void LTOCodeGenerator::
applyRestriction(GlobalValue &GV,
ArrayRef<StringRef> Libcalls,
std::vector<const char*> &MustPreserveList,
SmallPtrSetImpl<GlobalValue*> &AsmUsed,
Mangler &Mangler) {
// There are no restrictions to apply to declarations.
if (GV.isDeclaration())
return;
// There is nothing more restrictive than private linkage.
if (GV.hasPrivateLinkage())
return;
SmallString<64> Buffer;
TargetMach->getNameWithPrefix(Buffer, &GV, Mangler);
if (MustPreserveSymbols.count(Buffer))
MustPreserveList.push_back(GV.getName().data());
if (AsmUndefinedRefs.count(Buffer))
AsmUsed.insert(&GV);
// Conservatively append user-supplied runtime library functions to
// llvm.compiler.used. These could be internalized and deleted by
// optimizations like -globalopt, causing problems when later optimizations
// add new library calls (e.g., llvm.memset => memset and printf => puts).
// Leave it to the linker to remove any dead code (e.g. with -dead_strip).
if (isa<Function>(GV) &&
std::binary_search(Libcalls.begin(), Libcalls.end(), GV.getName()))
AsmUsed.insert(&GV);
}
static void DumpSymbolNameForGlobalValue(GlobalValue &GV) {
// Private linkage and available_externally linkage don't exist in symtab.
if (GV.hasPrivateLinkage() || GV.hasLinkerPrivateLinkage() ||
GV.hasLinkerPrivateWeakLinkage() || GV.hasAvailableExternallyLinkage())
return;
const std::string SymbolAddrStr = " "; // Not used yet...
char TypeChar = TypeCharForSymbol(GV);
if ((TypeChar != 'U') && UndefinedOnly)
return;
if ((TypeChar == 'U') && DefinedOnly)
return;
if (GV.hasLocalLinkage () && ExternalOnly)
return;
if (OutputFormat == posix) {
outs() << GV.getName () << " " << TypeCharForSymbol(GV) << " "
<< SymbolAddrStr << "\n";
} else if (OutputFormat == bsd) {
outs() << SymbolAddrStr << " " << TypeCharForSymbol(GV) << " "
<< GV.getName () << "\n";
} else if (OutputFormat == sysv) {
std::string PaddedName (GV.getName ());
while (PaddedName.length () < 20)
PaddedName += " ";
outs() << PaddedName << "|" << SymbolAddrStr << "| "
<< TypeCharForSymbol(GV)
<< " | | | |\n";
}
}
static bool shouldInternalize(const GlobalValue &GV,
const std::set<std::string> &ExternalNames,
const std::set<std::string> &DSONames) {
// Function must be defined here
if (GV.isDeclaration())
return false;
// Available externally is really just a "declaration with a body".
if (GV.hasAvailableExternallyLinkage())
return false;
// Already has internal linkage
if (GV.hasLocalLinkage())
return false;
// Marked to keep external?
if (ExternalNames.count(GV.getName()))
return false;
// Not needed for the symbol table?
if (!DSONames.count(GV.getName()))
return true;
// Not a linkonce. Someone can depend on it being on the symbol table.
if (!GV.hasLinkOnceLinkage())
return false;
// The address is not important, we can hide it.
if (GV.hasUnnamedAddr())
return true;
// FIXME: Check if the address is used.
return false;
}
void ModuleLinker::addLazyFor(GlobalValue &GV, IRMover::ValueAdder Add) {
if (!shouldLinkReferencedLinkOnce())
// For ThinLTO we don't import more than what was required.
// The client has to guarantee that the linkonce will be availabe at link
// time (by promoting it to weak for instance).
return;
// Add these to the internalize list
if (!GV.hasLinkOnceLinkage() && !shouldLinkOnlyNeeded())
return;
if (shouldInternalizeLinkedSymbols())
Internalize.insert(GV.getName());
Add(GV);
const Comdat *SC = GV.getComdat();
if (!SC)
return;
for (GlobalValue *GV2 : LazyComdatMembers[SC]) {
GlobalValue *DGV = getLinkedToGlobal(GV2);
bool LinkFromSrc = true;
if (DGV && shouldLinkFromSource(LinkFromSrc, *DGV, *GV2))
return;
if (!LinkFromSrc)
continue;
if (shouldInternalizeLinkedSymbols())
Internalize.insert(GV2->getName());
Add(*GV2);
}
}
static bool shouldInternalize(const GlobalValue &GV,
const std::set<std::string> &ExternalNames,
bool OnlyHidden) {
if (OnlyHidden && !GV.hasHiddenVisibility())
return false;
// Function must be defined here
if (GV.isDeclaration())
return false;
// Available externally is really just a "declaration with a body".
if (GV.hasAvailableExternallyLinkage())
return false;
// Assume that dllexported symbols are referenced elsewhere
if (GV.hasDLLExportStorageClass())
return false;
// Already has internal linkage
if (GV.hasLocalLinkage())
return false;
// Marked to keep external?
