[AIX][TOC] Add -mtocdata/-mno-tocdata options on AIX (#67999)

This patch enables support that the XL compiler had for AIX under
-qdatalocal/-qdataimported.

Author: syzaara

clang/docs/UsersManual.rst

@@ -4227,7 +4227,68 @@ Clang expects the GCC executable "gcc.exe" compiled for
 
 AIX
 ^^^
+TOC Data Transformation
+"""""""""""""""""""""""
+TOC data transformation is off by default (``-mno-tocdata``).
+When ``-mtocdata`` is specified, the TOC data transformation will be applied to
+all suitable variables with static storage duration, including static data
+members of classes and block-scope static variables (if not marked as exceptions,
+see further below).
 
+Suitable variables must:
+
+-  have complete types
+-  be independently generated (i.e., not placed in a pool)
+-  be at most as large as a pointer
+-  not be aligned more strictly than a pointer
+-  not be structs containing flexible array members
+-  not have internal linkage
+-  not have aliases
+-  not have section attributes
+-  not be thread local storage
+
+The TOC data transformation results in the variable, not its address,
+being placed in the TOC. This eliminates the need to load the address of the
+variable from the TOC.
+
+Note:
+If the TOC data transformation is applied to a variable whose definition
+is imported, the linker will generate fixup code for reading or writing to the
+variable.
+
+When multiple toc-data options are used, the last option used has the affect.
+For example: -mno-tocdata=g5,g1 -mtocdata=g1,g2 -mno-tocdata=g2 -mtocdata=g3,g4
+results in -mtocdata=g1,g3,g4
+
+Names of variables not having external linkage will be ignored.
+
+**Options:**
+
+.. option:: -mno-tocdata
+
+  This is the default behaviour. Only variables explicitly specified with
+  ``-mtocdata=`` will have the TOC data transformation applied.
+
+.. option:: -mtocdata
+
+  Apply the TOC data transformation to all suitable variables with static
+  storage duration (including static data members of classes and block-scope
+  static variables) that are not explicitly specified with ``-mno-tocdata=``.
+
+.. option:: -mno-tocdata=
+
+  Can be used in conjunction with ``-mtocdata`` to mark the comma-separated
+  list of external linkage variables, specified using their mangled names, as
+  exceptions to ``-mtocdata``.
+
+.. option:: -mtocdata=
+
+  Apply the TOC data transformation to the comma-separated list of external
+  linkage variables, specified using their mangled names, if they are suitable.
+  Emit diagnostics for all unsuitable variables specified.
+
+Default Visibility Export Mapping
+"""""""""""""""""""""""""""""""""
 The ``-mdefault-visibility-export-mapping=`` option can be used to control
 mapping of default visibility to an explicit shared object export
 (i.e. XCOFF exported visibility). Three values are provided for the option:

clang/include/clang/Basic/CodeGenOptions.h

@@ -404,6 +404,15 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// List of pass builder callbacks.
   std::vector<std::function<void(llvm::PassBuilder &)>> PassBuilderCallbacks;
 
+  /// List of global variables explicitly specified by the user as toc-data.
+  std::vector<std::string> TocDataVarsUserSpecified;
+
+  /// List of global variables that over-ride the toc-data default.
+  std::vector<std::string> NoTocDataVars;
+
+  /// Flag for all global variables to be treated as toc-data.
+  bool AllTocData;
+
   /// Path to allowlist file specifying which objects
   /// (files, functions) should exclusively be instrumented
   /// by sanitizer coverage pass.

