bool
TaintAnalysis::transfer(const Function& func, const DataflowNode& node_, NodeState& state, const std::vector<Lattice*>& dfInfo) {
static size_t ncalls = 0;
if (debug) {
*debug <<"TaintAnalysis::transfer-" <<++ncalls <<"(func=" <<func.get_name() <<",\n"
<<" node={" <<StringUtility::makeOneLine(node_.toString()) <<"},\n"
<<" state={" <<state.str(this, " ") <<",\n"
<<" dfInfo[" <<dfInfo.size() <<"]={...})\n";
}
SgNode *node = node_.getNode();
assert(!dfInfo.empty());
FiniteVarsExprsProductLattice *prodLat = dynamic_cast<FiniteVarsExprsProductLattice*>(dfInfo.front());
bool modified = magic_tainted(node, prodLat); // some values are automatically tainted based on their name
// Process AST nodes that transfer taintedness. Most of these operations have one or more inputs from which a result
// is always calculated the same way. So we just gather up the inputs and do the calculation at the very end of this
// function. The other operations are handled individually within their "if" bodies.
TaintLattice *result = NULL; // result pointer into the taint lattice
std::vector<TaintLattice*> inputs; // input pointers into the taint lattice
if (isSgAssignInitializer(node)) {
// as in "int a = b"
SgAssignInitializer *xop = isSgAssignInitializer(node);
TaintLattice *in1 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(SgExpr2Var(xop->get_operand())));
inputs.push_back(in1);
} else if (isSgAggregateInitializer(node)) {
// as in "int a[1] = {b}"
SgAggregateInitializer *xop = isSgAggregateInitializer(node);
const SgExpressionPtrList &exprs = xop->get_initializers()->get_expressions();
for (size_t i=0; i<exprs.size(); ++i) {
varID in_id = SgExpr2Var(exprs[i]);
TaintLattice *in = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in_id));
inputs.push_back(in);
}
} else if (isSgInitializedName(node)) {
SgInitializedName *xop = isSgInitializedName(node);
if (xop->get_initializer()) {
varID in1_id = SgExpr2Var(xop->get_initializer());
TaintLattice *in1 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in1_id));
inputs.push_back(in1);
}
} else if (isSgValueExp(node)) {
// numeric and character constants
SgValueExp *xop = isSgValueExp(node);
result = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(SgExpr2Var(xop)));
if (result)
modified = result->set_vertex(TaintLattice::VERTEX_UNTAINTED);
} else if (isSgAddressOfOp(node)) {
// as in "&x". The result taintedness has nothing to do with the value in x.
/*void*/
} else if (isSgBinaryOp(node)) {
// as in "a + b"
SgBinaryOp *xop = isSgBinaryOp(node);
varID in1_id = SgExpr2Var(isSgExpression(xop->get_lhs_operand()));
TaintLattice *in1 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in1_id));
inputs.push_back(in1);
varID in2_id = SgExpr2Var(isSgExpression(xop->get_rhs_operand()));
TaintLattice *in2 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in2_id));
inputs.push_back(in2);
if (isSgAssignOp(node)) { // copy the rhs lattice to the lhs lattice (as well as the entire '=' expression result)
assert(in1 && in2);
modified = in1->meetUpdate(in2);
}
} else if (isSgUnaryOp(node)) {
// as in "-a"
SgUnaryOp *xop = isSgUnaryOp(node);
varID in1_id = SgExpr2Var(xop->get_operand());
TaintLattice *in1 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in1_id));
inputs.push_back(in1);
} else if (isSgReturnStmt(node)) {
// as in "return a". The result will always be dead, so we're just doing this to get some debugging output. Most
// of our test inputs are functions, and the test examines the function's returned taintedness.
