[WebAssembly] Make Emscripten-specific behavior specific to the Emscripten target

[sunfish]

Summary

Make some behaviors added for Emscripten compatibility conditional on the target being Emscripten.

• Add an WASM_SYMBOL_IMPLICITLY_USED flag, so that attribute((used)) doesn't need to imply exporting. When targeting Emscripten, have WASM_SYMBOL_IMPLICITLY_USED imply exporting.
• Limit PIC support to the Emscripten target, since the current PIC support is Emscripten-specific.

[sbc100]

include/llvm/BinaryFormat/Wasm.h

316 ↗

(On Diff #201734)

Why "implicitly"? It seems like it a very explicit attribute :) How about just "WASM_SYMBOL_NO_STRIP" to match MCSA_NoDeadStrip and the mach-o implementation which is .no_dead_strip at the asm level.

[sbc100]

lib/MC/MCWasmStreamer.cpp

125 ↗

(On Diff #201734)

It looks like this is the only caller of setExported so you can remove setExported and isExported and probably also remove WASM_SYMBOL_EXPORTED?

In which case why not just rename WASM_SYMBOL_EXPORTED rather than adding another flag? Or can you imagine the EXPORTED flag getting set in some other way?

[dschuff]

LGTM in general. Maybe IMPLICITLY_EXPORTED would be an even better name?
Anyway, I'm hoping we can make that export behavior nicer soon; I find the attribute(used) -> export behavior a bit odd too. Once we drop fastcomp it will be easier to redefine EMSCRIPTEN_KEEPALIVE and other things.

[sunfish]

lib/MC/MCWasmStreamer.cpp

125 ↗

(On Diff #201734)

I can remove setExported and isExported. But we appear to still need WASM_SYMBOL_EXPORTED -- there's also the code in lib/MC/WasmObjectWriter.cpp which adds wasm::WASM_SYMBOL_EXPORTED to symbols with NoStrip in Emscripten mode, for compatibility with Emscripten, going by this comment.

[sunfish]

• Rename IMPLICITLY_USED to NO_STRIP.
• Remove isExported and setExported accessors.

[sbc100]

I was hoping you could have the linker default to exporting NO_STRIP symbols? If you don't like that idea then perhaps would have the "export NO_STRIP symbols" behaviour be emscripten-only in the linker and have it driven by the same signal we end up using for -shared/-pie.

So then we can just rename the existing EXPORTED to NO_STRIP to avoid adding another flag here?

Do do you think we will end up wanted a separate EXPORTED flag for symbol in any case? For example with WASI, will we will presumably want some way to export symbols based on source attributes? Should we add wasm-export-name attribute like we have for wasm-import-name?

[sunfish]

In D62542#1529441, @sbc100 wrote:

I was hoping you could have the linker default to exporting NO_STRIP symbols? If you don't like that idea then perhaps would have the "export NO_STRIP symbols" behaviour be emscripten-only in the linker and have it driven by the same signal we end up using for -shared/-pie.

It doesn't have that meaning in ELF targets, for example.

Do do you think we will end up wanted a separate EXPORTED flag for symbol in any case? For example with WASI, will we will presumably want some way to export symbols based on source attributes?

I don't know yet. There still are some questions about whether we want to have a separate layer of visibility between ESM's and C++ libraries.

Should we add wasm-export-name attribute like we have for wasm-import-name?

Yes, I expect we'll want this, though I myself don't have anything that urgently needs it.

[sunfish]

The PIC portion of this patch is now split off as r362638. This patch now just contains the __attribute__((used))/no_dead_strip changes.

[sunfish]

https://reviews.llvm.org/D66968 is now a patch implementing the lld side of this.

llvm/trunk/include/llvm/BinaryFormat/Wasm.h

 const unsigned WASM_SYMBOL_BINDING_GLOBAL = 0x0;
 const unsigned WASM_SYMBOL_BINDING_WEAK = 0x1;
 const unsigned WASM_SYMBOL_BINDING_LOCAL = 0x2;
 const unsigned WASM_SYMBOL_VISIBILITY_DEFAULT = 0x0;
 const unsigned WASM_SYMBOL_VISIBILITY_HIDDEN = 0x4;
 const unsigned WASM_SYMBOL_UNDEFINED = 0x10;
 const unsigned WASM_SYMBOL_EXPORTED = 0x20;
 const unsigned WASM_SYMBOL_EXPLICIT_NAME = 0x40;
+const unsigned WASM_SYMBOL_NO_STRIP = 0x80;
 
 #define WASM_RELOC(name, value) name = value,
 
 enum : unsigned {
 #include "WasmRelocs.def"
 };
 
