+void LowerBitSets::lowerBitSetCalls(
+ ArrayRef<Metadata *> BitSets, Constant *CombinedGlobalAddr,
+ const DenseMap<GlobalObject *, uint64_t> &GlobalLayout) {
+ Constant *CombinedGlobalIntAddr =
+ ConstantExpr::getPtrToInt(CombinedGlobalAddr, IntPtrTy);
+
+ // For each bitset in this disjoint set...
+ for (Metadata *BS : BitSets) {
+ // Build the bitset.
+ BitSetInfo BSI = buildBitSet(BS, GlobalLayout);
+ DEBUG({
+ if (auto BSS = dyn_cast<MDString>(BS))
+ dbgs() << BSS->getString() << ": ";
+ else
+ dbgs() << "<unnamed>: ";
+ BSI.print(dbgs());
+ });
+
+ ByteArrayInfo *BAI = 0;
+
+ // Lower each call to llvm.bitset.test for this bitset.
+ for (CallInst *CI : BitSetTestCallSites[BS]) {
+ ++NumBitSetCallsLowered;
+ Value *Lowered =
+ lowerBitSetCall(CI, BSI, BAI, CombinedGlobalIntAddr, GlobalLayout);
+ CI->replaceAllUsesWith(Lowered);
+ CI->eraseFromParent();
+ }
+ }
+}
+
+void LowerBitSets::verifyBitSetMDNode(MDNode *Op) {
+ if (Op->getNumOperands() != 3)
+ report_fatal_error(
+ "All operands of llvm.bitsets metadata must have 3 elements");
+ if (!Op->getOperand(1))
+ return;
+
+ auto OpConstMD = dyn_cast<ConstantAsMetadata>(Op->getOperand(1));
+ if (!OpConstMD)
+ report_fatal_error("Bit set element must be a constant");
+ auto OpGlobal = dyn_cast<GlobalObject>(OpConstMD->getValue());
+ if (!OpGlobal)
+ return;
+
+ if (OpGlobal->isThreadLocal())
+ report_fatal_error("Bit set element may not be thread-local");
+ if (OpGlobal->hasSection())
+ report_fatal_error("Bit set element may not have an explicit section");
+
+ if (isa<GlobalVariable>(OpGlobal) && OpGlobal->isDeclarationForLinker())
+ report_fatal_error("Bit set global var element must be a definition");
+
+ auto OffsetConstMD = dyn_cast<ConstantAsMetadata>(Op->getOperand(2));
+ if (!OffsetConstMD)
+ report_fatal_error("Bit set element offset must be a constant");
+ auto OffsetInt = dyn_cast<ConstantInt>(OffsetConstMD->getValue());
+ if (!OffsetInt)
+ report_fatal_error("Bit set element offset must be an integer constant");
+}
+
+static const unsigned kX86JumpTableEntrySize = 8;
+
+unsigned LowerBitSets::getJumpTableEntrySize() {
+ if (Arch != Triple::x86 && Arch != Triple::x86_64)
+ report_fatal_error("Unsupported architecture for jump tables");
+
+ return kX86JumpTableEntrySize;
+}
+
+// Create a constant representing a jump table entry for the target. This
+// consists of an instruction sequence containing a relative branch to Dest. The
+// constant will be laid out at address Src+(Len*Distance) where Len is the
+// target-specific jump table entry size.
+Constant *LowerBitSets::createJumpTableEntry(GlobalObject *Src, Function *Dest,
+ unsigned Distance) {
+ if (Arch != Triple::x86 && Arch != Triple::x86_64)
+ report_fatal_error("Unsupported architecture for jump tables");
+
+ const unsigned kJmpPCRel32Code = 0xe9;
+ const unsigned kInt3Code = 0xcc;
+
+ ConstantInt *Jmp = ConstantInt::get(Int8Ty, kJmpPCRel32Code);
+
+ // Build a constant representing the displacement between the constant's
+ // address and Dest. This will resolve to a PC32 relocation referring to Dest.
