// Note that this library should be as fast as possible, reentrant, and
// threadsafe!!
//
-// TODO: Make error message outputs be configurable depending on an option?
+// TODO: Return error messages to caller instead of printing them out directly.
// TODO: Allow passing in an option to ignore the symbol table
//
//===----------------------------------------------------------------------===//
#include "ReaderInternals.h"
#include "llvm/Bytecode/Reader.h"
#include "llvm/Bytecode/Format.h"
-#include "llvm/GlobalVariable.h"
-#include "llvm/Module.h"
-#include "llvm/BasicBlock.h"
-#include "llvm/ConstantVals.h"
+#include "llvm/Constants.h"
#include "llvm/iPHINode.h"
#include "llvm/iOther.h"
-#include <sys/types.h>
-#include <sys/mman.h>
-#include <sys/stat.h>
-#include <fcntl.h>
-#include <unistd.h>
+#include "llvm/Module.h"
+#include "Support/StringExtras.h"
+#include "Config/unistd.h"
+#include "Config/sys/mman.h"
+#include "Config/sys/stat.h"
+#include "Config/sys/types.h"
#include <algorithm>
+#include <memory>
-bool BytecodeParser::getTypeSlot(const Type *Ty, unsigned &Slot) {
- if (Ty->isPrimitiveType()) {
- Slot = Ty->getPrimitiveID();
- } else {
- // Check the method level types first...
- TypeValuesListTy::iterator I = find(MethodTypeValues.begin(),
- MethodTypeValues.end(), Ty);
- if (I != MethodTypeValues.end()) {
- Slot = FirstDerivedTyID+ModuleTypeValues.size()+
- (&*I - &MethodTypeValues[0]);
- } else {
- I = find(ModuleTypeValues.begin(), ModuleTypeValues.end(), Ty);
- if (I == ModuleTypeValues.end()) return true; // Didn't find type!
- Slot = FirstDerivedTyID + (&*I - &ModuleTypeValues[0]);
- }
- }
- //cerr << "getTypeSlot '" << Ty->getName() << "' = " << Slot << endl;
- return false;
+static inline void ALIGN32(const unsigned char *&begin,
+ const unsigned char *end) {
+ if (align32(begin, end))
+ throw std::string("Alignment error in buffer: read past end of block.");
+}
+
+unsigned BytecodeParser::getTypeSlot(const Type *Ty) {
+ if (Ty->isPrimitiveType())
+ return Ty->getPrimitiveID();
+
+ // Check the function level types first...
+ TypeValuesListTy::iterator I = find(FunctionTypeValues.begin(),
+ FunctionTypeValues.end(), Ty);
+ if (I != FunctionTypeValues.end())
+ return FirstDerivedTyID + ModuleTypeValues.size() +
+ (&*I - &FunctionTypeValues[0]);
+
+ I = find(ModuleTypeValues.begin(), ModuleTypeValues.end(), Ty);
+ if (I == ModuleTypeValues.end())
+ throw std::string("Didn't find type in ModuleTypeValues.");
+ return FirstDerivedTyID + (&*I - &ModuleTypeValues[0]);
}
const Type *BytecodeParser::getType(unsigned ID) {
- const Type *T = Type::getPrimitiveType((Type::PrimitiveID)ID);
- if (T) return T;
+ if (ID < Type::NumPrimitiveIDs)
+ if (const Type *T = Type::getPrimitiveType((Type::PrimitiveID)ID))
+ return T;
- //cerr << "Looking up Type ID: " << ID << endl;
+ //cerr << "Looking up Type ID: " << ID << "\n";
- const Value *D = getValue(Type::TypeTy, ID, false);
- if (D == 0) return failure<const Type*>(0);
+ if (ID < Type::NumPrimitiveIDs)
+ if (const Type *T = Type::getPrimitiveType((Type::PrimitiveID)ID))
+ return T; // Asked for a primitive type...
- return cast<Type>(D);
+ // Otherwise, derived types need offset...
+ ID -= FirstDerivedTyID;
+
+ // Is it a module-level type?
+ if (ID < ModuleTypeValues.size())
+ return ModuleTypeValues[ID].get();
+
+ // Nope, is it a function-level type?
+ ID -= ModuleTypeValues.size();
+ if (ID < FunctionTypeValues.size())
+ return FunctionTypeValues[ID].get();
+
+ throw std::string("Illegal type reference!");
}
-int BytecodeParser::insertValue(Value *Val, vector<ValueList> &ValueTab) {
- unsigned type;
- if (getTypeSlot(Val->getType(), type)) return failure<int>(-1);
+unsigned BytecodeParser::insertValue(Value *Val, ValueTable &ValueTab) {
+ assert((!HasImplicitZeroInitializer || !isa<Constant>(Val) ||
+ Val->getType()->isPrimitiveType() ||
+ !cast<Constant>(Val)->isNullValue()) &&
+ "Cannot read null values from bytecode!");
+ unsigned type = getTypeSlot(Val->getType());
assert(type != Type::TypeTyID && "Types should never be insertValue'd!");
- if (ValueTab.size() <= type)
- ValueTab.resize(type+1, ValueList());
+ if (ValueTab.size() <= type) {
+ unsigned OldSize = ValueTab.size();
+ ValueTab.resize(type+1);
+ while (OldSize != type+1)
+ ValueTab[OldSize++] = new ValueList();
+ }
//cerr << "insertValue Values[" << type << "][" << ValueTab[type].size()
- // << "] = " << Val << endl;
- ValueTab[type].push_back(Val);
+ // << "] = " << Val << "\n";
+ ValueTab[type]->push_back(Val);
- return ValueTab[type].size()-1;
-}
-
-Value *BytecodeParser::getValue(const Type *Ty, unsigned oNum, bool Create) {
- unsigned Num = oNum;
- unsigned type; // The type plane it lives in...