 #undef WASM_RELOC

llvm/trunk/include/llvm/MC/MCSymbolWasm.h

public:
 
   bool isExported() const {
     return getFlags() & wasm::WASM_SYMBOL_EXPORTED;
   }
   void setExported() const {
     modifyFlags(wasm::WASM_SYMBOL_EXPORTED, wasm::WASM_SYMBOL_EXPORTED);
   }
 
+  bool isNoStrip() const {
+    return getFlags() & wasm::WASM_SYMBOL_NO_STRIP;
+  }
+  void setNoStrip() const {
+    modifyFlags(wasm::WASM_SYMBOL_NO_STRIP, wasm::WASM_SYMBOL_NO_STRIP);
+  }
+
   bool isWeak() const { return IsWeak; }
   void setWeak(bool isWeak) { IsWeak = isWeak; }
 
   bool isHidden() const { return IsHidden; }
   void setHidden(bool isHidden) { IsHidden = isHidden; }
 
   bool isComdat() const { return IsComdat; }
   void setComdat(bool isComdat) { IsComdat = isComdat; }

llvm/trunk/include/llvm/MC/MCWasmObjectWriter.h

 namespace llvm {
 
 class MCFixup;
 class MCValue;
 class raw_pwrite_stream;
 
 class MCWasmObjectTargetWriter : public MCObjectTargetWriter {
   const unsigned Is64Bit : 1;
+  const unsigned IsEmscripten : 1;
 
 protected:
-  explicit MCWasmObjectTargetWriter(bool Is64Bit_);
+  explicit MCWasmObjectTargetWriter(bool Is64Bit_, bool IsEmscripten);
 
 public:
   virtual ~MCWasmObjectTargetWriter();
 
   virtual Triple::ObjectFormatType getFormat() const { return Triple::Wasm; }
   static bool classof(const MCObjectTargetWriter *W) {
     return W->getFormat() == Triple::Wasm;
   }
 
   virtual unsigned getRelocType(const MCValue &Target,
                                 const MCFixup &Fixup) const = 0;
 
   /// \name Accessors
   /// @{
   bool is64Bit() const { return Is64Bit; }
+  bool isEmscripten() const { return IsEmscripten; }
   /// @}
 };
 
 /// Construct a new Wasm writer instance.
 ///
 /// \param MOTW - The target specific Wasm writer subclass.
 /// \param OS - The stream to write to.
 /// \returns The constructed object writer.
 std::unique_ptr<MCObjectWriter>
 createWasmObjectWriter(std::unique_ptr<MCWasmObjectTargetWriter> MOTW,
                        raw_pwrite_stream &OS);
 
 } // namespace llvm
 
 #endif

llvm/trunk/lib/MC/MCWasmObjectTargetWriter.cpp

 //===-- MCWasmObjectTargetWriter.cpp - Wasm Target Writer Subclass --------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCWasmObjectWriter.h"
 
 using namespace llvm;
 
-MCWasmObjectTargetWriter::MCWasmObjectTargetWriter(bool Is64Bit)
-    : Is64Bit(Is64Bit) {}
+MCWasmObjectTargetWriter::MCWasmObjectTargetWriter(bool Is64Bit,
+                                                   bool IsEmscripten)
+    : Is64Bit(Is64Bit), IsEmscripten(IsEmscripten) {}
 
 // Pin the vtable to this object file
 MCWasmObjectTargetWriter::~MCWasmObjectTargetWriter() = default;

llvm/trunk/lib/MC/MCWasmStreamer.cpp

case MCSA_ELF_TypeFunction:
     Symbol->setType(wasm::WASM_SYMBOL_TYPE_FUNCTION);
     break;
 
   case MCSA_ELF_TypeObject:
   case MCSA_Cold:
     break;
 
   case MCSA_NoDeadStrip:
-    Symbol->setExported();
+    Symbol->setNoStrip();
     break;
 
   default:
     // unrecognized directive
     llvm_unreachable("unexpected MCSymbolAttr");
     return false;
   }
 

llvm/trunk/lib/MC/WasmObjectWriter.cpp

class WasmObjectWriter : public MCObjectWriter {
   SmallVector<WasmDataSegment, 4> DataSegments;
   unsigned NumFunctionImports = 0;
   unsigned NumGlobalImports = 0;
   unsigned NumEventImports = 0;
   uint32_t SectionCount = 0;
 