clang/include/clang/Basic/DiagnosticDriverKinds.td

@@ -587,6 +587,9 @@ def warn_drv_unsupported_gpopt : Warning<
   "ignoring '-mgpopt' option as it cannot be used with %select{|the implicit"
   " usage of }0-mabicalls">,
   InGroup<UnsupportedGPOpt>;
+def warn_drv_unsupported_tocdata: Warning<
+  "ignoring '-mtocdata' as it is only supported for -mcmodel=small">,
+  InGroup<OptionIgnored>;
 def warn_drv_unsupported_sdata : Warning<
   "ignoring '-msmall-data-limit=' with -mcmodel=large for -fpic or RV64">,
   InGroup<OptionIgnored>;

clang/include/clang/Basic/DiagnosticFrontendKinds.td

@@ -94,6 +94,8 @@ def err_fe_backend_error_attr :
 def warn_fe_backend_warning_attr :
   Warning<"call to '%0' declared with 'warning' attribute: %1">, BackendInfo,
   InGroup<BackendWarningAttributes>;
+def warn_toc_unsupported_type : Warning<"-mtocdata option is ignored "
+  "for %0 because %1">, InGroup<BackendWarningAttributes>;
 
 def err_fe_invalid_code_complete_file : Error<
     "cannot locate code-completion file %0">, DefaultFatal;

clang/include/clang/Driver/Options.td

@@ -3609,6 +3609,27 @@ def fpass_plugin_EQ : Joined<["-"], "fpass-plugin=">,
   MetaVarName<"<dsopath>">,
   HelpText<"Load pass plugin from a dynamic shared object file (only with new pass manager).">,
   MarshallingInfoStringVector<CodeGenOpts<"PassPlugins">>;
+defm tocdata : BoolOption<"m","tocdata",
+  CodeGenOpts<"AllTocData">, DefaultFalse,
+  PosFlag<SetTrue, [], [ClangOption, CC1Option],
+  "All suitable variables will have the TOC data transformation applied">,
+  NegFlag<SetFalse, [], [ClangOption, CC1Option],
+  "This is the default. TOC data transformation is not applied to any"
+  "variables. Only variables specified explicitly in -mtocdata= will"
+  "have the TOC data transformation.">,
+  BothFlags<[TargetSpecific], [ClangOption, CLOption]>>, Group<m_Group>;
+def mtocdata_EQ : CommaJoined<["-"], "mtocdata=">,
+  Visibility<[ClangOption, CC1Option]>,
+  Flags<[TargetSpecific]>, Group<m_Group>,
+  HelpText<"Specifies a list of variables to which the TOC data transformation"
+           "will be applied.">,
+  MarshallingInfoStringVector<CodeGenOpts<"TocDataVarsUserSpecified">>;
+def mno_tocdata_EQ : CommaJoined<["-"], "mno-tocdata=">,
+  Visibility<[ClangOption, CC1Option]>,
+  Flags<[TargetSpecific]>, Group<m_Group>,
+  HelpText<"Specifies a list of variables to be exempt from the TOC data"
+           "transformation.">,
+  MarshallingInfoStringVector<CodeGenOpts<"NoTocDataVars">>;
 defm preserve_as_comments : BoolFOption<"preserve-as-comments",
   CodeGenOpts<"PreserveAsmComments">, DefaultTrue,
   NegFlag<SetFalse, [], [ClangOption, CC1Option],

clang/lib/CodeGen/CGDecl.cpp

@@ -284,6 +284,7 @@ llvm::Constant *CodeGenModule::getOrCreateStaticVarDecl(
     setTLSMode(GV, D);
 
   setGVProperties(GV, &D);
+  getTargetCodeGenInfo().setTargetAttributes(cast<Decl>(&D), GV, *this);
 