+ bool HasOffset = HasImplicitZeroInitializer &&
+ !Val->getType()->isPrimitiveType();
- if (getTypeSlot(Ty, type)) return failure<Value*>(0); // TODO: true
-
- if (type == Type::TypeTyID) { // The 'type' plane has implicit values
- assert(Create == false);
- const Type *T = Type::getPrimitiveType((Type::PrimitiveID)Num);
- if (T) return (Value*)T; // Asked for a primitive type...
+ return ValueTab[type]->size()-1 + HasOffset;
+}
- // Otherwise, derived types need offset...
- Num -= FirstDerivedTyID;
- // Is it a module level type?
- if (Num < ModuleTypeValues.size())
- return (Value*)ModuleTypeValues[Num].get();
+Value *BytecodeParser::getValue(const Type *Ty, unsigned oNum, bool Create) {
+ return getValue(getTypeSlot(Ty), oNum, Create);
+}
- // Nope, is it a method level type?
- Num -= ModuleTypeValues.size();
- if (Num < MethodTypeValues.size())
- return (Value*)MethodTypeValues[Num].get();
+Value *BytecodeParser::getValue(unsigned type, unsigned oNum, bool Create) {
+ assert(type != Type::TypeTyID && "getValue() cannot get types!");
+ assert(type != Type::LabelTyID && "getValue() cannot get blocks!");
+ unsigned Num = oNum;
- return 0;
+ if (HasImplicitZeroInitializer && type >= FirstDerivedTyID) {
+ if (Num == 0)
+ return Constant::getNullValue(getType(type));
+ --Num;
}
if (type < ModuleValues.size()) {
- if (Num < ModuleValues[type].size())
- return ModuleValues[type][Num];
- Num -= ModuleValues[type].size();
+ if (Num < ModuleValues[type]->size())
+ return ModuleValues[type]->getOperand(Num);
+ Num -= ModuleValues[type]->size();
}
- if (Values.size() > type && Values[type].size() > Num)
- return Values[type][Num];
-
- if (!Create) return failure<Value*>(0); // Do not create a placeholder?
-
- Value *d = 0;
- switch (Ty->getPrimitiveID()) {
- case Type::LabelTyID: d = new BBPHolder(Ty, oNum); break;
- case Type::MethodTyID:
- cerr << "Creating method pholder! : " << type << ":" << oNum << " "
- << Ty->getName() << endl;
- d = new MethPHolder(Ty, oNum);
- if (insertValue(d, LateResolveModuleValues) ==-1) return failure<Value*>(0);
- return d;
- default: d = new DefPHolder(Ty, oNum); break;
- }
+ if (Values.size() > type && Values[type]->size() > Num)
+ return Values[type]->getOperand(Num);
- assert(d != 0 && "How did we not make something?");
- if (insertValue(d, LateResolveValues) == -1) return failure<Value*>(0);
- return d;
-}
+ if (!Create) return 0; // Do not create a placeholder?
-bool BytecodeParser::postResolveValues(ValueTable &ValTab) {
- bool Error = false;
- for (unsigned ty = 0; ty < ValTab.size(); ++ty) {
- ValueList &DL = ValTab[ty];
- unsigned Size;
- while ((Size = DL.size())) {
- unsigned IDNumber = getValueIDNumberFromPlaceHolder(DL[Size-1]);
-
- Value *D = DL[Size-1];
- DL.pop_back();
-
- Value *NewDef = getValue(D->getType(), IDNumber, false);
- if (NewDef == 0) {
- Error = true; // Unresolved thinger
- cerr << "Unresolvable reference found: <"
- << D->getType()->getDescription() << ">:" << IDNumber << "!\n";
- } else {
- // Fixup all of the uses of this placeholder def...
- D->replaceAllUsesWith(NewDef);
-
- // Now that all the uses are gone, delete the placeholder...
- // If we couldn't find a def (error case), then leak a little
- delete D; // memory, 'cause otherwise we can't remove all uses!
- }
- }
- }
+ std::pair<unsigned,unsigned> KeyValue(type, oNum);
+ std::map<std::pair<unsigned,unsigned>, Value*>::iterator I =
+ ForwardReferences.lower_bound(KeyValue);
+ if (I != ForwardReferences.end() && I->first == KeyValue)
+ return I->second; // We have already created this placeholder
- return Error;
+ Value *Val = new Argument(getType(type));
+ ForwardReferences.insert(I, std::make_pair(KeyValue, Val));
+ return Val;
}
-bool BytecodeParser::ParseBasicBlock(const uchar *&Buf, const uchar *EndBuf,
- BasicBlock *&BB) {
- BB = new BasicBlock();
+/// getBasicBlock - Get a particular numbered basic block, which might be a
+/// forward reference. This works together with ParseBasicBlock to handle these
+/// forward references in a clean manner.
+///
+BasicBlock *BytecodeParser::getBasicBlock(unsigned ID) {
+ // Make sure there is room in the table...
+ if (ParsedBasicBlocks.size() <= ID) ParsedBasicBlocks.resize(ID+1);
+
+ // First check to see if this is a backwards reference, i.e., ParseBasicBlock
+ // has already created this block, or if the forward reference has already
+ // been created.
+ if (ParsedBasicBlocks[ID])
+ return ParsedBasicBlocks[ID];
+
+ // Otherwise, the basic block has not yet been created. Do so and add it to
+ // the ParsedBasicBlocks list.
+ return ParsedBasicBlocks[ID] = new BasicBlock();
+}
- while (Buf < EndBuf) {
- Instruction *Inst;
- if (ParseInstruction(Buf, EndBuf, Inst)) {
- delete BB;
- return failure(true);
- }
+/// getConstantValue - Just like getValue, except that it returns a null pointer
+/// only on error. It always returns a constant (meaning that if the value is
+/// defined, but is not a constant, that is an error). If the specified
+/// constant hasn't been parsed yet, a placeholder is defined and used. Later,
+/// after the real value is parsed, the placeholder is eliminated.