if (ExternalNames.count(GV.getName()))
return false;
return true;
}
void LTOCodeGenerator::
applyRestriction(GlobalValue &GV,
std::vector<const char*> &MustPreserveList,
std::vector<const char*> &DSOList,
SmallPtrSet<GlobalValue*, 8> &AsmUsed,
Mangler &Mangler) {
SmallString<64> Buffer;
Mangler.getNameWithPrefix(Buffer, &GV, false);
if (GV.isDeclaration())
return;
if (MustPreserveSymbols.count(Buffer))
MustPreserveList.push_back(GV.getName().data());
if (DSOSymbols.count(Buffer))
DSOList.push_back(GV.getName().data());
if (AsmUndefinedRefs.count(Buffer))
AsmUsed.insert(&GV);
}
Module * llvmutil_extractmodule(Module * OrigMod, TargetMachine * TM, std::vector<llvm::GlobalValue*> * livevalues, std::vector<std::string> * symbolnames, bool internalizeandoptimize) {
assert(symbolnames == NULL || livevalues->size() == symbolnames->size());
ValueToValueMapTy VMap;
#if LLVM_VERSION >= 34
Module * M = llvmutil_extractmodulewithproperties(OrigMod->getModuleIdentifier(), OrigMod, (llvm::GlobalValue **)&(*livevalues)[0], livevalues->size(), AlwaysCopy, NULL, VMap);
#else
Module * M = CloneModule(OrigMod, VMap);
internalizeandoptimize = true; //we need to do this regardless of the input because it is the only way we can extract just the needed functions from the module
#endif
//rename values to symbolsnames
std::vector<const char *> names;
for(size_t i = 0; i < livevalues->size(); i++) {
GlobalValue * fn = cast<GlobalValue>(VMap[(*livevalues)[i]]);
const std::string & name = (*symbolnames)[i];
GlobalValue * gv = M->getNamedValue(name);
if(gv) { //if there is already a symbol with this name, rename it
gv->setName(Twine((*symbolnames)[i],"_renamed"));
}
fn->setName(name); //and set our function to this name
assert(fn->getName() == name);
names.push_back(name.c_str()); //internalize pass has weird interface, so we need to copy the names here
}
if (!internalizeandoptimize)
return M;
//at this point we run optimizations on the module
//first internalize all functions not mentioned in "names" using an internalize pass and then perform
//standard optimizations
PassManager MPM;
llvmutil_addtargetspecificpasses(&MPM, TM);
MPM.add(createVerifierPass()); //make sure we haven't messed stuff up yet
MPM.add(createInternalizePass(names));
MPM.add(createGlobalDCEPass()); //run this early since anything not in the table of exported functions is still in this module
//this will remove dead functions
PassManagerBuilder PMB;
PMB.OptLevel = 3;
PMB.SizeLevel = 0;
#if LLVM_VERSION >= 35
PMB.LoopVectorize = true;
PMB.SLPVectorize = true;
#endif
PMB.populateModulePassManager(MPM);
MPM.run(*M);
return M;
}
void Variables::changeGlobal(Change* change, Module &module) {
GlobalValue* oldTarget = dyn_cast<GlobalValue>(change->getValue());
Type* oldType = oldTarget->getType()->getElementType();
Type* newType = change->getType()[0];
errs() << "Changing the precision of variable \"" << oldTarget->getName() << "\" from " << *oldType << " to " << *newType << ".\n";
if (diffTypes(oldType, newType)) {
Constant *initializer;
GlobalVariable* newTarget;
if (PointerType *newPointerType = dyn_cast<PointerType>(newType)) {
initializer = ConstantPointerNull::get(newPointerType);
newTarget = new GlobalVariable(module, newType, false, GlobalValue::CommonLinkage, initializer, "");
}
else if (ArrayType * atype = dyn_cast<ArrayType>(newType)) {
// preparing initializer
Type *temp = Type::getFloatTy(module.getContext());
vector<Constant*> operands;
operands.push_back(ConstantFP::get(temp, 0));
ArrayRef<Constant*> *arrayRef = new ArrayRef<Constant*>(operands);
initializer = ConstantArray::get(atype, *arrayRef);
newTarget = new GlobalVariable(module, newType, false, GlobalValue::CommonLinkage, initializer, "");
}
else {
initializer = ConstantFP::get(newType, 0);
newTarget = new GlobalVariable(module, newType, false, GlobalValue::CommonLinkage, initializer, "");
}
/*
GlobalVariable* newTarget = new GlobalVariable(module, newType, false, GlobalValue::CommonLinkage, initializer, "");
*/
unsigned alignment = getAlignment(newType);
newTarget->setAlignment(alignment);
newTarget->takeName(oldTarget);
// iterating through instructions using old AllocaInst
Value::use_iterator it = oldTarget->use_begin();
for(; it != oldTarget->use_end(); it++) {
Transformer::transform(it, newTarget, oldTarget, newType, oldType, alignment);
}
//oldTarget->eraseFromParent();
}
else {
errs() << "No changes required.\n";
}
return;
}
static void DumpSymbolNameForGlobalValue(GlobalValue &GV) {
const std::string SymbolAddrStr = " "; // Not used yet...
char TypeChar = TypeCharForSymbol (GV);
if ((TypeChar != 'U') && UndefinedOnly)
return;
if ((TypeChar == 'U') && DefinedOnly)
return;
if (GV.hasLocalLinkage () && ExternalOnly)
return;
if (OutputFormat == posix) {
std::cout << GV.getName () << " " << TypeCharForSymbol (GV) << " "
<< SymbolAddrStr << "\n";
} else if (OutputFormat == bsd) {
std::cout << SymbolAddrStr << " " << TypeCharForSymbol (GV) << " "
<< GV.getName () << "\n";
} else if (OutputFormat == sysv) {
std::string PaddedName (GV.getName ());
while (PaddedName.length () < 20)
PaddedName += " ";
std::cout << PaddedName << "|" << SymbolAddrStr << "| "
<< TypeCharForSymbol (GV)
<< " | | | |\n";
}
}
static void DumpSymbolNameForGlobalValue(GlobalValue &GV) {
// Private linkage and available_externally linkage don't exist in symtab.
