/// StripSymbolNames - Strip symbol names.
static bool StripSymbolNames(Module &M, bool PreserveDbgInfo) {
SmallPtrSet<const GlobalValue*, 8> llvmUsedValues;
findUsedValues(M.getGlobalVariable("llvm.used"), llvmUsedValues);
findUsedValues(M.getGlobalVariable("llvm.compiler.used"), llvmUsedValues);
for (Module::global_iterator I = M.global_begin(), E = M.global_end();
I != E; ++I) {
if (I->hasLocalLinkage() && llvmUsedValues.count(I) == 0)
if (!PreserveDbgInfo || !I->getName().startswith("llvm.dbg"))
I->setName(""); // Internal symbols can't participate in linkage
}
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
if (I->hasLocalLinkage() && llvmUsedValues.count(I) == 0)
if (!PreserveDbgInfo || !I->getName().startswith("llvm.dbg"))
I->setName(""); // Internal symbols can't participate in linkage
StripSymtab(I->getValueSymbolTable(), PreserveDbgInfo);
}
// Remove all names from types.
StripTypeSymtab(M.getTypeSymbolTable(), PreserveDbgInfo);
return true;
}
/// AnalyzeGlobals - Scan through the users of all of the internal
/// GlobalValue's in the program. If none of them have their "address taken"
/// (really, their address passed to something nontrivial), record this fact,
/// and record the functions that they are used directly in.
void GlobalsModRef::AnalyzeGlobals(Module &M) {
std::vector<Function*> Readers, Writers;
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
if (I->hasLocalLinkage()) {
if (!AnalyzeUsesOfPointer(I, Readers, Writers)) {
// Remember that we are tracking this global.
NonAddressTakenGlobals.insert(I);
++NumNonAddrTakenFunctions;
}
Readers.clear(); Writers.clear();
}
for (Module::global_iterator I = M.global_begin(), E = M.global_end();
I != E; ++I)
if (I->hasLocalLinkage()) {
if (!AnalyzeUsesOfPointer(I, Readers, Writers)) {
// Remember that we are tracking this global, and the mod/ref fns
NonAddressTakenGlobals.insert(I);
for (unsigned i = 0, e = Readers.size(); i != e; ++i)
FunctionInfo[Readers[i]].GlobalInfo[I] |= Ref;
if (!I->isConstant()) // No need to keep track of writers to constants
for (unsigned i = 0, e = Writers.size(); i != e; ++i)
FunctionInfo[Writers[i]].GlobalInfo[I] |= Mod;
++NumNonAddrTakenGlobalVars;
// If this global holds a pointer type, see if it is an indirect global.
if (I->getType()->getElementType()->isPointerTy() &&
AnalyzeIndirectGlobalMemory(I))
++NumIndirectGlobalVars;
}
Readers.clear(); Writers.clear();
}
}
bool MergeFunctions::runOnModule(Module &M) {
bool Changed = false;
std::map<unsigned long, std::vector<Function *> > FnMap;
for (Module::iterator F = M.begin(), E = M.end(); F != E; ++F) {
if (F->isDeclaration() || F->isIntrinsic())
continue;
if (!F->hasLocalLinkage() && !F->hasExternalLinkage() &&
!F->hasWeakLinkage())
continue;
if (hasAddressTaken(F))
continue;
FnMap[hash(F)].push_back(F);
}
// TODO: instead of running in a loop, we could also fold functions in callgraph
// order. Constructing the CFG probably isn't cheaper than just running in a loop.
bool LocalChanged;
do {
LocalChanged = false;
for (std::map<unsigned long, std::vector<Function *> >::iterator
I = FnMap.begin(), E = FnMap.end(); I != E; ++I) {
DOUT << "size: " << FnMap.size() << "\n";
std::vector<Function *> &FnVec = I->second;
DOUT << "hash (" << I->first << "): " << FnVec.size() << "\n";
for (int i = 0, e = FnVec.size(); i != e; ++i) {
for (int j = i + 1; j != e; ++j) {
bool isEqual = equals(FnVec[i], FnVec[j]);
DOUT << " " << FnVec[i]->getName()
<< (isEqual ? " == " : " != ")
<< FnVec[j]->getName() << "\n";
if (isEqual) {
if (fold(FnVec, i, j)) {
LocalChanged = true;
FnVec.erase(FnVec.begin() + j);
--j, --e;
}
}
}
}
}
Changed |= LocalChanged;
} while (LocalChanged);
return Changed;
}
/// GetAllUndefinedSymbols - calculates the set of undefined symbols that still
/// exist in an LLVM module. This is a bit tricky because there may be two
/// symbols with the same name but different LLVM types that will be resolved to
/// each other but aren't currently (thus we need to treat it as resolved).