+///
+Constant *BytecodeParser::getConstantValue(const Type *Ty, unsigned Slot) {
+ if (Value *V = getValue(Ty, Slot, false))
+ return dyn_cast<Constant>(V); // If we already have the value parsed...
+
+ std::pair<const Type*, unsigned> Key(Ty, Slot);
+ GlobalRefsType::iterator I = GlobalRefs.lower_bound(Key);
+
+ if (I != GlobalRefs.end() && I->first == Key) {
+ BCR_TRACE(5, "Previous forward ref found!\n");
+ return cast<Constant>(I->second);
+ } else {
+ // Create a placeholder for the constant reference and
+ // keep track of the fact that we have a forward ref to recycle it
+ BCR_TRACE(5, "Creating new forward ref to a constant!\n");
+ Constant *C = new ConstPHolder(Ty, Slot);
+
+ // Keep track of the fact that we have a forward ref to recycle it
+ GlobalRefs.insert(I, std::make_pair(Key, C));
+ return C;
+ }
+}
- if (Inst == 0) { delete BB; return failure(true); }
- if (insertValue(Inst, Values) == -1) { delete BB; return failure(true); }
- BB->getInstList().push_back(Inst);
+BasicBlock *BytecodeParser::ParseBasicBlock(const unsigned char *&Buf,
+ const unsigned char *EndBuf,
+ unsigned BlockNo) {
+ BasicBlock *BB;
+ if (ParsedBasicBlocks.size() == BlockNo)
+ ParsedBasicBlocks.push_back(BB = new BasicBlock());
+ else if (ParsedBasicBlocks[BlockNo] == 0)
+ BB = ParsedBasicBlocks[BlockNo] = new BasicBlock();
+ else
+ BB = ParsedBasicBlocks[BlockNo];
+ while (Buf < EndBuf) {
+ Instruction *Inst = ParseInstruction(Buf, EndBuf);
+ insertValue(Inst, Values);
+ BB->getInstList().push_back(Inst);
BCR_TRACE(4, Inst);
}
- return false;
+ return BB;
}
-bool BytecodeParser::ParseSymbolTable(const uchar *&Buf, const uchar *EndBuf,
- SymbolTable *ST) {
+void BytecodeParser::ParseSymbolTable(const unsigned char *&Buf,
+ const unsigned char *EndBuf,
+ SymbolTable *ST,
+ Function *CurrentFunction) {
while (Buf < EndBuf) {
// Symtab block header: [num entries][type id number]
unsigned NumEntries, Typ;
if (read_vbr(Buf, EndBuf, NumEntries) ||
- read_vbr(Buf, EndBuf, Typ)) return failure(true);
+ read_vbr(Buf, EndBuf, Typ)) throw Error_readvbr;
const Type *Ty = getType(Typ);
- if (Ty == 0) return failure(true);
-
- BCR_TRACE(3, "Plane Type: '" << Ty << "' with " << NumEntries <<
- " entries\n");
+ BCR_TRACE(3, "Plane Type: '" << *Ty << "' with " << NumEntries <<
+ " entries\n");
for (unsigned i = 0; i < NumEntries; ++i) {
// Symtab entry: [def slot #][name]
unsigned slot;
- if (read_vbr(Buf, EndBuf, slot)) return failure(true);
- string Name;
+ if (read_vbr(Buf, EndBuf, slot)) throw Error_readvbr;
+ std::string Name;
if (read(Buf, EndBuf, Name, false)) // Not aligned...
- return failure(true);
-
- Value *D = getValue(Ty, slot, false); // Find mapping...
- if (D == 0) {
- BCR_TRACE(3, "FAILED LOOKUP: Slot #" << slot << endl);
- return failure(true);
- }
- BCR_TRACE(4, "Map: '" << Name << "' to #" << slot << ":" << D;
- if (!isa<Instruction>(D)) cerr << endl);
-
- D->setName(Name, ST);
+ throw std::string("Buffer not aligned.");
+
+ Value *V = 0;
+ if (Typ == Type::TypeTyID)
+ V = (Value*)getType(slot);
+ else if (Typ == Type::LabelTyID) {
+ if (CurrentFunction) {
+ // FIXME: THIS IS N^2!!!
+ Function::iterator BlockIterator = CurrentFunction->begin();
+ std::advance(BlockIterator, slot);
+ V = BlockIterator;
+ }
+ } else
+ V = getValue(Typ, slot, false); // Find mapping...
+ if (V == 0) throw std::string("Failed value look-up.");
+ BCR_TRACE(4, "Map: '" << Name << "' to #" << slot << ":" << *V;
+ if (!isa<Instruction>(V)) std::cerr << "\n");
+
+ V->setName(Name, ST);
}
}
- if (Buf > EndBuf) return failure(true);
- return false;
+ if (Buf > EndBuf) throw std::string("Tried to read past end of buffer.");
}
-// DeclareNewGlobalValue - Patch up forward references to global values in the
-// form of ConstantPointerRef.
-//
-void BytecodeParser::DeclareNewGlobalValue(GlobalValue *GV, unsigned Slot) {
- // Check to see if there is a forward reference to this global variable...
- // if there is, eliminate it and patch the reference to use the new def'n.
- GlobalRefsType::iterator I = GlobalRefs.find(make_pair(GV->getType(), Slot));
-
- if (I != GlobalRefs.end()) {
- GlobalVariable *OldGV = I->second; // Get the placeholder...
- BCR_TRACE(3, "Mutating CPPR Forward Ref!\n");
-
- // Loop over all of the uses of the GlobalValue. The only thing they are
- // allowed to be at this point is ConstantPointerRef's.
- assert(OldGV->use_size() == 1 && "Only one reference should exist!");
- while (!OldGV->use_empty()) {
- User *U = OldGV->use_back(); // Must be a ConstantPointerRef...