if (GV.hasPrivateLinkage() ||
GV.hasLinkerPrivateLinkage() ||
GV.hasLinkerPrivateWeakLinkage() ||
GV.hasAvailableExternallyLinkage())
return;
char TypeChar = TypeCharForSymbol(GV);
if (GV.hasLocalLinkage () && ExternalOnly)
return;
NMSymbol s;
s.Address = object::UnknownAddressOrSize;
s.Size = object::UnknownAddressOrSize;
s.TypeChar = TypeChar;
s.Name = GV.getName();
SymbolList.push_back(s);
}
void ModuleLinker::addLazyFor(GlobalValue &GV, IRMover::ValueAdder Add) {
// Add these to the internalize list
if (!GV.hasLinkOnceLinkage())
return;
if (shouldInternalizeLinkedSymbols())
Internalize.insert(GV.getName());
Add(GV);
const Comdat *SC = GV.getComdat();
if (!SC)
return;
for (GlobalValue *GV2 : ComdatMembers[SC]) {
if (!GV2->hasLocalLinkage() && shouldInternalizeLinkedSymbols())
Internalize.insert(GV2->getName());
Add(*GV2);
}
}
bool UnsafeTypeCastingCheck::doInitialization(llvm::Module &mod) {
errs() << "initialization... \n";
for (Module::global_iterator git = mod.global_begin() ;
git != mod.global_end() ;
git++) {
GlobalValue *gv = dyn_cast<GlobalValue>(git);
string gv_name = gv->getName().str();
if (gv_name.compare("blockDim") == 0 ||
gv_name.compare("gridDim") == 0 ||
gv_name.compare("blockIdx") == 0 ||
gv_name.compare("threadIdx") == 0) {
setPointedType(gv, UINT_UT);
}
}
// get utcc_assert function call (assertion)
code_modified = false;
assert_func = mod.getFunction("utcc_assert");
assert(assert_func != NULL);
return false;
}
void ModuleLinker::addLazyFor(GlobalValue &GV, IRMover::ValueAdder Add) {
// Add these to the internalize list
if (!GV.hasLinkOnceLinkage())
return;
if (shouldInternalizeLinkedSymbols())
Internalize.insert(GV.getName());
Add(GV);
const Comdat *SC = GV.getComdat();
if (!SC)
return;
for (GlobalValue *GV2 : LazyComdatMembers[SC]) {
GlobalValue *DGV = getLinkedToGlobal(GV2);
bool LinkFromSrc = true;
if (DGV && shouldLinkFromSource(LinkFromSrc, *DGV, *GV2))
return;
if (!LinkFromSrc)
continue;
if (shouldInternalizeLinkedSymbols())
Internalize.insert(GV2->getName());
Add(*GV2);
}
}
void BitcodeCompiler::add(BitcodeFile &F) {
std::unique_ptr<IRObjectFile> Obj = std::move(F.Obj);
std::vector<GlobalValue *> Keep;
unsigned BodyIndex = 0;
ArrayRef<Symbol *> Syms = F.getSymbols();
Module &M = Obj->getModule();
if (M.getDataLayoutStr().empty())
fatal("invalid bitcode file: " + F.getName() + " has no datalayout");
// Discard non-compatible debug infos if necessary.
M.materializeMetadata();
UpgradeDebugInfo(M);
// If a symbol appears in @llvm.used, the linker is required
// to treat the symbol as there is a reference to the symbol
// that it cannot see. Therefore, we can't internalize.
SmallPtrSet<GlobalValue *, 8> Used;
collectUsedGlobalVariables(M, Used, /* CompilerUsed */ false);
for (const BasicSymbolRef &Sym : Obj->symbols()) {
uint32_t Flags = Sym.getFlags();
GlobalValue *GV = Obj->getSymbolGV(Sym.getRawDataRefImpl());
if (GV && GV->hasAppendingLinkage())
Keep.push_back(GV);
if (BitcodeFile::shouldSkip(Flags))
continue;
Symbol *S = Syms[BodyIndex++];
if (Flags & BasicSymbolRef::SF_Undefined) {
handleUndefinedAsmRefs(Sym, GV, AsmUndefinedRefs);
continue;
}
auto *B = dyn_cast<DefinedBitcode>(S->body());
if (!B || B->File != &F)
continue;
// We collect the set of symbols we want to internalize here
// and change the linkage after the IRMover executed, i.e. after
// we imported the symbols and satisfied undefined references
// to it. We can't just change linkage here because otherwise
// the IRMover will just rename the symbol.
if (GV && shouldInternalize(Used, S, GV))
InternalizedSyms.insert(GV->getName());
// At this point we know that either the combined LTO object will provide a
// definition of a symbol, or we will internalize it. In either case, we
// need to undefine the symbol. In the former case, the real definition
// needs to be able to replace the original definition without conflicting.