///
/// Inputs:
/// M - The module in which to find undefined symbols.
///
/// Outputs:
/// UndefinedSymbols - A set of C++ strings containing the name of all
/// undefined symbols.
///
static void
GetAllUndefinedSymbols(Module *M, std::set<std::string> &UndefinedSymbols) {
std::set<std::string> DefinedSymbols;
UndefinedSymbols.clear();
// If the program doesn't define a main, try pulling one in from a .a file.
// This is needed for programs where the main function is defined in an
// archive, such f2c'd programs.
Function *Main = M->getFunction("main");
if (Main == 0 || Main->isDeclaration())
UndefinedSymbols.insert("main");
for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
assert(!I->hasDLLImportLinkage()
&& "Found dllimported non-external symbol!");
DefinedSymbols.insert(I->getName());
}
}
for (Module::global_iterator I = M->global_begin(), E = M->global_end();
I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
assert(!I->hasDLLImportLinkage()
&& "Found dllimported non-external symbol!");
DefinedSymbols.insert(I->getName());
}
}
for (Module::alias_iterator I = M->alias_begin(), E = M->alias_end();
I != E; ++I)
if (I->hasName())
DefinedSymbols.insert(I->getName());
// Prune out any defined symbols from the undefined symbols set...
for (std::set<std::string>::iterator I = UndefinedSymbols.begin();
I != UndefinedSymbols.end(); )
if (DefinedSymbols.count(*I))
UndefinedSymbols.erase(I++); // This symbol really is defined!
else
++I; // Keep this symbol in the undefined symbols list
}
void
AndroidBitcodeLinker::GetAllSymbols(Module *M,
std::set<std::string> &UndefinedSymbols,
std::set<std::string> &DefinedSymbols) {
UndefinedSymbols.clear();
DefinedSymbols.clear();
Function *Main = M->getFunction("main");
if (Main == 0 || Main->isDeclaration())
UndefinedSymbols.insert("main");
for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
assert(!I->hasDLLImportStorageClass()
&& "Found dllimported non-external symbol!");
DefinedSymbols.insert(I->getName());
}
}
for (Module::global_iterator I = M->global_begin(), E = M->global_end();
I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
assert(!I->hasDLLImportStorageClass()
&& "Found dllimported non-external symbol!");
DefinedSymbols.insert(I->getName());
}
}
for (Module::alias_iterator I = M->alias_begin(), E = M->alias_end();
I != E; ++I)
if (I->hasName())
DefinedSymbols.insert(I->getName());
for (std::set<std::string>::iterator I = UndefinedSymbols.begin();
I != UndefinedSymbols.end(); )
if (DefinedSymbols.count(*I))
UndefinedSymbols.erase(I++);
else
++I;
}
static void getSymbols(Module*M, std::vector<std::string>& symbols) {
// Loop over global variables
for (Module::global_iterator GI = M->global_begin(), GE=M->global_end(); GI != GE; ++GI)
if (!GI->isDeclaration() && !GI->hasLocalLinkage())
if (!GI->getName().empty())
symbols.push_back(GI->getName());
// Loop over functions
for (Module::iterator FI = M->begin(), FE = M->end(); FI != FE; ++FI)
if (!FI->isDeclaration() && !FI->hasLocalLinkage())
if (!FI->getName().empty())
symbols.push_back(FI->getName());
// Loop over aliases
for (Module::alias_iterator AI = M->alias_begin(), AE = M->alias_end();
AI != AE; ++AI) {
if (AI->hasName())
symbols.push_back(AI->getName());
}
}
bool IPCP::runOnModule(Module &M) {
bool Changed = false;
bool LocalChange = true;
// FIXME: instead of using smart algorithms, we just iterate until we stop
// making changes.
while (LocalChange) {
LocalChange = false;
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
if (!I->isDeclaration()) {
// Delete any klingons.
I->removeDeadConstantUsers();
if (I->hasLocalLinkage())
LocalChange |= PropagateConstantsIntoArguments(*I);
Changed |= PropagateConstantReturn(*I);
}
Changed |= LocalChange;
}
return Changed;
}
bool InternalizePass::runOnModule(Module &M) {
CallGraph *CG = getAnalysisIfAvailable<CallGraph>();
CallGraphNode *ExternalNode = CG ? CG->getExternalCallingNode() : 0;
if (ExternalNames.empty()) {
// Return if we're not in 'all but main' mode and have no external api
if (!AllButMain)
return false;
// If no list or file of symbols was specified, check to see if there is a
// "main" symbol defined in the module. If so, use it, otherwise do not
// internalize the module, it must be a library or something.