- ConstantPointerRef *CPPR = cast<ConstantPointerRef>(U);
- assert(CPPR->getValue() == OldGV && "Something isn't happy");
-
- BCR_TRACE(4, "Mutating Forward Ref!\n");
-
- // Change the const pool reference to point to the real global variable
- // now. This should drop a use from the OldGV.
- CPPR->mutateReference(GV);
- }
-
- // Remove GV from the module...
- GV->getParent()->getGlobalList().remove(OldGV);
- delete OldGV; // Delete the old placeholder
-
- // Remove the map entry for the global now that it has been created...
- GlobalRefs.erase(I);
- }
-}
+void BytecodeParser::ResolveReferencesToValue(Value *NewV, unsigned Slot) {
+ GlobalRefsType::iterator I = GlobalRefs.find(std::make_pair(NewV->getType(),
+ Slot));
+ if (I == GlobalRefs.end()) return; // Never forward referenced?
-bool BytecodeParser::ParseMethod(const uchar *&Buf, const uchar *EndBuf,
- Module *C) {
- // Clear out the local values table...
- Values.clear();
- if (MethodSignatureList.empty()) {
- Error = "Method found, but MethodSignatureList empty!";
- return failure(true); // Unexpected method!
- }
+ BCR_TRACE(3, "Mutating forward refs!\n");
+ Value *VPH = I->second; // Get the placeholder...
- const PointerType *PMTy = MethodSignatureList.front().first; // PtrMeth
- const MethodType *MTy = dyn_cast<const MethodType>(PMTy->getElementType());
- if (MTy == 0) return failure(true); // Not ptr to method!
+ VPH->replaceAllUsesWith(NewV);
- unsigned isInternal;
- if (read_vbr(Buf, EndBuf, isInternal)) return failure(true);
+ // If this is a global variable being resolved, remove the placeholder from
+ // the module...
+ if (GlobalValue* GVal = dyn_cast<GlobalValue>(NewV))
+ GVal->getParent()->getGlobalList().remove(cast<GlobalVariable>(VPH));
- unsigned MethSlot = MethodSignatureList.front().second;
- MethodSignatureList.pop_front();
- Method *M = new Method(MTy, isInternal != 0);
+ delete VPH; // Delete the old placeholder
+ GlobalRefs.erase(I); // Remove the map entry for it
+}
- BCR_TRACE(2, "METHOD TYPE: " << MTy << endl);
+void BytecodeParser::ParseFunction(const unsigned char *&Buf,
+ const unsigned char *EndBuf) {
+ if (FunctionSignatureList.empty())
+ throw std::string("FunctionSignatureList empty!");
+
+ Function *F = FunctionSignatureList.back().first;
+ unsigned FunctionSlot = FunctionSignatureList.back().second;
+ FunctionSignatureList.pop_back();
+
+ // Save the information for future reading of the function
+ LazyFunctionInfo *LFI = new LazyFunctionInfo();
+ LFI->Buf = Buf; LFI->EndBuf = EndBuf; LFI->FunctionSlot = FunctionSlot;
+ LazyFunctionLoadMap[F] = LFI;
+ // Pretend we've `parsed' this function
+ Buf = EndBuf;
+}
- const MethodType::ParamTypes &Params = MTy->getParamTypes();
- for (MethodType::ParamTypes::const_iterator It = Params.begin();
- It != Params.end(); ++It) {
- MethodArgument *MA = new MethodArgument(*It);
- if (insertValue(MA, Values) == -1) {
- Error = "Error reading method arguments!\n";
- delete M; return failure(true);
- }
- M->getArgumentList().push_back(MA);
+void BytecodeParser::materializeFunction(Function* F) {
+ // Find {start, end} pointers and slot in the map. If not there, we're done.
+ std::map<Function*, LazyFunctionInfo*>::iterator Fi =
+ LazyFunctionLoadMap.find(F);
+ if (Fi == LazyFunctionLoadMap.end()) return;
+
+ LazyFunctionInfo *LFI = Fi->second;
+ const unsigned char *Buf = LFI->Buf;
+ const unsigned char *EndBuf = LFI->EndBuf;
+ unsigned FunctionSlot = LFI->FunctionSlot;
+ LazyFunctionLoadMap.erase(Fi);
+ delete LFI;
+
+ GlobalValue::LinkageTypes Linkage = GlobalValue::ExternalLinkage;
+
+ if (!hasInternalMarkerOnly) {
+ unsigned LinkageType;
+ if (read_vbr(Buf, EndBuf, LinkageType))
+ throw std::string("ParseFunction: Error reading from buffer.");
+ if (LinkageType & ~0x3)
+ throw std::string("Invalid linkage type for Function.");
+ Linkage = (GlobalValue::LinkageTypes)LinkageType;
+ } else {
+ // We used to only support two linkage models: internal and external
+ unsigned isInternal;
+ if (read_vbr(Buf, EndBuf, isInternal))
+ throw std::string("ParseFunction: Error reading from buffer.");
+ if (isInternal) Linkage = GlobalValue::InternalLinkage;
}
+ F->setLinkage(Linkage);
+
+ const FunctionType::ParamTypes &Params =F->getFunctionType()->getParamTypes();
+ Function::aiterator AI = F->abegin();
+ for (FunctionType::ParamTypes::const_iterator It = Params.begin();
+ It != Params.end(); ++It, ++AI)
+ insertValue(AI, Values);
+
+ // Keep track of how many basic blocks we have read in...