SgReturnStmt *xop = isSgReturnStmt(node);
varID in1_id = SgExpr2Var(xop->get_expression());
TaintLattice *in1 = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(in1_id));
inputs.push_back(in1);
}
// Update the result lattice (unless dead) with the inputs (unless dead) by using the meedUpdate() method. All this
// means is that the new result will be the maximum of the old result and all inputs, where "maximum" is defined such
// that "tainted" is greater than "untainted" (and both of them are greater than bottom/unknown).
for (size_t i=0; i<inputs.size(); ++i)
if (debug)
*debug <<"TaintAnalysis::transfer: input " <<(i+1) <<" is " <<lattice_info(inputs[i]) <<"\n";
if (!result && varID::isValidVarExp(node)) {
varID result_id(node); // NOTE: constructor doesn't handle all SgExpression nodes, thus the next "if"
result = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(result_id));
}
if (!result && isSgExpression(node)) {
varID result_id = SgExpr2Var(isSgExpression(node));
result = dynamic_cast<TaintLattice*>(prodLat->getVarLattice(result_id));
}
if (result) {
for (size_t i=0; i<inputs.size(); ++i) {
if (inputs[i])
modified = result->meetUpdate(inputs[i]) || modified;
}
}
if (debug)
*debug <<"TaintAnalysis::transfer: result is " <<lattice_info(result) <<(modified?" (modified)":" (not modified)") <<"\n";
return modified;
}
std::string getSgInitializedName(SgInitializedName* initName) {
std::string exprStr;
SgSymbol* initNameSym = initName->search_for_symbol_from_symbol_table();
std::string varInit = initializeVariable(initName);
std::string retString;
if (initName->get_initptr() != NULL) {
SgInitializer* nameInitializer = initName->get_initializer();
VariantT var = nameInitializer->variantT();
switch (var) {
case V_SgAggregateInitializer:
{
SgAggregateInitializer* aggInit = isSgAggregateInitializer(nameInitializer);
if (!isSgArrayType(aggInit->get_type())) {
std::cout << "currently only arrays use aggregate initializers, you are using " << aggInit->class_name() << std::endl;
ROSE_ASSERT(false);
}
SgExprListExp* members = aggInit->get_initializers();
SgExpressionPtrList member_expressions = members->get_expressions();
std::string symName = SymbolToZ3[initNameSym];
ROSE_ASSERT(SymbolToInstances[initNameSym] == 0);
int arrmem = 0;
std::stringstream exprStream;
for (SgExpressionPtrList::iterator i = member_expressions.begin(); i != member_expressions.end(); i++) {
exprStream << "\n(assert (= (select " << symName << "_0 " << arrmem << ") " << getSgExpressionString((isSgAssignInitializer((*i))->get_operand())) << ")";
arrmem = arrmem+1;
}
retString = varInit + "\n" + exprStream.str();
#ifdef ARRAY_TEST
std::cout << "retString: " << retString << std::endl;
#endif
break;
}
case V_SgCompoundInitializer:
{
std::cout << "SgCompoundInitializer not yet supported" << std::endl;
ROSE_ASSERT(false);
break;
}
case V_SgConstructorInitializer:
{
std::cout << "SgConstructorInitializer is not yet supported" << std::endl;
ROSE_ASSERT(false);
break;
}
case V_SgDesignatedInitializer:
{
std::cout << "SgDesignatedInitializer is not yet supported" << std::endl;
ROSE_ASSERT(false);
break;
}
case V_SgAssignInitializer:
{
SgAssignInitializer* assignInit = isSgAssignInitializer(nameInitializer);
std::string symName = SymbolToZ3[initNameSym];
ROSE_ASSERT(SymbolToInstances[initNameSym] == 0);
exprStr = "(assert (= " + symName + "_0 " + getSgExpressionString(assignInit->get_operand()) + "))";
retString = varInit + "\n" + exprStr;
break;
}
default:
{
std::cout << "unknown initializer of type: " << nameInitializer->class_name() << std::endl;
ROSE_ASSERT(false);
break;
}
}
}
else {
retString = varInit;
}
return retString;
}