// In the latter case, we need to allow the combined LTO object to provide a
// definition with the same name, for example when doing parallel codegen.
undefine(S);
if (!GV)
// Module asm symbol.
continue;
switch (GV->getLinkage()) {
default:
break;
case llvm::GlobalValue::LinkOnceAnyLinkage:
GV->setLinkage(GlobalValue::WeakAnyLinkage);
break;
case llvm::GlobalValue::LinkOnceODRLinkage:
GV->setLinkage(GlobalValue::WeakODRLinkage);
break;
}
Keep.push_back(GV);
}
if (Error E = Mover.move(Obj->takeModule(), Keep,
[](GlobalValue &, IRMover::ValueAdder) {})) {
handleAllErrors(std::move(E), [&](const llvm::ErrorInfoBase &EIB) {
fatal("failed to link module " + F.getName() + ": " + EIB.message());
});
}
}
bool ModuleLinker::shouldLinkFromSource(bool &LinkFromSrc,
const GlobalValue &Dest,
const GlobalValue &Src) {
// Should we unconditionally use the Src?
if (shouldOverrideFromSrc()) {
LinkFromSrc = true;
return false;
}
// We always have to add Src if it has appending linkage.
if (Src.hasAppendingLinkage()) {
// Should have prevented importing for appending linkage in linkIfNeeded.
assert(!isPerformingImport());
LinkFromSrc = true;
return false;
}
bool SrcIsDeclaration = Src.isDeclarationForLinker();
bool DestIsDeclaration = Dest.isDeclarationForLinker();
if (isPerformingImport()) {
if (isa<Function>(&Src)) {
// For functions, LinkFromSrc iff this is a function requested
// for importing. For variables, decide below normally.
LinkFromSrc = GlobalsToImport->count(&Src);
return false;
}
// Check if this is an alias with an already existing definition
// in Dest, which must have come from a prior importing pass from
// the same Src module. Unlike imported function and variable
// definitions, which are imported as available_externally and are
// not definitions for the linker, that is not a valid linkage for
// imported aliases which must be definitions. Simply use the existing
// Dest copy.
if (isa<GlobalAlias>(&Src) && !DestIsDeclaration) {
assert(isa<GlobalAlias>(&Dest));
LinkFromSrc = false;
return false;
}
}
if (SrcIsDeclaration) {
// If Src is external or if both Src & Dest are external.. Just link the
// external globals, we aren't adding anything.
if (Src.hasDLLImportStorageClass()) {
// If one of GVs is marked as DLLImport, result should be dllimport'ed.
LinkFromSrc = DestIsDeclaration;
return false;
}
// If the Dest is weak, use the source linkage.
if (Dest.hasExternalWeakLinkage()) {
LinkFromSrc = true;
return false;
}
// Link an available_externally over a declaration.
LinkFromSrc = !Src.isDeclaration() && Dest.isDeclaration();
return false;
}
if (DestIsDeclaration) {
// If Dest is external but Src is not:
LinkFromSrc = true;
return false;
}
if (Src.hasCommonLinkage()) {
if (Dest.hasLinkOnceLinkage() || Dest.hasWeakLinkage()) {
LinkFromSrc = true;
return false;
}
if (!Dest.hasCommonLinkage()) {
LinkFromSrc = false;
return false;
}
const DataLayout &DL = Dest.getParent()->getDataLayout();
uint64_t DestSize = DL.getTypeAllocSize(Dest.getValueType());
uint64_t SrcSize = DL.getTypeAllocSize(Src.getValueType());
LinkFromSrc = SrcSize > DestSize;
return false;
}
if (Src.isWeakForLinker()) {
assert(!Dest.hasExternalWeakLinkage());
assert(!Dest.hasAvailableExternallyLinkage());
if (Dest.hasLinkOnceLinkage() && Src.hasWeakLinkage()) {
LinkFromSrc = true;
return false;
}
LinkFromSrc = false;
return false;
}
if (Dest.isWeakForLinker()) {
assert(Src.hasExternalLinkage());
LinkFromSrc = true;
return false;
}
assert(!Src.hasExternalWeakLinkage());
assert(!Dest.hasExternalWeakLinkage());
assert(Dest.hasExternalLinkage() && Src.hasExternalLinkage() &&
"Unexpected linkage type!");
return emitError("Linking globals named '" + Src.getName() +
"': symbol multiply defined!");
}
static std::unique_ptr<Module>
getModuleForFile(LLVMContext &Context, claimed_file &F,
ld_plugin_input_file &Info, raw_fd_ostream *ApiFile,
StringSet<> &Internalize, StringSet<> &Maybe) {
if (get_symbols(F.handle, F.syms.size(), F.syms.data()) != LDPS_OK)
message(LDPL_FATAL, "Failed to get symbol information");
const void *View;
if (get_view(F.handle, &View) != LDPS_OK)
message(LDPL_FATAL, "Failed to get a view of file");
MemoryBufferRef BufferRef(StringRef((const char *)View, Info.filesize),
Info.name);
ErrorOr<std::unique_ptr<object::IRObjectFile>> ObjOrErr =
object::IRObjectFile::create(BufferRef, Context);
if (std::error_code EC = ObjOrErr.getError())
message(LDPL_FATAL, "Could not read bitcode from file : %s",
EC.message().c_str());
object::IRObjectFile &Obj = **ObjOrErr;
Module &M = Obj.getModule();
M.materializeMetadata();
UpgradeDebugInfo(M);
SmallPtrSet<GlobalValue *, 8> Used;
collectUsedGlobalVariables(M, Used, /*CompilerUsed*/ false);
DenseSet<GlobalValue *> Drop;
std::vector<GlobalAlias *> KeptAliases;
unsigned SymNum = 0;
for (auto &ObjSym : Obj.symbols()) {
if (shouldSkip(ObjSym.getFlags()))
continue;
ld_plugin_symbol &Sym = F.syms[SymNum];
++SymNum;
ld_plugin_symbol_resolution Resolution =
(ld_plugin_symbol_resolution)Sym.resolution;
if (options::generate_api_file)
*ApiFile << Sym.name << ' ' << getResolutionName(Resolution) << '\n';
GlobalValue *GV = Obj.getSymbolGV(ObjSym.getRawDataRefImpl());
if (!GV) {
freeSymName(Sym);
continue; // Asm symbol.