+ unsigned BlockNum = 0;
+
while (Buf < EndBuf) {
unsigned Type, Size;
- const uchar *OldBuf = Buf;
- if (readBlock(Buf, EndBuf, Type, Size)) {
- Error = "Error reading Method level block!";
- delete M; return failure(true);
- }
+ const unsigned char *OldBuf = Buf;
+ readBlock(Buf, EndBuf, Type, Size);
switch (Type) {
- case BytecodeFormat::ConstantPool:
+ case BytecodeFormat::ConstantPool: {
BCR_TRACE(2, "BLOCK BytecodeFormat::ConstantPool: {\n");
- if (ParseConstantPool(Buf, Buf+Size, Values, MethodTypeValues)) {
- delete M; return failure(true);
- }
+ ParseConstantPool(Buf, Buf+Size, Values, FunctionTypeValues);
break;
+ }
case BytecodeFormat::BasicBlock: {
BCR_TRACE(2, "BLOCK BytecodeFormat::BasicBlock: {\n");
- BasicBlock *BB;
- if (ParseBasicBlock(Buf, Buf+Size, BB) ||
- insertValue(BB, Values) == -1) {
- delete M; return failure(true); // Parse error... :(
- }
-
- M->getBasicBlocks().push_back(BB);
+ BasicBlock *BB = ParseBasicBlock(Buf, Buf+Size, BlockNum++);
+ F->getBasicBlockList().push_back(BB);
break;
}
- case BytecodeFormat::SymbolTable:
+ case BytecodeFormat::SymbolTable: {
BCR_TRACE(2, "BLOCK BytecodeFormat::SymbolTable: {\n");
- if (ParseSymbolTable(Buf, Buf+Size, M->getSymbolTableSure())) {
- delete M; return failure(true);
- }
+ ParseSymbolTable(Buf, Buf+Size, &F->getSymbolTable(), F);
break;
+ }
default:
BCR_TRACE(2, "BLOCK <unknown>:ignored! {\n");
Buf += Size;
- if (OldBuf > Buf) return failure(true); // Wrap around!
+ if (OldBuf > Buf)
+ throw std::string("Wrapped around reading bytecode.");
break;
}
BCR_TRACE(2, "} end block\n");
- if (align32(Buf, EndBuf)) {
- Error = "Error aligning Method level block!";
- delete M; // Malformed bc file, read past end of block.
- return failure(true);
- }
- }
-
- if (postResolveValues(LateResolveValues) ||
- postResolveValues(LateResolveModuleValues)) {
- Error = "Error resolving method values!";
- delete M; return failure(true); // Unresolvable references!
+ // Malformed bc file if read past end of block.
+ ALIGN32(Buf, EndBuf);
}
- Value *MethPHolder = getValue(PMTy, MethSlot, false);
- assert(MethPHolder && "Something is broken no placeholder found!");
- assert(isa<Method>(MethPHolder) && "Not a method?");
-
- unsigned type; // Type slot
- assert(!getTypeSlot(MTy, type) && "How can meth type not exist?");
- getTypeSlot(PMTy, type);
+ // Make sure there were no references to non-existant basic blocks.
+ if (BlockNum != ParsedBasicBlocks.size())
+ throw std::string("Illegal basic block operand reference");
+ ParsedBasicBlocks.clear();
- C->getMethodList().push_back(M);
- // Replace placeholder with the real method pointer...
- ModuleValues[type][MethSlot] = M;
+ // Resolve forward references
+ while (!ForwardReferences.empty()) {
+ std::map<std::pair<unsigned,unsigned>, Value*>::iterator I = ForwardReferences.begin();
+ unsigned type = I->first.first;
+ unsigned Slot = I->first.second;
+ Value *PlaceHolder = I->second;
+ ForwardReferences.erase(I);
- // Clear out method level types...
- MethodTypeValues.clear();
+ Value *NewVal = getValue(type, Slot, false);
+ if (NewVal == 0)
+ throw std::string("Unresolvable reference found: <" +
+ PlaceHolder->getType()->getDescription() + ">:" +
+ utostr(Slot) + ".");
- // If anyone is using the placeholder make them use the real method instead
- MethPHolder->replaceAllUsesWith(M);
-
- // We don't need the placeholder anymore!
- delete MethPHolder;
-
- // If the method is empty, we don't need the method argument entries...
- if (M->isExternal())
- M->getArgumentList().delete_all();
+ // Fixup all of the uses of this placeholder def...
+ PlaceHolder->replaceAllUsesWith(NewVal);
+
+ // Now that all the uses are gone, delete the placeholder...
+ // If we couldn't find a def (error case), then leak a little
+ // memory, because otherwise we can't remove all uses!
+ delete PlaceHolder;
+ }
- DeclareNewGlobalValue(M, MethSlot);
+ // Clear out function-level types...
+ FunctionTypeValues.clear();
- return false;
+ freeTable(Values);
}
-bool BytecodeParser::ParseModuleGlobalInfo(const uchar *&Buf, const uchar *End,
- Module *Mod) {
- if (!MethodSignatureList.empty()) {
- Error = "Two ModuleGlobalInfo packets found!";
- return failure(true); // Two ModuleGlobal blocks?
- }
+void BytecodeParser::ParseModuleGlobalInfo(const unsigned char *&Buf,
+ const unsigned char *End) {
+ if (!FunctionSignatureList.empty())
+ throw std::string("Two ModuleGlobalInfo packets found!");
// Read global variables...
unsigned VarType;
- if (read_vbr(Buf, End, VarType)) return failure(true);
+ if (read_vbr(Buf, End, VarType)) throw Error_readvbr;
while (VarType != Type::VoidTyID) { // List is terminated by Void
- // VarType Fields: bit0 = isConstant, bit1 = hasInitializer,
- // bit2 = isInternal, bit3+ = slot#
- const Type *Ty = getType(VarType >> 3);
- if (!Ty || !Ty->isPointerType()) {
- Error = "Global not pointer type! Ty = " + Ty->getDescription();
- return failure(true);
+ unsigned SlotNo;
+ GlobalValue::LinkageTypes Linkage;
+
+ if (!hasInternalMarkerOnly) {
+ // VarType Fields: bit0 = isConstant, bit1 = hasInitializer,
+ // bit2,3 = Linkage, bit4+ = slot#
+ SlotNo = VarType >> 4;
+ Linkage = (GlobalValue::LinkageTypes)((VarType >> 2) & 3);
+ } else {
+ // VarType Fields: bit0 = isConstant, bit1 = hasInitializer,
+ // bit2 = isInternal, bit3+ = slot#
+ SlotNo = VarType >> 3;
+ Linkage = (VarType & 4) ? GlobalValue::InternalLinkage :
+ GlobalValue::ExternalLinkage;
}
- const PointerType *PTy = cast<const PointerType>(Ty);
- const Type *ElTy = PTy->getElementType();
+ const Type *Ty = getType(SlotNo);
+ if (!isa<PointerType>(Ty))
+ throw std::string("Global not pointer type! Ty = " +
+ Ty->getDescription());
- Constant *Initializer = 0;
- if (VarType & 2) { // Does it have an initalizer?