   // Make sure the result is of the correct type.
   LangAS ExpectedAS = Ty.getAddressSpace();

clang/lib/CodeGen/CodeGenModule.cpp

@@ -626,6 +626,26 @@ static bool checkAliasedGlobal(
   return true;
 }
 
+// Emit a warning if toc-data attribute is requested for global variables that
+// have aliases and remove the toc-data attribute.
+static void checkAliasForTocData(llvm::GlobalVariable *GVar,
+                                 const CodeGenOptions &CodeGenOpts,
+                                 DiagnosticsEngine &Diags,
+                                 SourceLocation Location) {
+  if (GVar->hasAttribute("toc-data")) {
+    auto GVId = GVar->getName();
+    // Is this a global variable specified by the user as local?
+    if ((llvm::binary_search(CodeGenOpts.TocDataVarsUserSpecified, GVId))) {
+      Diags.Report(Location, diag::warn_toc_unsupported_type)
+          << GVId << "the variable has an alias";
+    }
+    llvm::AttributeSet CurrAttributes = GVar->getAttributes();
+    llvm::AttributeSet NewAttributes =
+        CurrAttributes.removeAttribute(GVar->getContext(), "toc-data");
+    GVar->setAttributes(NewAttributes);
+  }
+}
+
 void CodeGenModule::checkAliases() {
   // Check if the constructed aliases are well formed. It is really unfortunate
   // that we have to do this in CodeGen, but we only construct mangled names
@@ -652,6 +672,12 @@ void CodeGenModule::checkAliases() {
       continue;
     }
 
+    if (getContext().getTargetInfo().getTriple().isOSAIX())
+      if (const llvm::GlobalVariable *GVar =
+              dyn_cast<const llvm::GlobalVariable>(GV))
+        checkAliasForTocData(const_cast<llvm::GlobalVariable *>(GVar),
+                             getCodeGenOpts(), Diags, Location);
+
     llvm::Constant *Aliasee =
         IsIFunc ? cast<llvm::GlobalIFunc>(Alias)->getResolver()
                 : cast<llvm::GlobalAlias>(Alias)->getAliasee();

clang/lib/CodeGen/Targets/PPC.cpp

@@ -8,6 +8,7 @@
 
 #include "ABIInfoImpl.h"
 #include "TargetInfo.h"
+#include "clang/Basic/DiagnosticFrontend.h"
 
 using namespace clang;
 using namespace clang::CodeGen;
@@ -145,6 +146,9 @@ class AIXTargetCodeGenInfo : public TargetCodeGenInfo {
 
   bool initDwarfEHRegSizeTable(CodeGen::CodeGenFunction &CGF,
                                llvm::Value *Address) const override;
+
+  void setTargetAttributes(const Decl *D, llvm::GlobalValue *GV,
+                           CodeGen::CodeGenModule &M) const override;
 };
 } // namespace
 
@@ -265,6 +269,61 @@ bool AIXTargetCodeGenInfo::initDwarfEHRegSizeTable(
   return PPC_initDwarfEHRegSizeTable(CGF, Address, Is64Bit, /*IsAIX*/ true);
 }
 