//
Function *MainFunc = M.getFunction("main");
if (MainFunc == 0 || MainFunc->isDeclaration())
return false; // No main found, must be a library...
// Preserve main, internalize all else.
ExternalNames.insert(MainFunc->getName());
}
bool Changed = false;
// Never internalize functions which code-gen might insert.
ExternalNames.insert("__stack_chk_fail");
// Mark all functions not in the api as internal.
// FIXME: maybe use private linkage?
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
if (!I->isDeclaration() && // Function must be defined here
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!I->hasLocalLinkage() && // Can't already have internal linkage
!ExternalNames.count(I->getName())) {// Not marked to keep external?
I->setLinkage(GlobalValue::InternalLinkage);
// Remove a callgraph edge from the external node to this function.
if (ExternalNode) ExternalNode->removeOneAbstractEdgeTo((*CG)[I]);
Changed = true;
++NumFunctions;
DEBUG(dbgs() << "Internalizing func " << I->getName() << "\n");
}
// Never internalize the llvm.used symbol. It is used to implement
// attribute((used)).
// FIXME: Shouldn't this just filter on llvm.metadata section??
ExternalNames.insert("llvm.used");
ExternalNames.insert("llvm.compiler.used");
// Never internalize anchors used by the machine module info, else the info
// won't find them. (see MachineModuleInfo.)
ExternalNames.insert("llvm.global_ctors");
ExternalNames.insert("llvm.global_dtors");
ExternalNames.insert("llvm.global.annotations");
// Never internalize symbols code-gen inserts.
ExternalNames.insert("__stack_chk_guard");
// Mark all global variables with initializers that are not in the api as
// internal as well.
// FIXME: maybe use private linkage?
for (Module::global_iterator I = M.global_begin(), E = M.global_end();
I != E; ++I)
if (!I->isDeclaration() && !I->hasLocalLinkage() &&
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!ExternalNames.count(I->getName())) {
I->setLinkage(GlobalValue::InternalLinkage);
Changed = true;
++NumGlobals;
DEBUG(dbgs() << "Internalized gvar " << I->getName() << "\n");
}
// Mark all aliases that are not in the api as internal as well.
for (Module::alias_iterator I = M.alias_begin(), E = M.alias_end();
I != E; ++I)
if (!I->isDeclaration() && !I->hasInternalLinkage() &&
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!ExternalNames.count(I->getName())) {
I->setLinkage(GlobalValue::InternalLinkage);
Changed = true;
++NumAliases;
DEBUG(dbgs() << "Internalized alias " << I->getName() << "\n");
}
return Changed;
}
bool InternalizePass::runOnModule(Module &M) {
CallGraph *CG = getAnalysisIfAvailable<CallGraph>();
CallGraphNode *ExternalNode = CG ? CG->getExternalCallingNode() : 0;
bool Changed = false;
// Never internalize functions which code-gen might insert.
// FIXME: We should probably add this (and the __stack_chk_guard) via some
// type of call-back in CodeGen.
ExternalNames.insert("__stack_chk_fail");
// Mark all functions not in the api as internal.
// FIXME: maybe use private linkage?
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
if (!I->isDeclaration() && // Function must be defined here
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!I->hasLocalLinkage() && // Can't already have internal linkage
!ExternalNames.count(I->getName())) {// Not marked to keep external?
I->setLinkage(GlobalValue::InternalLinkage);
// Remove a callgraph edge from the external node to this function.
if (ExternalNode) ExternalNode->removeOneAbstractEdgeTo((*CG)[I]);
Changed = true;
++NumFunctions;
DEBUG(dbgs() << "Internalizing func " << I->getName() << "\n");
}
// Never internalize the llvm.used symbol. It is used to implement
// attribute((used)).
// FIXME: Shouldn't this just filter on llvm.metadata section??
ExternalNames.insert("llvm.used");
ExternalNames.insert("llvm.compiler.used");
// Never internalize anchors used by the machine module info, else the info
// won't find them. (see MachineModuleInfo.)
ExternalNames.insert("llvm.global_ctors");
ExternalNames.insert("llvm.global_dtors");
ExternalNames.insert("llvm.global.annotations");
// Never internalize symbols code-gen inserts.
ExternalNames.insert("__stack_chk_guard");
// Mark all global variables with initializers that are not in the api as
// internal as well.