- // Do not improvise... values must have been stored in the constant pool,
- // which should have been read before now.
- //
- unsigned InitSlot;
- if (read_vbr(Buf, End, InitSlot)) return failure(true);
-
- Value *V = getValue(ElTy, InitSlot, false);
- if (V == 0) return failure(true);
- Initializer = cast<Constant>(V);
- }
+ const Type *ElTy = cast<PointerType>(Ty)->getElementType();
// Create the global variable...
- GlobalVariable *GV = new GlobalVariable(ElTy, VarType & 1, VarType & 4,
- Initializer);
- int DestSlot = insertValue(GV, ModuleValues);
- if (DestSlot == -1) return failure(true);
-
- Mod->getGlobalList().push_back(GV);
-
- DeclareNewGlobalValue(GV, unsigned(DestSlot));
-
- BCR_TRACE(2, "Global Variable of type: " << PTy->getDescription()
- << " into slot #" << DestSlot << endl);
+ GlobalVariable *GV = new GlobalVariable(ElTy, VarType & 1, Linkage,
+ 0, "", TheModule);
+ BCR_TRACE(2, "Global Variable of type: " << *Ty << "\n");
+ ResolveReferencesToValue(GV, insertValue(GV, ModuleValues));
- if (read_vbr(Buf, End, VarType)) return failure(true);
+ if (VarType & 2) { // Does it have an initializer?
+ unsigned InitSlot;
+ if (read_vbr(Buf, End, InitSlot)) throw Error_readvbr;
+ GlobalInits.push_back(std::make_pair(GV, InitSlot));
+ }
+ if (read_vbr(Buf, End, VarType)) throw Error_readvbr;
}
- // Read the method signatures for all of the methods that are coming, and
- // create fillers in the Value tables.
- unsigned MethSignature;
- if (read_vbr(Buf, End, MethSignature)) return failure(true);
- while (MethSignature != Type::VoidTyID) { // List is terminated by Void
- const Type *Ty = getType(MethSignature);
- if (!Ty || !isa<PointerType>(Ty) ||
- !isa<MethodType>(cast<PointerType>(Ty)->getElementType())) {
- Error = "Method not ptr to meth type! Ty = " + Ty->getDescription();
- return failure(true);
- }
-
- // We create methods by passing the underlying MethodType to create...
+ // Read the function objects for all of the functions that are coming
+ unsigned FnSignature;
+ if (read_vbr(Buf, End, FnSignature)) throw Error_readvbr;
+ while (FnSignature != Type::VoidTyID) { // List is terminated by Void
+ const Type *Ty = getType(FnSignature);
+ if (!isa<PointerType>(Ty) ||
+ !isa<FunctionType>(cast<PointerType>(Ty)->getElementType()))
+ throw std::string("Function not ptr to func type! Ty = " +
+ Ty->getDescription());
+
+ // We create functions by passing the underlying FunctionType to create...
Ty = cast<PointerType>(Ty)->getElementType();
- // When the ModuleGlobalInfo section is read, we load the type of each
- // method and the 'ModuleValues' slot that it lands in. We then load a
- // placeholder into its slot to reserve it. When the method is loaded, this
- // placeholder is replaced.
+ // When the ModuleGlobalInfo section is read, we load the type of each
+ // function and the 'ModuleValues' slot that it lands in. We then load a
+ // placeholder into its slot to reserve it. When the function is loaded,
+ // this placeholder is replaced.
// Insert the placeholder...
- Value *Val = new MethPHolder(Ty, 0);
- if (insertValue(Val, ModuleValues) == -1) return failure(true);
+ Function *Func = new Function(cast<FunctionType>(Ty),
+ GlobalValue::InternalLinkage, "", TheModule);
+ unsigned DestSlot = insertValue(Func, ModuleValues);
+ ResolveReferencesToValue(Func, DestSlot);
- // Figure out which entry of its typeslot it went into...
- unsigned TypeSlot;
- if (getTypeSlot(Val->getType(), TypeSlot)) return failure(true);
-
- unsigned SlotNo = ModuleValues[TypeSlot].size()-1;
-
- // Keep track of this information in a linked list that is emptied as
- // methods are loaded...
+ // Keep track of this information in a list that is emptied as functions are
+ // loaded...
//
- MethodSignatureList.push_back(
- make_pair(cast<const PointerType>(Val->getType()), SlotNo));
- if (read_vbr(Buf, End, MethSignature)) return failure(true);
- BCR_TRACE(2, "Method of type: " << Ty << endl);
+ FunctionSignatureList.push_back(std::make_pair(Func, DestSlot));
+
+ if (read_vbr(Buf, End, FnSignature)) throw Error_readvbr;
+ BCR_TRACE(2, "Function of type: " << Ty << "\n");
}
- if (align32(Buf, End)) return failure(true);
+ ALIGN32(Buf, End);
+
+ // Now that the function signature list is set up, reverse it so that we can
+ // remove elements efficiently from the back of the vector.
+ std::reverse(FunctionSignatureList.begin(), FunctionSignatureList.end());
// This is for future proofing... in the future extra fields may be added that
// we don't understand, so we transparently ignore them.