   // TargetObjectWriter wrappers.
   bool is64Bit() const { return TargetObjectWriter->is64Bit(); }
+  bool isEmscripten() const { return TargetObjectWriter->isEmscripten(); }
 
   void startSection(SectionBookkeeping &Section, unsigned SectionId);
   void startCustomSection(SectionBookkeeping &Section, StringRef Name);
   void endSection(SectionBookkeeping &Section);
 
 public:
   WasmObjectWriter(std::unique_ptr<MCWasmObjectTargetWriter> MOTW,
                    raw_pwrite_stream &OS)
for (const MCSymbol &S : Asm.symbols()) {
     if (WS.isWeak())
       Flags |= wasm::WASM_SYMBOL_BINDING_WEAK;
     if (WS.isHidden())
       Flags |= wasm::WASM_SYMBOL_VISIBILITY_HIDDEN;
     if (!WS.isExternal() && WS.isDefined())
       Flags |= wasm::WASM_SYMBOL_BINDING_LOCAL;
     if (WS.isUndefined())
       Flags |= wasm::WASM_SYMBOL_UNDEFINED;
-    if (WS.isExported())
+    if (WS.isNoStrip()) {
+      Flags |= wasm::WASM_SYMBOL_NO_STRIP;
+      if (isEmscripten()) {
         Flags |= wasm::WASM_SYMBOL_EXPORTED;
+      }
+    }
     if (WS.getName() != WS.getImportName())
       Flags |= wasm::WASM_SYMBOL_EXPLICIT_NAME;
 
     wasm::WasmSymbolInfo Info;
     Info.Name = WS.getName();
     Info.Kind = WS.getType();
     Info.Flags = Flags;
     if (!WS.isData()) {

llvm/trunk/lib/ObjectYAML/WasmYAML.cpp

#define BCaseMask(M, X)                                                        \
   // BCaseMask(BINDING_MASK, BINDING_GLOBAL);
   BCaseMask(BINDING_MASK, BINDING_WEAK);
   BCaseMask(BINDING_MASK, BINDING_LOCAL);
   // BCaseMask(VISIBILITY_MASK, VISIBILITY_DEFAULT);
   BCaseMask(VISIBILITY_MASK, VISIBILITY_HIDDEN);
   BCaseMask(UNDEFINED, UNDEFINED);
   BCaseMask(EXPORTED, EXPORTED);
   BCaseMask(EXPLICIT_NAME, EXPLICIT_NAME);
+  BCaseMask(NO_STRIP, NO_STRIP);
 #undef BCaseMask
 }
 
 void ScalarEnumerationTraits<WasmYAML::SymbolKind>::enumeration(
     IO &IO, WasmYAML::SymbolKind &Kind) {
 #define ECase(X) IO.enumCase(Kind, #X, wasm::WASM_SYMBOL_TYPE_##X);
   ECase(FUNCTION);
   ECase(DATA);

llvm/trunk/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyAsmBackend.cpp

 #include "llvm/Support/raw_ostream.h"
 
 using namespace llvm;
 
 namespace {
 
 class WebAssemblyAsmBackend final : public MCAsmBackend {
   bool Is64Bit;
+  bool IsEmscripten;
 
 public:
-  explicit WebAssemblyAsmBackend(bool Is64Bit)
-      : MCAsmBackend(support::little), Is64Bit(Is64Bit) {}
+  explicit WebAssemblyAsmBackend(bool Is64Bit, bool IsEmscripten)
+      : MCAsmBackend(support::little), Is64Bit(Is64Bit),
+        IsEmscripten(IsEmscripten) {}
 
   unsigned getNumFixupKinds() const override {
     return WebAssembly::NumTargetFixupKinds;
   }
 
   const MCFixupKindInfo &getFixupKindInfo(MCFixupKind Kind) const override;
 
   void applyFixup(const MCAssembler &Asm, const MCFixup &Fixup,
void WebAssemblyAsmBackend::applyFixup(const MCAssembler &Asm,
   // For each byte of the fragment that the fixup touches, mask in the
   // bits from the fixup value.
   for (unsigned I = 0; I != NumBytes; ++I)
     Data[Offset + I] |= uint8_t((Value >> (I * 8)) & 0xff);
 }
 
 std::unique_ptr<MCObjectTargetWriter>
 WebAssemblyAsmBackend::createObjectTargetWriter() const {
-  return createWebAssemblyWasmObjectWriter(Is64Bit);
+  return createWebAssemblyWasmObjectWriter(Is64Bit, IsEmscripten);
 }
 
 } // end anonymous namespace
 
 MCAsmBackend *llvm::createWebAssemblyAsmBackend(const Triple &TT) {
-  return new WebAssemblyAsmBackend(TT.isArch64Bit());
+  return new WebAssemblyAsmBackend(TT.isArch64Bit(), TT.isOSEmscripten());
 }

llvm/trunk/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyMCTargetDesc.h

 class Triple;
 class raw_pwrite_stream;
 
 MCCodeEmitter *createWebAssemblyMCCodeEmitter(const MCInstrInfo &MCII);
 
