From: Dan Gohman Date: Sat, 7 Jan 2017 00:34:54 +0000 (+0000) Subject: [WebAssembly] Add a pass to create wrappers for function bitcasts. X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=91f4652a9e4d8c3dc419c3485e22939720cc684d;p=llvm [WebAssembly] Add a pass to create wrappers for function bitcasts. WebAssembly requires caller and callee signatures to match exactly. In LLVM, there are a variety of circumstances where signatures may be mismatched in practice, and one can bitcast a function address to another type to call it as that type. This patch adds a pass which replaces bitcasted function addresses with wrappers to replace the bitcasts. This doesn't catch everything, but it does match many common cases. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@291315 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Target/WebAssembly/CMakeLists.txt b/lib/Target/WebAssembly/CMakeLists.txt index f4d46383e5b..d9c53ecc8d0 100644 --- a/lib/Target/WebAssembly/CMakeLists.txt +++ b/lib/Target/WebAssembly/CMakeLists.txt @@ -17,6 +17,7 @@ add_llvm_target(WebAssemblyCodeGen WebAssemblyExplicitLocals.cpp WebAssemblyFastISel.cpp WebAssemblyFixIrreducibleControlFlow.cpp + WebAssemblyFixFunctionBitcasts.cpp WebAssemblyFrameLowering.cpp WebAssemblyISelDAGToDAG.cpp WebAssemblyISelLowering.cpp diff --git a/lib/Target/WebAssembly/WebAssembly.h b/lib/Target/WebAssembly/WebAssembly.h index 09c35b4825f..8738263ad84 100644 --- a/lib/Target/WebAssembly/WebAssembly.h +++ b/lib/Target/WebAssembly/WebAssembly.h @@ -28,6 +28,7 @@ class FunctionPass; // LLVM IR passes. ModulePass *createWebAssemblyLowerEmscriptenEHSjLj(bool DoEH, bool DoSjLj); void initializeWebAssemblyLowerEmscriptenEHSjLjPass(PassRegistry &); +ModulePass *createWebAssemblyFixFunctionBitcasts(); FunctionPass *createWebAssemblyOptimizeReturned(); // ISel and immediate followup passes. diff --git a/lib/Target/WebAssembly/WebAssemblyFixFunctionBitcasts.cpp b/lib/Target/WebAssembly/WebAssemblyFixFunctionBitcasts.cpp new file mode 100644 index 00000000000..dfad5b8facd --- /dev/null +++ b/lib/Target/WebAssembly/WebAssemblyFixFunctionBitcasts.cpp @@ -0,0 +1,148 @@ +//===-- WebAssemblyFixFunctionBitcasts.cpp - Fix function bitcasts --------===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +/// +/// \file +/// \brief Fix bitcasted functions. +/// +/// WebAssembly requires caller and callee signatures to match, however in LLVM, +/// some amount of slop is vaguely permitted. Detect mismatch by looking for +/// bitcasts of functions and rewrite them to use wrapper functions instead. +/// +/// This doesn't catch all cases, such as when a function's address is taken in +/// one place and casted in another, but it works for many common cases. +/// +/// Note that LLVM already optimizes away function bitcasts in common cases by +/// dropping arguments as needed, so this pass only ends up getting used in less +/// common cases. +/// +//===----------------------------------------------------------------------===// + +#include "WebAssembly.h" +#include "llvm/IR/Constants.h" +#include "llvm/IR/Instructions.h" +#include "llvm/IR/Module.h" +#include "llvm/IR/Operator.h" +#include "llvm/Pass.h" +#include "llvm/Support/Debug.h" +#include "llvm/Support/raw_ostream.h" +using namespace llvm; + +#define DEBUG_TYPE "wasm-fix-function-bitcasts" + +namespace { +class FixFunctionBitcasts final : public ModulePass { + StringRef getPassName() const override { + return "WebAssembly Fix Function Bitcasts"; + } + + void getAnalysisUsage(AnalysisUsage &AU) const override { + AU.setPreservesCFG(); + ModulePass::getAnalysisUsage(AU); + } + + bool runOnModule(Module &M) override; + + SmallVector, 0> Uses; + +public: + static char ID; + FixFunctionBitcasts() : ModulePass(ID) {} +}; +} // End anonymous namespace + +char FixFunctionBitcasts::ID = 0; +ModulePass *llvm::createWebAssemblyFixFunctionBitcasts() { + return new FixFunctionBitcasts(); +} + +// Recursively descend the def-use lists from V to find non-bitcast users of +// bitcasts of V. +static void FindUses(Value *V, Function &F, + SmallVectorImpl> &Uses) { + for (Use &U : V->uses()) { + if (BitCastOperator *BC = dyn_cast(U.getUser())) + FindUses(BC, F, Uses); + else if (U.get()->getType() != F.getType()) + Uses.push_back(std::make_pair(&U, &F)); + } +} + +// Create a wrapper function with type Ty that calls F (which may have a +// different type). Attempt to support common bitcasted function idioms: +// - Call with more arguments than needed: arguments are dropped +// - Call with fewer arguments than needed: arguments are filled in with undef +// - Return value is not needed: drop it +// - Return value needed but not present: supply an undef +static Function *CreateWrapper(Function *F, FunctionType *Ty) { + Module *M = F->getParent(); + + Function *Wrapper = + Function::Create(Ty, Function::PrivateLinkage, "bitcast", M); + BasicBlock *BB = BasicBlock::Create(M->getContext(), "body", Wrapper); + + // Determine what arguments to pass. + SmallVector Args; + Function::arg_iterator AI = Wrapper->arg_begin(); + FunctionType::param_iterator PI = F->getFunctionType()->param_begin(); + FunctionType::param_iterator PE = F->getFunctionType()->param_end(); + for (; AI != Wrapper->arg_end() && PI != PE; ++AI, ++PI) { + assert(AI->getType() == *PI && + "mismatched argument types not supported yet"); + Args.push_back(&*AI); + } + for (; PI != PE; ++PI) + Args.push_back(UndefValue::get(*PI)); + + CallInst *Call = CallInst::Create(F, Args, "", BB); + + // Determine what value to return. + if (Ty->getReturnType()->isVoidTy()) + ReturnInst::Create(M->getContext(), BB); + else if (F->getFunctionType()->getReturnType()->isVoidTy()) + ReturnInst::Create(M->getContext(), UndefValue::get(Ty->getReturnType()), + BB); + else if (F->getFunctionType()->getReturnType() == Ty->getReturnType()) + ReturnInst::Create(M->getContext(), Call, BB); + else + llvm_unreachable("mismatched return types not supported yet"); + + return Wrapper; +} + +bool FixFunctionBitcasts::runOnModule(Module &M) { + // Collect all the places that need wrappers. + for (Function &F : M) + FindUses(&F, F, Uses); + + DenseMap, Function *> Wrappers; + + for (auto &UseFunc : Uses) { + Use *U = UseFunc.first; + Function *F = UseFunc.second; + PointerType *PTy = cast(U->get()->getType()); + FunctionType *Ty = dyn_cast(PTy->getElementType()); + + // If the function is casted to something like i8* as a "generic pointer" + // to be later casted to something else, we can't generate a wrapper for it. + // Just ignore such casts for now. + if (!Ty) + continue; + + auto Pair = Wrappers.insert(std::make_pair(std::make_pair(F, Ty), nullptr)); + if (Pair.second) + Pair.first->second = CreateWrapper(F, Ty); + + if (isa(U->get())) + U->get()->replaceAllUsesWith(Pair.first->second); + else + U->set(Pair.first->second); + } + + return true; +} diff --git a/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp b/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp index b61bc0a0814..f5ef35a2ad4 100644 --- a/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp +++ b/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp @@ -163,6 +163,10 @@ void WebAssemblyPassConfig::addIRPasses() { // control specifically what gets lowered. addPass(createAtomicExpandPass(TM)); + // Fix function bitcasts, as WebAssembly requires caller and callee signatures + // to match. + addPass(createWebAssemblyFixFunctionBitcasts()); + // Optimize "returned" function attributes. if (getOptLevel() != CodeGenOpt::None) addPass(createWebAssemblyOptimizeReturned()); diff --git a/test/CodeGen/WebAssembly/function-bitcasts.ll b/test/CodeGen/WebAssembly/function-bitcasts.ll new file mode 100644 index 00000000000..b3953c4f100 --- /dev/null +++ b/test/CodeGen/WebAssembly/function-bitcasts.ll @@ -0,0 +1,56 @@ +; RUN: llc < %s -asm-verbose=false | FileCheck %s + +; Test that function pointer casts are replaced with wrappers. + +target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128" +target triple = "wasm32-unknown-unknown" + +; CHECK-LABEL: test: +; CHECK-NEXT: call .Lbitcast@FUNCTION{{$}} +; CHECK-NEXT: call .Lbitcast.1@FUNCTION{{$}} +; CHECK-NEXT: i32.const $push[[L0:[0-9]*]]=, 0 +; CHECK-NEXT: call .Lbitcast.2@FUNCTION, $pop[[L0]]{{$}} +; CHECK-NEXT: i32.call $drop=, .Lbitcast.3@FUNCTION{{$}} +; CHECK-NEXT: call foo2@FUNCTION{{$}} +; CHECK-NEXT: call foo3@FUNCTION{{$}} +; CHECK-NEXT: .endfunc + +; CHECK-LABEL: .Lbitcast: +; CHECK-NEXT: .local i32 +; CHECK-NEXT: call has_i32_arg@FUNCTION, $0{{$}} +; CHECK-NEXT: .endfunc + +; CHECK-LABEL: .Lbitcast.1: +; CHECK-NEXT: call $drop=, has_i32_ret@FUNCTION{{$}} +; CHECK-NEXT: .endfunc + +; CHECK-LABEL: .Lbitcast.2: +; CHECK-NEXT: .param i32 +; CHECK-NEXT: call foo0@FUNCTION{{$}} +; CHECK-NEXT: .endfunc + +; CHECK-LABEL: .Lbitcast.3: +; CHECK-NEXT: .result i32 +; CHECK-NEXT: .local i32 +; CHECK-NEXT: call foo1@FUNCTION{{$}} +; CHECK-NEXT: copy_local $push0=, $0 +; CHECK-NEXT: .endfunc + +declare void @has_i32_arg(i32) +declare i32 @has_i32_ret() + +declare void @foo0() +declare void @foo1() +declare void @foo2() +declare void @foo3() + +define void @test() { +entry: + call void bitcast (void (i32)* @has_i32_arg to void ()*)() + call void bitcast (i32 ()* @has_i32_ret to void ()*)() + call void bitcast (void ()* @foo0 to void (i32)*)(i32 0) + %t = call i32 bitcast (void ()* @foo1 to i32 ()*)() + call void bitcast (void ()* @foo2 to void ()*)() + call void @foo3() + ret void +}