//
Buf = End;
- return false;
}
-bool BytecodeParser::ParseModule(const uchar *Buf, const uchar *EndBuf,
- Module *&C) {
+void BytecodeParser::ParseVersionInfo(const unsigned char *&Buf,
+ const unsigned char *EndBuf) {
+ unsigned Version;
+ if (read_vbr(Buf, EndBuf, Version)) throw Error_readvbr;
- unsigned Type, Size;
- if (readBlock(Buf, EndBuf, Type, Size)) return failure(true);
- if (Type != BytecodeFormat::Module || Buf+Size != EndBuf) {
- Error = "Expected Module packet!";
- return failure(true); // Hrm, not a class?
+ // Unpack version number: low four bits are for flags, top bits = version
+ Module::Endianness Endianness;
+ Module::PointerSize PointerSize;
+ Endianness = (Version & 1) ? Module::BigEndian : Module::LittleEndian;
+ PointerSize = (Version & 2) ? Module::Pointer64 : Module::Pointer32;
+
+ bool hasNoEndianness = Version & 4;
+ bool hasNoPointerSize = Version & 8;
+
+ RevisionNum = Version >> 4;
+
+ // Default values for the current bytecode version
+ HasImplicitZeroInitializer = true;
+ hasInternalMarkerOnly = false;
+ FirstDerivedTyID = 14;
+
+ switch (RevisionNum) {
+ case 0: // Initial revision
+ // Version #0 didn't have any of the flags stored correctly, and in fact as
+ // only valid with a 14 in the flags values. Also, it does not support
+ // encoding zero initializers for arrays compactly.
+ //
+ if (Version != 14) throw std::string("Unknown revision 0 flags?");
+ HasImplicitZeroInitializer = false;
+ Endianness = Module::BigEndian;
+ PointerSize = Module::Pointer64;
+ hasInternalMarkerOnly = true;
+ hasNoEndianness = hasNoPointerSize = false;
+ break;
+ case 1:
+ // Version #1 has four bit fields: isBigEndian, hasLongPointers,
+ // hasNoEndianness, and hasNoPointerSize.
+ hasInternalMarkerOnly = true;
+ break;
+ case 2:
+ // Version #2 added information about all 4 linkage types instead of just
+ // having internal and external.
+ break;
+ default:
+ throw std::string("Unknown bytecode version number!");
}
+ if (hasNoEndianness) Endianness = Module::AnyEndianness;
+ if (hasNoPointerSize) PointerSize = Module::AnyPointerSize;
+
+ TheModule->setEndianness(Endianness);
+ TheModule->setPointerSize(PointerSize);
+ BCR_TRACE(1, "Bytecode Rev = " << (unsigned)RevisionNum << "\n");
+ BCR_TRACE(1, "Endianness/PointerSize = " << Endianness << ","
+ << PointerSize << "\n");
+ BCR_TRACE(1, "HasImplicitZeroInit = " << HasImplicitZeroInitializer << "\n");
+}
+
+void BytecodeParser::ParseModule(const unsigned char *Buf,
+ const unsigned char *EndBuf) {
+ unsigned Type, Size;
+ readBlock(Buf, EndBuf, Type, Size);
+ if (Type != BytecodeFormat::Module || Buf+Size != EndBuf)
+ throw std::string("Expected Module packet! B: "+
+ utostr((unsigned)(intptr_t)Buf) + ", S: "+utostr(Size)+
+ " E: "+utostr((unsigned)(intptr_t)EndBuf)); // Hrm, not a class?
+
BCR_TRACE(0, "BLOCK BytecodeFormat::Module: {\n");
- MethodSignatureList.clear(); // Just in case...
+ FunctionSignatureList.clear(); // Just in case...
// Read into instance variables...
- if (read_vbr(Buf, EndBuf, FirstDerivedTyID)) return failure(true);
- if (align32(Buf, EndBuf)) return failure(true);
- BCR_TRACE(1, "FirstDerivedTyID = " << FirstDerivedTyID << "\n");
+ ParseVersionInfo(Buf, EndBuf);
+ ALIGN32(Buf, EndBuf);
- TheModule = C = new Module();
while (Buf < EndBuf) {
- const uchar *OldBuf = Buf;
- if (readBlock(Buf, EndBuf, Type, Size)) { delete C; return failure(true); }
+ const unsigned char *OldBuf = Buf;
+ readBlock(Buf, EndBuf, Type, Size);
switch (Type) {
- case BytecodeFormat::ConstantPool:
- BCR_TRACE(1, "BLOCK BytecodeFormat::ConstantPool: {\n");
- if (ParseConstantPool(Buf, Buf+Size, ModuleValues, ModuleTypeValues)) {
- delete C; return failure(true);
- }
+ case BytecodeFormat::GlobalTypePlane:
+ BCR_TRACE(1, "BLOCK BytecodeFormat::GlobalTypePlane: {\n");
+ ParseGlobalTypes(Buf, Buf+Size);
break;
case BytecodeFormat::ModuleGlobalInfo:
BCR_TRACE(1, "BLOCK BytecodeFormat::ModuleGlobalInfo: {\n");
+ ParseModuleGlobalInfo(Buf, Buf+Size);
+ break;
- if (ParseModuleGlobalInfo(Buf, Buf+Size, C)) {
- delete C; return failure(true);
- }
+ case BytecodeFormat::ConstantPool:
+ BCR_TRACE(1, "BLOCK BytecodeFormat::ConstantPool: {\n");
+ ParseConstantPool(Buf, Buf+Size, ModuleValues, ModuleTypeValues);
break;
- case BytecodeFormat::Method: {
- BCR_TRACE(1, "BLOCK BytecodeFormat::Method: {\n");
- if (ParseMethod(Buf, Buf+Size, C)) {
- delete C; return failure(true); // Error parsing method
- }
+ case BytecodeFormat::Function: {
+ BCR_TRACE(1, "BLOCK BytecodeFormat::Function: {\n");
+ ParseFunction(Buf, Buf+Size);
break;
}
case BytecodeFormat::SymbolTable:
BCR_TRACE(1, "BLOCK BytecodeFormat::SymbolTable: {\n");
- if (ParseSymbolTable(Buf, Buf+Size, C->getSymbolTableSure())) {
- delete C; return failure(true);
- }
+ ParseSymbolTable(Buf, Buf+Size, &TheModule->getSymbolTable(), 0);
break;
default:
- Error = "Expected Module Block!";
Buf += Size;
- if (OldBuf > Buf) return failure(true); // Wrap around!