 MCAsmBackend *createWebAssemblyAsmBackend(const Triple &TT);
 
 std::unique_ptr<MCObjectTargetWriter>
-createWebAssemblyWasmObjectWriter(bool Is64Bit);
+createWebAssemblyWasmObjectWriter(bool Is64Bit, bool IsEmscripten);
 
 namespace WebAssembly {
 enum OperandType {
   /// Basic block label in a branch construct.
   OPERAND_BASIC_BLOCK = MCOI::OPERAND_FIRST_TARGET,
   /// Local index.
   OPERAND_LOCAL,
   /// Global index.

llvm/trunk/lib/Target/WebAssembly/MCTargetDesc/WebAssemblyWasmObjectWriter.cpp

 #include "llvm/Support/Casting.h"
 #include "llvm/Support/ErrorHandling.h"
 
 using namespace llvm;
 
 namespace {
 class WebAssemblyWasmObjectWriter final : public MCWasmObjectTargetWriter {
 public:
-  explicit WebAssemblyWasmObjectWriter(bool Is64Bit);
+  explicit WebAssemblyWasmObjectWriter(bool Is64Bit, bool IsEmscripten);
 
 private:
   unsigned getRelocType(const MCValue &Target,
                         const MCFixup &Fixup) const override;
 };
 } // end anonymous namespace
 
-WebAssemblyWasmObjectWriter::WebAssemblyWasmObjectWriter(bool Is64Bit)
-    : MCWasmObjectTargetWriter(Is64Bit) {}
+WebAssemblyWasmObjectWriter::WebAssemblyWasmObjectWriter(bool Is64Bit,
+                                                         bool IsEmscripten)
+    : MCWasmObjectTargetWriter(Is64Bit, IsEmscripten) {}
 
 static const MCSection *getFixupSection(const MCExpr *Expr) {
   if (auto SyExp = dyn_cast<MCSymbolRefExpr>(Expr)) {
     if (SyExp->getSymbol().isInSection())
       return &SyExp->getSymbol().getSection();
     return nullptr;
   }
 
case FK_Data_4:
     }
     return wasm::R_WASM_MEMORY_ADDR_I32;
   default:
     llvm_unreachable("unimplemented fixup kind");
   }
 }
 
 std::unique_ptr<MCObjectTargetWriter>
-llvm::createWebAssemblyWasmObjectWriter(bool Is64Bit) {
-  return std::make_unique<WebAssemblyWasmObjectWriter>(Is64Bit);
+llvm::createWebAssemblyWasmObjectWriter(bool Is64Bit, bool IsEmscripten) {
+  return std::make_unique<WebAssemblyWasmObjectWriter>(Is64Bit, IsEmscripten);
 }

llvm/trunk/test/MC/WebAssembly/no-dead-strip.ll

 ; RUN: llc -filetype=obj -wasm-keep-registers %s -o - | llvm-readobj --symbols | FileCheck %s
 
 target triple = "wasm32-unknown-unknown"
 
 @llvm.used = appending global [1 x i8*] [i8* bitcast (i32 ()* @foo to i8*)], section "llvm.metadata"
 
 define i32 @foo() {
 entry:
     ret i32 0
 }
 
 ; CHECK:      Symbols [
 ; CHECK-NEXT:   Symbol {
 ; CHECK-NEXT:     Name: foo
 ; CHECK-NEXT:     Type: FUNCTION (0x0)
-; CHECK-NEXT:     Flags [ (0x20)
-; CHECK-NEXT:       EXPORTED (0x20)
+; CHECK-NEXT:     Flags [ (0x80)
+; CHECK-NEXT:       NO_STRIP (0x80)
 ; CHECK-NEXT:     ]
 ; CHECK-NEXT:     ElementIndex: 0x0
 ; CHECK-NEXT:   }
 ; CHECK-NEXT: ]

llvm/trunk/tools/llvm-readobj/WasmDumper.cpp

#define ENUM_ENTRY(X)                                                          \
   ENUM_ENTRY(BINDING_GLOBAL),
   ENUM_ENTRY(BINDING_WEAK),
   ENUM_ENTRY(BINDING_LOCAL),
   ENUM_ENTRY(VISIBILITY_DEFAULT),
   ENUM_ENTRY(VISIBILITY_HIDDEN),
   ENUM_ENTRY(UNDEFINED),
   ENUM_ENTRY(EXPORTED),
   ENUM_ENTRY(EXPLICIT_NAME),
+  ENUM_ENTRY(NO_STRIP),
 #undef ENUM_ENTRY
 };
 
 class WasmDumper : public ObjDumper {
 public:
   WasmDumper(const WasmObjectFile *Obj, ScopedPrinter &Writer)
       : ObjDumper(Writer), Obj(Obj) {}