+void AIXTargetCodeGenInfo::setTargetAttributes(
+    const Decl *D, llvm::GlobalValue *GV, CodeGen::CodeGenModule &M) const {
+  if (!isa<llvm::GlobalVariable>(GV))
+    return;
+
+  auto *GVar = dyn_cast<llvm::GlobalVariable>(GV);
+  auto GVId = GV->getName();
+
+  // Is this a global variable specified by the user as toc-data?
+  bool UserSpecifiedTOC =
+      llvm::binary_search(M.getCodeGenOpts().TocDataVarsUserSpecified, GVId);
+  // Assumes the same variable cannot be in both TocVarsUserSpecified and
+  // NoTocVars.
+  if (UserSpecifiedTOC ||
+      ((M.getCodeGenOpts().AllTocData) &&
+       !llvm::binary_search(M.getCodeGenOpts().NoTocDataVars, GVId))) {
+    const unsigned long PointerSize =
+        GV->getParent()->getDataLayout().getPointerSizeInBits() / 8;
+    auto *VarD = dyn_cast<VarDecl>(D);
+    assert(VarD && "Invalid declaration of global variable.");
+
+    ASTContext &Context = D->getASTContext();
+    unsigned Alignment = Context.toBits(Context.getDeclAlign(D)) / 8;
+    const auto *Ty = VarD->getType().getTypePtr();
+    const RecordDecl *RDecl =
+        Ty->isRecordType() ? Ty->getAs<RecordType>()->getDecl() : nullptr;
+
+    bool EmitDiagnostic = UserSpecifiedTOC && GV->hasExternalLinkage();
+    auto reportUnsupportedWarning = [&](bool ShouldEmitWarning, StringRef Msg) {
+      if (ShouldEmitWarning)
+        M.getDiags().Report(D->getLocation(), diag::warn_toc_unsupported_type)
+            << GVId << Msg;
+    };
+    if (!Ty || Ty->isIncompleteType())
+      reportUnsupportedWarning(EmitDiagnostic, "of incomplete type");
+    else if (RDecl && RDecl->hasFlexibleArrayMember())
+      reportUnsupportedWarning(EmitDiagnostic,
+                               "it contains a flexible array member");
+    else if (VarD->getTLSKind() != VarDecl::TLS_None)
+      reportUnsupportedWarning(EmitDiagnostic, "of thread local storage");
+    else if (PointerSize < Context.getTypeInfo(VarD->getType()).Width / 8)
+      reportUnsupportedWarning(EmitDiagnostic,
+                               "variable is larger than a pointer");
+    else if (PointerSize < Alignment)
+      reportUnsupportedWarning(EmitDiagnostic,
+                               "variable is aligned wider than a pointer");
+    else if (D->hasAttr<SectionAttr>())
+      reportUnsupportedWarning(EmitDiagnostic,
+                               "variable has a section attribute");
+    else if (GV->hasExternalLinkage() ||
+             (M.getCodeGenOpts().AllTocData && !GV->hasLocalLinkage()))
+      GVar->addAttribute("toc-data");
+  }
+}
+
 // PowerPC-32
 namespace {
 /// PPC32_SVR4_ABIInfo - The 32-bit PowerPC ELF (SVR4) ABI information.

clang/lib/Driver/ToolChains/AIX.cpp

@@ -433,13 +433,100 @@ void AIX::AddCXXStdlibLibArgs(const llvm::opt::ArgList &Args,
   llvm_unreachable("Unexpected C++ library type; only libc++ is supported.");
 }
 