}
if (Resolution != LDPR_PREVAILING_DEF_IRONLY && GV->hasCommonLinkage()) {
// Common linkage is special. There is no single symbol that wins the
// resolution. Instead we have to collect the maximum alignment and size.
// The IR linker does that for us if we just pass it every common GV.
// We still have to keep track of LDPR_PREVAILING_DEF_IRONLY so we
// internalize once the IR linker has done its job.
freeSymName(Sym);
continue;
}
switch (Resolution) {
case LDPR_UNKNOWN:
llvm_unreachable("Unexpected resolution");
case LDPR_RESOLVED_IR:
case LDPR_RESOLVED_EXEC:
case LDPR_RESOLVED_DYN:
assert(GV->isDeclarationForLinker());
break;
case LDPR_UNDEF:
if (!GV->isDeclarationForLinker()) {
assert(GV->hasComdat());
Drop.insert(GV);
}
break;
case LDPR_PREVAILING_DEF_IRONLY: {
keepGlobalValue(*GV, KeptAliases);
if (!Used.count(GV)) {
// Since we use the regular lib/Linker, we cannot just internalize GV
// now or it will not be copied to the merged module. Instead we force
// it to be copied and then internalize it.
Internalize.insert(GV->getName());
}
break;
}
case LDPR_PREVAILING_DEF:
keepGlobalValue(*GV, KeptAliases);
break;
case LDPR_PREEMPTED_IR:
// Gold might have selected a linkonce_odr and preempted a weak_odr.
// In that case we have to make sure we don't end up internalizing it.
if (!GV->isDiscardableIfUnused())
Maybe.erase(GV->getName());
// fall-through
case LDPR_PREEMPTED_REG:
Drop.insert(GV);
break;
case LDPR_PREVAILING_DEF_IRONLY_EXP: {
// We can only check for address uses after we merge the modules. The
// reason is that this GV might have a copy in another module
// and in that module the address might be significant, but that
// copy will be LDPR_PREEMPTED_IR.
if (GV->hasLinkOnceODRLinkage())
Maybe.insert(GV->getName());
keepGlobalValue(*GV, KeptAliases);
break;
}
}
freeSymName(Sym);
}
ValueToValueMapTy VM;
LocalValueMaterializer Materializer(Drop);
for (GlobalAlias *GA : KeptAliases) {
// Gold told us to keep GA. It is possible that a GV usied in the aliasee
// expression is being dropped. If that is the case, that GV must be copied.
Constant *Aliasee = GA->getAliasee();
Constant *Replacement = mapConstantToLocalCopy(Aliasee, VM, &Materializer);
GA->setAliasee(Replacement);
}
for (auto *GV : Drop)
drop(*GV);
return Obj.takeModule();
}
void FunctionImportGlobalProcessing::processGlobalForThinLTO(GlobalValue &GV) {
ValueInfo VI;
if (GV.hasName()) {
VI = ImportIndex.getValueInfo(GV.getGUID());
// Set synthetic function entry counts.
if (VI && ImportIndex.hasSyntheticEntryCounts()) {
if (Function *F = dyn_cast<Function>(&GV)) {
if (!F->isDeclaration()) {
for (auto &S : VI.getSummaryList()) {
FunctionSummary *FS = dyn_cast<FunctionSummary>(S->getBaseObject());
if (FS->modulePath() == M.getModuleIdentifier()) {
F->setEntryCount(Function::ProfileCount(FS->entryCount(),
Function::PCT_Synthetic));
break;
}
}
}
}
}
// Check the summaries to see if the symbol gets resolved to a known local
// definition.
if (VI && VI.isDSOLocal()) {
GV.setDSOLocal(true);
if (GV.hasDLLImportStorageClass())
GV.setDLLStorageClass(GlobalValue::DefaultStorageClass);
}
}
// Mark read-only variables which can be imported with specific attribute.
// We can't internalize them now because IRMover will fail to link variable
// definitions to their external declarations during ThinLTO import. We'll
// internalize read-only variables later, after import is finished.
// See internalizeImmutableGVs.
//
// If global value dead stripping is not enabled in summary then
// propagateConstants hasn't been run. We can't internalize GV
// in such case.
if (!GV.isDeclaration() && VI && ImportIndex.withGlobalValueDeadStripping()) {
const auto &SL = VI.getSummaryList();
auto *GVS = SL.empty() ? nullptr : dyn_cast<GlobalVarSummary>(SL[0].get());
if (GVS && GVS->isReadOnly())
cast<GlobalVariable>(&GV)->addAttribute("thinlto-internalize");
}
bool DoPromote = false;
if (GV.hasLocalLinkage() &&
((DoPromote = shouldPromoteLocalToGlobal(&GV)) || isPerformingImport())) {
// Save the original name string before we rename GV below.
auto Name = GV.getName().str();
// Once we change the name or linkage it is difficult to determine
// again whether we should promote since shouldPromoteLocalToGlobal needs
// to locate the summary (based on GUID from name and linkage). Therefore,
// use DoPromote result saved above.