// FIXME: maybe use private linkage?
for (Module::global_iterator I = M.global_begin(), E = M.global_end();
I != E; ++I)
if (!I->isDeclaration() && !I->hasLocalLinkage() &&
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!ExternalNames.count(I->getName())) {
I->setLinkage(GlobalValue::InternalLinkage);
Changed = true;
++NumGlobals;
DEBUG(dbgs() << "Internalized gvar " << I->getName() << "\n");
}
// Mark all aliases that are not in the api as internal as well.
for (Module::alias_iterator I = M.alias_begin(), E = M.alias_end();
I != E; ++I)
if (!I->isDeclaration() && !I->hasInternalLinkage() &&
// Available externally is really just a "declaration with a body".
!I->hasAvailableExternallyLinkage() &&
!ExternalNames.count(I->getName())) {
I->setLinkage(GlobalValue::InternalLinkage);
Changed = true;
++NumAliases;
DEBUG(dbgs() << "Internalized alias " << I->getName() << "\n");
}
return Changed;
}
/// Based on GetAllUndefinedSymbols() from LLVM3.2
///
/// GetAllUndefinedSymbols - calculates the set of undefined symbols that still
/// exist in an LLVM module. This is a bit tricky because there may be two
/// symbols with the same name but different LLVM types that will be resolved to
/// each other but aren't currently (thus we need to treat it as resolved).
///
/// Inputs:
/// M - The module in which to find undefined symbols.
///
/// Outputs:
/// UndefinedSymbols - A set of C++ strings containing the name of all
/// undefined symbols.
///
static void
GetAllUndefinedSymbols(Module *M, std::set<std::string> &UndefinedSymbols) {
static const std::string llvmIntrinsicPrefix="llvm.";
std::set<std::string> DefinedSymbols;
UndefinedSymbols.clear();
KLEE_DEBUG_WITH_TYPE("klee_linker",
dbgs() << "*** Computing undefined symbols ***\n");
for (Module::iterator I = M->begin(), E = M->end(); I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
#if LLVM_VERSION_CODE < LLVM_VERSION(3, 5)
assert(!I->hasDLLImportLinkage() && "Found dllimported non-external symbol!");
#else
assert(!I->hasDLLImportStorageClass() && "Found dllimported non-external symbol!");
#endif
DefinedSymbols.insert(I->getName());
}
}
for (Module::global_iterator I = M->global_begin(), E = M->global_end();
I != E; ++I)
if (I->hasName()) {
if (I->isDeclaration())
UndefinedSymbols.insert(I->getName());
else if (!I->hasLocalLinkage()) {
#if LLVM_VERSION_CODE < LLVM_VERSION(3, 5)
assert(!I->hasDLLImportLinkage() && "Found dllimported non-external symbol!");
#else
assert(!I->hasDLLImportStorageClass() && "Found dllimported non-external symbol!");
#endif
DefinedSymbols.insert(I->getName());
}
}
for (Module::alias_iterator I = M->alias_begin(), E = M->alias_end();
I != E; ++I)
if (I->hasName())
DefinedSymbols.insert(I->getName());
// Prune out any defined symbols from the undefined symbols set
// and other symbols we don't want to treat as an undefined symbol
std::vector<std::string> SymbolsToRemove;
for (std::set<std::string>::iterator I = UndefinedSymbols.begin();
I != UndefinedSymbols.end(); ++I )
{
if (DefinedSymbols.count(*I))
{
SymbolsToRemove.push_back(*I);
continue;
}
// Strip out llvm intrinsics
if ( (I->size() >= llvmIntrinsicPrefix.size() ) &&
(I->compare(0, llvmIntrinsicPrefix.size(), llvmIntrinsicPrefix) == 0) )
{
KLEE_DEBUG_WITH_TYPE("klee_linker", dbgs() << "LLVM intrinsic " << *I <<
" has will be removed from undefined symbols"<< "\n");
SymbolsToRemove.push_back(*I);
continue;
}
// Symbol really is undefined
KLEE_DEBUG_WITH_TYPE("klee_linker",
dbgs() << "Symbol " << *I << " is undefined.\n");
}
// Remove KLEE intrinsics from set of undefined symbols
for (SpecialFunctionHandler::const_iterator sf = SpecialFunctionHandler::begin(),
se = SpecialFunctionHandler::end(); sf != se; ++sf)
{
if (UndefinedSymbols.find(sf->name) == UndefinedSymbols.end())
continue;
SymbolsToRemove.push_back(sf->name);
KLEE_DEBUG_WITH_TYPE("klee_linker",
dbgs() << "KLEE intrinsic " << sf->name <<
" has will be removed from undefined symbols"<< "\n");
}
// Now remove the symbols from undefined set.
for (size_t i = 0, j = SymbolsToRemove.size(); i < j; ++i )
UndefinedSymbols.erase(SymbolsToRemove[i]);
KLEE_DEBUG_WITH_TYPE("klee_linker",
dbgs() << "*** Finished computing undefined symbols ***\n");
}