+ Constant *DestInt = ConstantExpr::getPtrToInt(Dest, IntPtrTy);
+ Constant *SrcInt = ConstantExpr::getPtrToInt(Src, IntPtrTy);
+ Constant *Disp = ConstantExpr::getSub(DestInt, SrcInt);
+ ConstantInt *DispOffset =
+ ConstantInt::get(IntPtrTy, Distance * kX86JumpTableEntrySize + 5);
+ Constant *OffsetedDisp = ConstantExpr::getSub(Disp, DispOffset);
+ OffsetedDisp = ConstantExpr::getTrunc(OffsetedDisp, Int32Ty);
+
+ ConstantInt *Int3 = ConstantInt::get(Int8Ty, kInt3Code);
+
+ Constant *Fields[] = {
+ Jmp, OffsetedDisp, Int3, Int3, Int3,
+ };
+ return ConstantStruct::getAnon(Fields, /*Packed=*/true);
+}
+
+Type *LowerBitSets::getJumpTableEntryType() {
+ if (Arch != Triple::x86 && Arch != Triple::x86_64)
+ report_fatal_error("Unsupported architecture for jump tables");
+
+ return StructType::get(M->getContext(),
+ {Int8Ty, Int32Ty, Int8Ty, Int8Ty, Int8Ty},
+ /*Packed=*/true);
+}
+
+/// Given a disjoint set of bitsets and functions, build a jump table for the
+/// functions, build the bit sets and lower the llvm.bitset.test calls.
+void LowerBitSets::buildBitSetsFromFunctions(ArrayRef<Metadata *> BitSets,
+ ArrayRef<Function *> Functions) {
+ // Unlike the global bitset builder, the function bitset builder cannot
+ // re-arrange functions in a particular order and base its calculations on the
+ // layout of the functions' entry points, as we have no idea how large a
+ // particular function will end up being (the size could even depend on what
+ // this pass does!) Instead, we build a jump table, which is a block of code
+ // consisting of one branch instruction for each of the functions in the bit
+ // set that branches to the target function, and redirect any taken function
+ // addresses to the corresponding jump table entry. In the object file's
+ // symbol table, the symbols for the target functions also refer to the jump
+ // table entries, so that addresses taken outside the module will pass any
+ // verification done inside the module.
+ //
+ // In more concrete terms, suppose we have three functions f, g, h which are
+ // members of a single bitset, and a function foo that returns their
+ // addresses:
+ //
+ // f:
+ // mov 0, %eax
+ // ret
+ //
+ // g:
+ // mov 1, %eax
+ // ret
+ //
+ // h:
+ // mov 2, %eax
+ // ret
+ //
+ // foo:
+ // mov f, %eax
+ // mov g, %edx
+ // mov h, %ecx
+ // ret
+ //
+ // To create a jump table for these functions, we instruct the LLVM code
+ // generator to output a jump table in the .text section. This is done by
+ // representing the instructions in the jump table as an LLVM constant and
+ // placing them in a global variable in the .text section. The end result will
+ // (conceptually) look like this:
+ //
+ // f:
+ // jmp .Ltmp0 ; 5 bytes
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ //
+ // g:
+ // jmp .Ltmp1 ; 5 bytes
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ //
+ // h:
+ // jmp .Ltmp2 ; 5 bytes
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ // int3 ; 1 byte
+ //
+ // .Ltmp0:
+ // mov 0, %eax
+ // ret
+ //
+ // .Ltmp1:
+ // mov 1, %eax
+ // ret
+ //
+ // .Ltmp2:
+ // mov 2, %eax
+ // ret
+ //
+ // foo:
+ // mov f, %eax
+ // mov g, %edx
+ // mov h, %ecx
+ // ret
+ //
+ // Because the addresses of f, g, h are evenly spaced at a power of 2, in the
+ // normal case the check can be carried out using the same kind of simple
+ // arithmetic that we normally use for globals.
+
+ assert(!Functions.empty());
+
+ // Build a simple layout based on the regular layout of jump tables.
+ DenseMap<GlobalObject *, uint64_t> GlobalLayout;
+ unsigned EntrySize = getJumpTableEntrySize();
+ for (unsigned I = 0; I != Functions.size(); ++I)
+ GlobalLayout[Functions[I]] = I * EntrySize;
+
+ // Create a constant to hold the jump table.
+ ArrayType *JumpTableType =
+ ArrayType::get(getJumpTableEntryType(), Functions.size());
+ auto JumpTable = new GlobalVariable(*M, JumpTableType,
+ /*isConstant=*/true,
+ GlobalValue::PrivateLinkage, nullptr);
+ JumpTable->setSection(ObjectFormat == Triple::MachO
+ ? "__TEXT,__text,regular,pure_instructions"
+ : ".text");
+ lowerBitSetCalls(BitSets, JumpTable, GlobalLayout);
+
+ // Build aliases pointing to offsets into the jump table, and replace
+ // references to the original functions with references to the aliases.