GV.setName(getName(&GV, DoPromote));
GV.setLinkage(getLinkage(&GV, DoPromote));
if (!GV.hasLocalLinkage())
GV.setVisibility(GlobalValue::HiddenVisibility);
// If we are renaming a COMDAT leader, ensure that we record the COMDAT
// for later renaming as well. This is required for COFF.
if (const auto *C = GV.getComdat())
if (C->getName() == Name)
RenamedComdats.try_emplace(C, M.getOrInsertComdat(GV.getName()));
} else
GV.setLinkage(getLinkage(&GV, /* DoPromote */ false));
// Remove functions imported as available externally defs from comdats,
// as this is a declaration for the linker, and will be dropped eventually.
// It is illegal for comdats to contain declarations.
auto *GO = dyn_cast<GlobalObject>(&GV);
if (GO && GO->isDeclarationForLinker() && GO->hasComdat()) {
// The IRMover should not have placed any imported declarations in
// a comdat, so the only declaration that should be in a comdat
// at this point would be a definition imported as available_externally.
assert(GO->hasAvailableExternallyLinkage() &&
"Expected comdat on definition (possibly available external)");
GO->setComdat(nullptr);
}
}
bool ModuleLinker::shouldLinkFromSource(bool &LinkFromSrc,
const GlobalValue &Dest,
const GlobalValue &Src) {
// Should we unconditionally use the Src?
if (shouldOverrideFromSrc()) {
LinkFromSrc = true;
return false;
}
// We always have to add Src if it has appending linkage.
if (Src.hasAppendingLinkage()) {
// Should have prevented importing for appending linkage in linkIfNeeded.
assert(!isPerformingImport());
LinkFromSrc = true;
return false;
}
if (isPerformingImport()) {
// LinkFromSrc iff this is a global requested for importing.
LinkFromSrc = GlobalsToImport->count(&Src);
return false;
}
bool SrcIsDeclaration = Src.isDeclarationForLinker();
bool DestIsDeclaration = Dest.isDeclarationForLinker();
if (SrcIsDeclaration) {
// If Src is external or if both Src & Dest are external.. Just link the
// external globals, we aren't adding anything.
if (Src.hasDLLImportStorageClass()) {
// If one of GVs is marked as DLLImport, result should be dllimport'ed.
LinkFromSrc = DestIsDeclaration;
return false;
}
// If the Dest is weak, use the source linkage.
if (Dest.hasExternalWeakLinkage()) {
LinkFromSrc = true;
return false;
}
// Link an available_externally over a declaration.
LinkFromSrc = !Src.isDeclaration() && Dest.isDeclaration();
return false;
}
if (DestIsDeclaration) {
// If Dest is external but Src is not:
LinkFromSrc = true;
return false;
}
if (Src.hasCommonLinkage()) {
if (Dest.hasLinkOnceLinkage() || Dest.hasWeakLinkage()) {
LinkFromSrc = true;
return false;
}
if (!Dest.hasCommonLinkage()) {
LinkFromSrc = false;
return false;
}
const DataLayout &DL = Dest.getParent()->getDataLayout();
uint64_t DestSize = DL.getTypeAllocSize(Dest.getValueType());
uint64_t SrcSize = DL.getTypeAllocSize(Src.getValueType());
LinkFromSrc = SrcSize > DestSize;
return false;
}
if (Src.isWeakForLinker()) {
assert(!Dest.hasExternalWeakLinkage());
assert(!Dest.hasAvailableExternallyLinkage());
if (Dest.hasLinkOnceLinkage() && Src.hasWeakLinkage()) {
LinkFromSrc = true;
return false;
}
LinkFromSrc = false;
return false;
}
if (Dest.isWeakForLinker()) {
assert(Src.hasExternalLinkage());
LinkFromSrc = true;
return false;
}
assert(!Src.hasExternalWeakLinkage());
assert(!Dest.hasExternalWeakLinkage());
assert(Dest.hasExternalLinkage() && Src.hasExternalLinkage() &&
"Unexpected linkage type!");
return emitError("Linking globals named '" + Src.getName() +
"': symbol multiply defined!");
}
void BitcodeCompiler::add(BitcodeFile &F) {
std::unique_ptr<IRObjectFile> Obj =
check(IRObjectFile::create(F.MB, Context));
std::vector<GlobalValue *> Keep;
unsigned BodyIndex = 0;
ArrayRef<SymbolBody *> Bodies = F.getSymbols();
Module &M = Obj->getModule();
if (M.getDataLayoutStr().empty())
fatal("invalid bitcode file: " + F.getName() + " has no datalayout");
// If a symbol appears in @llvm.used, the linker is required
// to treat the symbol as there is a reference to the symbol
// that it cannot see. Therefore, we can't internalize.
SmallPtrSet<GlobalValue *, 8> Used;
collectUsedGlobalVariables(M, Used, /* CompilerUsed */ false);
for (const BasicSymbolRef &Sym : Obj->symbols()) {
GlobalValue *GV = Obj->getSymbolGV(Sym.getRawDataRefImpl());
// Ignore module asm symbols.
if (!GV)
continue;
if (GV->hasAppendingLinkage()) {
Keep.push_back(GV);
continue;
}
if (BitcodeFile::shouldSkip(Sym))
continue;
SymbolBody *B = Bodies[BodyIndex++];
if (!B || &B->repl() != B || !isa<DefinedBitcode>(B))
continue;
switch (GV->getLinkage()) {
default:
break;
case llvm::GlobalValue::LinkOnceAnyLinkage:
GV->setLinkage(GlobalValue::WeakAnyLinkage);
break;
case llvm::GlobalValue::LinkOnceODRLinkage:
GV->setLinkage(GlobalValue::WeakODRLinkage);
break;
}
// We collect the set of symbols we want to internalize here
// and change the linkage after the IRMover executed, i.e. after
// we imported the symbols and satisfied undefined references
// to it. We can't just change linkage here because otherwise
// the IRMover will just rename the symbol.
