::endl;
}
}
return false;
}
}