+ for (unsigned I = 0; I != Functions.size(); ++I) {
+ Constant *CombinedGlobalElemPtr = ConstantExpr::getBitCast(
+ ConstantExpr::getGetElementPtr(
+ JumpTableType, JumpTable,
+ ArrayRef<Constant *>{ConstantInt::get(IntPtrTy, 0),
+ ConstantInt::get(IntPtrTy, I)}),
+ Functions[I]->getType());
+ if (LinkerSubsectionsViaSymbols || Functions[I]->isDeclarationForLinker()) {
+ Functions[I]->replaceAllUsesWith(CombinedGlobalElemPtr);
+ } else {
+ GlobalAlias *GAlias = GlobalAlias::create(Functions[I]->getType(),
+ Functions[I]->getLinkage(), "",
+ CombinedGlobalElemPtr, M);
+ GAlias->setVisibility(Functions[I]->getVisibility());
+ GAlias->takeName(Functions[I]);
+ Functions[I]->replaceAllUsesWith(GAlias);
+ }
+ if (!Functions[I]->isDeclarationForLinker())
+ Functions[I]->setLinkage(GlobalValue::PrivateLinkage);
+ }
+
+ // Build and set the jump table's initializer.
+ std::vector<Constant *> JumpTableEntries;
+ for (unsigned I = 0; I != Functions.size(); ++I)
+ JumpTableEntries.push_back(
+ createJumpTableEntry(JumpTable, Functions[I], I));
+ JumpTable->setInitializer(
+ ConstantArray::get(JumpTableType, JumpTableEntries));
+}
+
+void LowerBitSets::buildBitSetsFromDisjointSet(
+ ArrayRef<Metadata *> BitSets, ArrayRef<GlobalObject *> Globals) {
+ llvm::DenseMap<Metadata *, uint64_t> BitSetIndices;
+ llvm::DenseMap<GlobalObject *, uint64_t> GlobalIndices;
+ for (auto B : BitSets)
+ BitSetIndices[B] = BitSetIndices.size();
+ for (auto G : Globals)
+ GlobalIndices[G] = GlobalIndices.size();
+
+ // For each bitset, build a set of indices that refer to globals referenced by
+ // the bitset.
+ std::vector<std::set<uint64_t>> BitSetMembers(BitSets.size());
+ if (BitSetNM) {
+ for (MDNode *Op : BitSetNM->operands()) {
+ // Op = { bitset name, global, offset }
+ if (!Op->getOperand(1))
+ continue;
+ auto I = BitSetIndices.find(Op->getOperand(0));
+ if (I == BitSetIndices.end())
+ continue;
+
+ auto OpGlobal = dyn_cast<GlobalObject>(
+ cast<ConstantAsMetadata>(Op->getOperand(1))->getValue());
+ if (!OpGlobal)
+ continue;
+ BitSetMembers[I->second].insert(GlobalIndices[OpGlobal]);
+ }
+ }
+
+ // Order the sets of indices by size. The GlobalLayoutBuilder works best
+ // when given small index sets first.
+ std::stable_sort(
+ BitSetMembers.begin(), BitSetMembers.end(),
+ [](const std::set<uint64_t> &O1, const std::set<uint64_t> &O2) {
+ return O1.size() < O2.size();
+ });
+
+ // Create a GlobalLayoutBuilder and provide it with index sets as layout
+ // fragments. The GlobalLayoutBuilder tries to lay out members of fragments as
+ // close together as possible.
+ GlobalLayoutBuilder GLB(Globals.size());
+ for (auto &&MemSet : BitSetMembers)
+ GLB.addFragment(MemSet);
+
+ // Build the bitsets from this disjoint set.
+ if (Globals.empty() || isa<GlobalVariable>(Globals[0])) {
+ // Build a vector of global variables with the computed layout.
+ std::vector<GlobalVariable *> OrderedGVs(Globals.size());
+ auto OGI = OrderedGVs.begin();
+ for (auto &&F : GLB.Fragments) {
+ for (auto &&Offset : F) {
+ auto GV = dyn_cast<GlobalVariable>(Globals[Offset]);
+ if (!GV)
+ report_fatal_error(
+ "Bit set may not contain both global variables and functions");
+ *OGI++ = GV;
+ }
+ }
+
+ buildBitSetsFromGlobalVariables(BitSets, OrderedGVs);
+ } else {
+ // Build a vector of functions with the computed layout.
+ std::vector<Function *> OrderedFns(Globals.size());
+ auto OFI = OrderedFns.begin();
+ for (auto &&F : GLB.Fragments) {
+ for (auto &&Offset : F) {
+ auto Fn = dyn_cast<Function>(Globals[Offset]);
+ if (!Fn)
+ report_fatal_error(
+ "Bit set may not contain both global variables and functions");
+ *OFI++ = Fn;
+ }
+ }
+
+ buildBitSetsFromFunctions(BitSets, OrderedFns);
+ }
+}
+
projects / oota-llvm.git / commitdiff