// Shared libraries need to be handled slightly differently.
// For now, let's be conservative and just never internalize
// symbols when creating a shared library.
if (!Config->Shared && !Config->ExportDynamic && !B->isUsedInRegularObj() &&
!B->MustBeInDynSym)
if (!Used.count(GV))
InternalizedSyms.insert(GV->getName());
Keep.push_back(GV);
}
Mover.move(Obj->takeModule(), Keep,
[](GlobalValue &, IRMover::ValueAdder) {});
}
static std::unique_ptr<Module>
getModuleForFile(LLVMContext &Context, claimed_file &F, const void *View,
ld_plugin_input_file &Info, raw_fd_ostream *ApiFile,
StringSet<> &Internalize, StringSet<> &Maybe,
std::vector<GlobalValue *> &Keep,
StringMap<unsigned> &Realign) {
MemoryBufferRef BufferRef(StringRef((const char *)View, Info.filesize),
Info.name);
ErrorOr<std::unique_ptr<object::IRObjectFile>> ObjOrErr =
object::IRObjectFile::create(BufferRef, Context);
if (std::error_code EC = ObjOrErr.getError())
message(LDPL_FATAL, "Could not read bitcode from file : %s",
EC.message().c_str());
object::IRObjectFile &Obj = **ObjOrErr;
Module &M = Obj.getModule();
M.materializeMetadata();
UpgradeDebugInfo(M);
SmallPtrSet<GlobalValue *, 8> Used;
collectUsedGlobalVariables(M, Used, /*CompilerUsed*/ false);
unsigned SymNum = 0;
for (auto &ObjSym : Obj.symbols()) {
GlobalValue *GV = Obj.getSymbolGV(ObjSym.getRawDataRefImpl());
if (GV && GV->hasAppendingLinkage())
Keep.push_back(GV);
if (shouldSkip(ObjSym.getFlags()))
continue;
ld_plugin_symbol &Sym = F.syms[SymNum];
++SymNum;
ld_plugin_symbol_resolution Resolution =
(ld_plugin_symbol_resolution)Sym.resolution;
if (options::generate_api_file)
*ApiFile << Sym.name << ' ' << getResolutionName(Resolution) << '\n';
if (!GV) {
freeSymName(Sym);
continue; // Asm symbol.
}
ResolutionInfo &Res = ResInfo[Sym.name];
if (Resolution == LDPR_PREVAILING_DEF_IRONLY_EXP && !Res.IsLinkonceOdr)
Resolution = LDPR_PREVAILING_DEF;
// In ThinLTO mode change all prevailing resolutions to LDPR_PREVAILING_DEF.
// For ThinLTO the IR files are compiled through the backend independently,
// so we need to ensure that any prevailing linkonce copy will be emitted
// into the object file by making it weak. Additionally, we can skip the
// IRONLY handling for internalization, which isn't performed in ThinLTO
// mode currently anyway.
if (options::thinlto && (Resolution == LDPR_PREVAILING_DEF_IRONLY_EXP ||
Resolution == LDPR_PREVAILING_DEF_IRONLY))
Resolution = LDPR_PREVAILING_DEF;
GV->setUnnamedAddr(Res.UnnamedAddr);
GV->setVisibility(Res.Visibility);
// Override gold's resolution for common symbols. We want the largest
// one to win.
if (GV->hasCommonLinkage()) {
if (Resolution == LDPR_PREVAILING_DEF_IRONLY)
Res.CommonInternal = true;
if (Resolution == LDPR_PREVAILING_DEF_IRONLY ||
Resolution == LDPR_PREVAILING_DEF)
Res.UseCommon = true;
const DataLayout &DL = GV->getParent()->getDataLayout();
uint64_t Size = DL.getTypeAllocSize(GV->getType()->getElementType());
unsigned Align = GV->getAlignment();
if (Res.UseCommon && Size >= Res.CommonSize) {
// Take GV.
if (Res.CommonInternal)
Resolution = LDPR_PREVAILING_DEF_IRONLY;
else
Resolution = LDPR_PREVAILING_DEF;
cast<GlobalVariable>(GV)->setAlignment(
std::max(Res.CommonAlign, Align));
} else {
// Do not take GV, it's smaller than what we already have in the
// combined module.
Resolution = LDPR_PREEMPTED_IR;
if (Align > Res.CommonAlign)
// Need to raise the alignment though.
Realign[Sym.name] = Align;
}
Res.CommonSize = std::max(Res.CommonSize, Size);
Res.CommonAlign = std::max(Res.CommonAlign, Align);
}
switch (Resolution) {
case LDPR_UNKNOWN:
llvm_unreachable("Unexpected resolution");
case LDPR_RESOLVED_IR:
case LDPR_RESOLVED_EXEC:
case LDPR_RESOLVED_DYN:
case LDPR_PREEMPTED_IR:
case LDPR_PREEMPTED_REG:
break;
case LDPR_UNDEF:
if (!GV->isDeclarationForLinker())
assert(GV->hasComdat());
break;
case LDPR_PREVAILING_DEF_IRONLY: {
Keep.push_back(GV);
// The IR linker has to be able to map this value to a declaration,
// so we can only internalize after linking.
if (!Used.count(GV))
Internalize.insert(GV->getName());
break;
}
case LDPR_PREVAILING_DEF:
Keep.push_back(GV);
// There is a non IR use, so we have to force optimizations to keep this.
switch (GV->getLinkage()) {
default:
break;
case GlobalValue::LinkOnceAnyLinkage:
GV->setLinkage(GlobalValue::WeakAnyLinkage);
break;
case GlobalValue::LinkOnceODRLinkage:
GV->setLinkage(GlobalValue::WeakODRLinkage);
break;
}
break;
case LDPR_PREVAILING_DEF_IRONLY_EXP: {
// We can only check for address uses after we merge the modules. The
// reason is that this GV might have a copy in another module
// and in that module the address might be significant, but that
// copy will be LDPR_PREEMPTED_IR.