+// This function processes all the mtocdata options to build the final
+// simplified toc data options to pass to CC1.
+static void addTocDataOptions(const llvm::opt::ArgList &Args,
+                              llvm::opt::ArgStringList &CC1Args,
+                              const Driver &D) {
+
+  // Check the global toc-data setting. The default is -mno-tocdata.
+  // To enable toc-data globally, -mtocdata must be specified.
+  // Additionally, it must be last to take effect.
+  const bool TOCDataGloballyinEffect = [&Args]() {
+    if (const Arg *LastArg =
+            Args.getLastArg(options::OPT_mtocdata, options::OPT_mno_tocdata))
+      return LastArg->getOption().matches(options::OPT_mtocdata);
+    else
+      return false;
+  }();
+
+  // Currently only supported for small code model.
+  if (TOCDataGloballyinEffect &&
+      (Args.getLastArgValue(options::OPT_mcmodel_EQ).equals("large") ||
+       Args.getLastArgValue(options::OPT_mcmodel_EQ).equals("medium"))) {
+    D.Diag(clang::diag::warn_drv_unsupported_tocdata);
+    return;
+  }
+
+  enum TOCDataSetting {
+    AddressInTOC = 0, // Address of the symbol stored in the TOC.
+    DataInTOC = 1     // Symbol defined in the TOC.
+  };
+
+  const TOCDataSetting DefaultTocDataSetting =
+      TOCDataGloballyinEffect ? DataInTOC : AddressInTOC;
+
+  // Process the list of variables in the explicitly specified options
+  // -mtocdata= and -mno-tocdata= to see which variables are opposite to
+  // the global setting of tocdata in TOCDataGloballyinEffect.
+  // Those that have the opposite setting to TOCDataGloballyinEffect, are added
+  // to ExplicitlySpecifiedGlobals.
+  llvm::StringSet<> ExplicitlySpecifiedGlobals;
+  for (const auto Arg :
+       Args.filtered(options::OPT_mtocdata_EQ, options::OPT_mno_tocdata_EQ)) {
+    TOCDataSetting ArgTocDataSetting =
+        Arg->getOption().matches(options::OPT_mtocdata_EQ) ? DataInTOC
+                                                           : AddressInTOC;
+
+    if (ArgTocDataSetting != DefaultTocDataSetting)
+      for (const char *Val : Arg->getValues())
+        ExplicitlySpecifiedGlobals.insert(Val);
+    else
+      for (const char *Val : Arg->getValues())
+        ExplicitlySpecifiedGlobals.erase(Val);
+  }
+
+  auto buildExceptionList = [](const llvm::StringSet<> &ExplicitValues,
+                               const char *OptionSpelling) {
+    std::string Option(OptionSpelling);
+    bool IsFirst = true;
+    for (const auto &E : ExplicitValues) {
+      if (!IsFirst)
+        Option += ",";
+
+      IsFirst = false;
+      Option += E.first();
+    }
+    return Option;
+  };
+
+  // Pass the final tocdata options to CC1 consisting of the default
+  // tocdata option (-mtocdata/-mno-tocdata) along with the list
+  // option (-mno-tocdata=/-mtocdata=) if there are any explicitly specified
+  // variables which would be exceptions to the default setting.
+  const char *TocDataGlobalOption =
+      TOCDataGloballyinEffect ? "-mtocdata" : "-mno-tocdata";
+  CC1Args.push_back(TocDataGlobalOption);
+
+  const char *TocDataListOption =
+      TOCDataGloballyinEffect ? "-mno-tocdata=" : "-mtocdata=";
+  if (!ExplicitlySpecifiedGlobals.empty())
+    CC1Args.push_back(Args.MakeArgString(llvm::Twine(
+        buildExceptionList(ExplicitlySpecifiedGlobals, TocDataListOption))));
+}
+
 void AIX::addClangTargetOptions(
     const llvm::opt::ArgList &Args, llvm::opt::ArgStringList &CC1Args,
     Action::OffloadKind DeviceOffloadingKind) const {
   Args.AddLastArg(CC1Args, options::OPT_mignore_xcoff_visibility);
   Args.AddLastArg(CC1Args, options::OPT_mdefault_visibility_export_mapping_EQ);
   Args.addOptInFlag(CC1Args, options::OPT_mxcoff_roptr, options::OPT_mno_xcoff_roptr);
 
+  // Forward last mtocdata/mno_tocdata options to -cc1.
+  if (Args.hasArg(options::OPT_mtocdata_EQ, options::OPT_mno_tocdata_EQ,
+                  options::OPT_mtocdata))
+    addTocDataOptions(Args, CC1Args, getDriver());
+
   if (Args.hasFlag(options::OPT_fxl_pragma_pack,
                    options::OPT_fno_xl_pragma_pack, true))
     CC1Args.push_back("-fxl-pragma-pack");

clang/lib/Frontend/CompilerInstance.cpp

@@ -1047,6 +1047,11 @@ bool CompilerInstance::ExecuteAction(FrontendAction &Act) {
   if (getFrontendOpts().ShowStats || !getFrontendOpts().StatsFile.empty())
     llvm::EnableStatistics(false);
 
+  // Sort vectors containing toc data and no toc data variables to facilitate
+  // binary search later.
+  llvm::sort(getCodeGenOpts().TocDataVarsUserSpecified);
+  llvm::sort(getCodeGenOpts().NoTocDataVars);
+
   for (const FrontendInputFile &FIF : getFrontendOpts().Inputs) {
     // Reset the ID tables if we are reusing the SourceManager and parsing
     // regular files.