+ if (OldBuf > Buf) throw std::string("Expected Module Block!");
break;
}
BCR_TRACE(1, "} end block\n");
- if (align32(Buf, EndBuf)) { delete C; return failure(true); }
- }
-
- if (!MethodSignatureList.empty()) { // Expected more methods!
- Error = "Method expected, but bytecode stream at end!";
- return failure(true);
+ ALIGN32(Buf, EndBuf);
}
- BCR_TRACE(0, "} end block\n\n");
- return false;
-}
-
-Module *BytecodeParser::ParseBytecode(const uchar *Buf, const uchar *EndBuf) {
- LateResolveValues.clear();
- unsigned Sig;
- // Read and check signature...
- if (read(Buf, EndBuf, Sig) ||
- Sig != ('l' | ('l' << 8) | ('v' << 16) | 'm' << 24)) {
- Error = "Invalid bytecode signature!";
- return failure<Module*>(0); // Invalid signature!
+ // After the module constant pool has been read, we can safely initialize
+ // global variables...
+ while (!GlobalInits.empty()) {
+ GlobalVariable *GV = GlobalInits.back().first;
+ unsigned Slot = GlobalInits.back().second;
+ GlobalInits.pop_back();
+
+ // Look up the initializer value...
+ if (Value *V = getValue(GV->getType()->getElementType(), Slot, false)) {
+ if (GV->hasInitializer())
+ throw std::string("Global *already* has an initializer?!");
+ GV->setInitializer(cast<Constant>(V));
+ } else
+ throw std::string("Cannot find initializer value.");
}
- Module *Result;
- if (ParseModule(Buf, EndBuf, Result)) return 0;
- return Result;
-}
+ if (!FunctionSignatureList.empty())
+ throw std::string("Function expected, but bytecode stream ended!");
-
-Module *ParseBytecodeBuffer(const uchar *Buffer, unsigned Length) {
- BytecodeParser Parser;
- return Parser.ParseBytecode(Buffer, Buffer+Length);
+ BCR_TRACE(0, "} end block\n\n");
}
-// Parse and return a class file...
-//
-Module *ParseBytecodeFile(const string &Filename, string *ErrorStr) {
- struct stat StatBuf;
- Module *Result = 0;
-
- if (Filename != string("-")) { // Read from a file...
- int FD = open(Filename.c_str(), O_RDONLY);
- if (FD == -1) {
- if (ErrorStr) *ErrorStr = "Error opening file!";
- return failure<Module*>(0);
- }
-
- if (fstat(FD, &StatBuf) == -1) { close(FD); return failure<Module*>(0); }
+void
+BytecodeParser::ParseBytecode(const unsigned char *Buf, unsigned Length,
+ const std::string &ModuleID) {
- int Length = StatBuf.st_size;
- if (Length == 0) {
- if (ErrorStr) *ErrorStr = "Error stat'ing file!";
- close(FD); return failure<Module*>(0);
- }
- uchar *Buffer = (uchar*)mmap(0, Length, PROT_READ,
- MAP_PRIVATE, FD, 0);
- if (Buffer == (uchar*)-1) {
- if (ErrorStr) *ErrorStr = "Error mmapping file!";
- close(FD); return failure<Module*>(0);
- }
+ unsigned char *EndBuf = (unsigned char*)(Buf + Length);
- BytecodeParser Parser;
- Result = Parser.ParseBytecode(Buffer, Buffer+Length);
-
- munmap((char*)Buffer, Length);
- close(FD);
- if (ErrorStr) *ErrorStr = Parser.getError();
- } else { // Read from stdin
- size_t FileSize = 0;
- int BlockSize;
- uchar Buffer[4096], *FileData = 0;
- while ((BlockSize = read(0, Buffer, 4))) {
- if (BlockSize == -1) { free(FileData); return failure<Module*>(0); }
-
- FileData = (uchar*)realloc(FileData, FileSize+BlockSize);
- memcpy(FileData+FileSize, Buffer, BlockSize);
- FileSize += BlockSize;
- }
-
- if (FileSize == 0) {
- if (ErrorStr) *ErrorStr = "Standard Input empty!";
- free(FileData); return failure<Module*>(0);
- }
-
-#define ALIGN_PTRS 1
-#if ALIGN_PTRS
- uchar *Buf = (uchar*)mmap(0, FileSize, PROT_READ|PROT_WRITE,
- MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
- assert((Buf != (uchar*)-1) && "mmap returned error!");
- memcpy(Buf, FileData, FileSize);
- free(FileData);
-#else
- uchar *Buf = FileData;
-#endif
-
- BytecodeParser Parser;
- Result = Parser.ParseBytecode(Buf, Buf+FileSize);
-
-#if ALIGN_PTRS
- munmap((char*)Buf, FileSize); // Free mmap'd data area
-#else
- free(FileData); // Free realloc'd block of memory
-#endif
-
- if (ErrorStr) *ErrorStr = Parser.getError();
+ // Read and check signature...
+ unsigned Sig;
+ if (read(Buf, EndBuf, Sig) ||
+ Sig != ('l' | ('l' << 8) | ('v' << 16) | ('m' << 24)))
+ throw std::string("Invalid bytecode signature!");
+
+ TheModule = new Module(ModuleID);
+ try {
+ ParseModule(Buf, EndBuf);
+ } catch (std::string &Error) {
+ freeState(); // Must destroy handles before deleting module!
+ delete TheModule;
+ TheModule = 0;
+ throw;
}
-
- return Result;
}