Maybe.insert(GV->getName());
Keep.push_back(GV);
break;
}
}
freeSymName(Sym);
}
return Obj.takeModule();
}
bool InternalizePass::runOnModule(Module &M) {
CallGraphWrapperPass *CGPass = getAnalysisIfAvailable<CallGraphWrapperPass>();
CallGraph *CG = CGPass ? &CGPass->getCallGraph() : 0;
CallGraphNode *ExternalNode = CG ? CG->getExternalCallingNode() : 0;
bool Changed = false;
SmallPtrSet<GlobalValue *, 8> Used;
collectUsedGlobalVariables(M, Used, false);
// We must assume that globals in llvm.used have a reference that not even
// the linker can see, so we don't internalize them.
// For llvm.compiler.used the situation is a bit fuzzy. The assembler and
// linker can drop those symbols. If this pass is running as part of LTO,
// one might think that it could just drop llvm.compiler.used. The problem
// is that even in LTO llvm doesn't see every reference. For example,
// we don't see references from function local inline assembly. To be
// conservative, we internalize symbols in llvm.compiler.used, but we
// keep llvm.compiler.used so that the symbol is not deleted by llvm.
for (SmallPtrSet<GlobalValue *, 8>::iterator I = Used.begin(), E = Used.end();
I != E; ++I) {
GlobalValue *V = *I;
ExternalNames.insert(V->getName());
}
// Mark all functions not in the api as internal.
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
if (!shouldInternalize(*I, ExternalNames, OnlyHidden))
continue;
I->setLinkage(GlobalValue::InternalLinkage);
if (ExternalNode)
// Remove a callgraph edge from the external node to this function.
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.
// FIXME: We should probably add this (and the __stack_chk_guard) via some
// type of call-back in CodeGen.
ExternalNames.insert("__stack_chk_fail");
ExternalNames.insert("__stack_chk_guard");
// Mark all global variables with initializers that are not in the api as
// internal as well.
for (Module::global_iterator I = M.global_begin(), E = M.global_end();
I != E; ++I) {
if (!shouldInternalize(*I, ExternalNames, OnlyHidden))
continue;
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 (!shouldInternalize(*I, ExternalNames, OnlyHidden))
continue;
I->setLinkage(GlobalValue::InternalLinkage);
Changed = true;
++NumAliases;
DEBUG(dbgs() << "Internalized alias " << I->getName() << "\n");
}
return Changed;
}
std::unique_ptr<llvm::Module>
klee::linkModules(std::vector<std::unique_ptr<llvm::Module>> &modules,
llvm::StringRef entryFunction, std::string &errorMsg) {
assert(!modules.empty() && "modules list should not be empty");
if (entryFunction.empty()) {
// If no entry function is provided, link all modules together into one
std::unique_ptr<llvm::Module> composite = std::move(modules.back());
modules.pop_back();
// Just link all modules together
for (auto &module : modules) {
if (linkTwoModules(composite.get(), std::move(module), errorMsg))
continue;
// Linking failed
errorMsg = "Linking archive module with composite failed:" + errorMsg;
return nullptr;
}
// clean up every module as we already linked in every module
modules.clear();
return composite;
}
// Starting from the module containing the entry function, resolve unresolved
// dependencies recursively
// search for the module containing the entry function
std::unique_ptr<llvm::Module> composite;
for (auto &module : modules) {
if (!module || !module->getNamedValue(entryFunction))
continue;
if (composite) {
errorMsg =
"Function " + entryFunction.str() +
" defined in different modules (" + module->getModuleIdentifier() +
" already defined in: " + composite->getModuleIdentifier() + ")";
return nullptr;
}
composite = std::move(module);
}
// fail if not found
if (!composite) {
errorMsg = "'" + entryFunction.str() + "' function not found in module.";
return nullptr;
}
while (true) {
std::set<std::string> undefinedSymbols;
GetAllUndefinedSymbols(composite.get(), undefinedSymbols);
// Stop in nothing is undefined
if (undefinedSymbols.empty())
break;
bool merged = false;
for (auto &module : modules) {
if (!module)
continue;
for (auto symbol : undefinedSymbols) {
GlobalValue *GV =
dyn_cast_or_null<GlobalValue>(module->getNamedValue(symbol));
if (!GV || GV->isDeclaration())
continue;
// Found symbol, therefore merge in module
KLEE_DEBUG_WITH_TYPE("klee_linker",
dbgs() << "Found " << GV->getName() << " in "
<< module->getModuleIdentifier() << "\n");
if (linkTwoModules(composite.get(), std::move(module), errorMsg)) {
module = nullptr;
merged = true;
break;
}
// Linking failed
errorMsg = "Linking archive module with composite failed:" + errorMsg;
return nullptr;
}
}
if (!merged)
break;
}
// Condense the module array
std::vector<std::unique_ptr<llvm::Module>> LeftoverModules;
for (auto &module : modules) {
if (module)
LeftoverModules.emplace_back(std::move(module));
}
modules.swap(LeftoverModules);
return composite;
}