[clang][BPF] support type exist/size and enum exist/value relocations

This patch added the following additional compile-once
run-everywhere (CO-RE) relocations:
  - existence/size of typedef, struct/union or enum type
  - enum value and enum value existence

These additional relocations will make CO-RE bpf programs more
adaptive for potential kernel internal data structure changes.

For existence/size relocations, the following two code patterns
are supported:
  1. uint32_t __builtin_preserve_type_info(*(<type> *)0, flag);
  2. <type> var;
     uint32_t __builtin_preserve_field_info(var, flag);
flag = 0 for existence relocation and flag = 1 for size relocation.

For enum value existence and enum value relocations, the following code
pattern is supported:
  uint64_t __builtin_preserve_enum_value(*(<enum_type> *)<enum_value>,
                                         flag);
flag = 0 means existence relocation and flag = 1 for enum value.
relocation. In the above <enum_type> can be an enum type or
a typedef to enum type. The <enum_value> needs to be an enumerator
value from the same enum type. The return type is uint64_t to
permit potential 64bit enumerator values.

Differential Revision: https://reviews.llvm.org/D83242

clang/include/clang/Basic/BuiltinsBPF.def

@@ -23,5 +23,11 @@ TARGET_BUILTIN(__builtin_preserve_field_info, "Ui.", "t", "")
 // Get BTF type id.
 TARGET_BUILTIN(__builtin_btf_type_id, "Ui.", "t", "")
 
+// Get type information.
+TARGET_BUILTIN(__builtin_preserve_type_info, "Ui.", "t", "")
+
+// Preserve enum value.
+TARGET_BUILTIN(__builtin_preserve_enum_value, "Li.", "t", "")
+
 #undef BUILTIN
 #undef TARGET_BUILTIN

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -10865,6 +10865,14 @@ def err_preserve_field_info_not_const: Error<
   "__builtin_preserve_field_info argument %0 not a constant">;
 def err_btf_type_id_not_const: Error<
   "__builtin_btf_type_id argument %0 not a constant">;
+def err_preserve_type_info_invalid : Error<
+  "__builtin_preserve_type_info argument %0 invalid">;
+def err_preserve_type_info_not_const: Error<
+  "__builtin_preserve_type_info argument %0 not a constant">;
+def err_preserve_enum_value_invalid : Error<
+  "__builtin_preserve_enum_value argument %0 invalid">;
+def err_preserve_enum_value_not_const: Error<
+  "__builtin_preserve_enum_value argument %0 not a constant">;
 
 def err_bit_cast_non_trivially_copyable : Error<
   "__builtin_bit_cast %select{source|destination}0 type must be trivially copyable">;

clang/lib/CodeGen/CGBuiltin.cpp

@@ -10921,9 +10921,16 @@ Value *CodeGenFunction::EmitAArch64BuiltinExpr(unsigned BuiltinID,
 Value *CodeGenFunction::EmitBPFBuiltinExpr(unsigned BuiltinID,
                                            const CallExpr *E) {
   assert((BuiltinID == BPF::BI__builtin_preserve_field_info ||
-          BuiltinID == BPF::BI__builtin_btf_type_id) &&
+          BuiltinID == BPF::BI__builtin_btf_type_id ||
+          BuiltinID == BPF::BI__builtin_preserve_type_info ||
+          BuiltinID == BPF::BI__builtin_preserve_enum_value) &&
          "unexpected BPF builtin");
 
+  // A sequence number, injected into IR builtin functions, to
+  // prevent CSE given the only difference of the funciton
+  // may just be the debuginfo metadata.
+  static uint32_t BuiltinSeqNum;
+
   switch (BuiltinID) {
   default:
     llvm_unreachable("Unexpected BPF builtin");
@@ -11016,6 +11023,63 @@ Value *CodeGenFunction::EmitBPFBuiltinExpr(unsigned BuiltinID,
     Fn->setMetadata(LLVMContext::MD_preserve_access_index, DbgInfo);
     return Fn;
   }
+  case BPF::BI__builtin_preserve_type_info: {
+    if (!getDebugInfo()) {
+      CGM.Error(E->getExprLoc(), "using builtin function without -g");
+      return nullptr;
+    }
+
+    const Expr *Arg0 = E->getArg(0);
+    llvm::DIType *DbgInfo = getDebugInfo()->getOrCreateStandaloneType(
+        Arg0->getType(), Arg0->getExprLoc());
+
+    ConstantInt *Flag = cast<ConstantInt>(EmitScalarExpr(E->getArg(1)));
+    Value *FlagValue = ConstantInt::get(Int64Ty, Flag->getSExtValue());
+    Value *SeqNumVal = ConstantInt::get(Int32Ty, BuiltinSeqNum++);
+
+    llvm::Function *FnPreserveTypeInfo = llvm::Intrinsic::getDeclaration(
+        &CGM.getModule(), llvm::Intrinsic::bpf_preserve_type_info, {});
+    CallInst *Fn =
+        Builder.CreateCall(FnPreserveTypeInfo, {SeqNumVal, FlagValue});
+    Fn->setMetadata(LLVMContext::MD_preserve_access_index, DbgInfo);
+    return Fn;
+  }
+  case BPF::BI__builtin_preserve_enum_value: {
+    if (!getDebugInfo()) {
+      CGM.Error(E->getExprLoc(), "using builtin function without -g");
+      return nullptr;
+    }
+
+    const Expr *Arg0 = E->getArg(0);
+    llvm::DIType *DbgInfo = getDebugInfo()->getOrCreateStandaloneType(
+        Arg0->getType(), Arg0->getExprLoc());
+
+    // Find enumerator
+    const auto *UO = cast<UnaryOperator>(Arg0->IgnoreParens());
+    const auto *CE = cast<CStyleCastExpr>(UO->getSubExpr());
+    const auto *DR = cast<DeclRefExpr>(CE->getSubExpr());
+    const auto *Enumerator = cast<EnumConstantDecl>(DR->getDecl());
+
+    auto &InitVal = Enumerator->getInitVal();
+    std::string InitValStr;
+    if (InitVal.isNegative() || InitVal > uint64_t(INT64_MAX))
+      InitValStr = std::to_string(InitVal.getSExtValue());
+    else
+      InitValStr = std::to_string(InitVal.getZExtValue());
+    std::string EnumStr = Enumerator->getNameAsString() + ":" + InitValStr;
+    Value *EnumStrVal = Builder.CreateGlobalStringPtr(EnumStr);
+
+    ConstantInt *Flag = cast<ConstantInt>(EmitScalarExpr(E->getArg(1)));
+    Value *FlagValue = ConstantInt::get(Int64Ty, Flag->getSExtValue());
+    Value *SeqNumVal = ConstantInt::get(Int32Ty, BuiltinSeqNum++);
+
+    llvm::Function *IntrinsicFn = llvm::Intrinsic::getDeclaration(
+        &CGM.getModule(), llvm::Intrinsic::bpf_preserve_enum_value, {});
+    CallInst *Fn =
+        Builder.CreateCall(IntrinsicFn, {SeqNumVal, EnumStrVal, FlagValue});
+    Fn->setMetadata(LLVMContext::MD_preserve_access_index, DbgInfo);
+    return Fn;
+  }
   }
 }
 

clang/lib/Sema/SemaChecking.cpp

@@ -2557,52 +2557,151 @@ bool Sema::CheckAArch64BuiltinFunctionCall(const TargetInfo &TI,
   return SemaBuiltinConstantArgRange(TheCall, i, l, u + l);
 }
 
+static bool isValidBPFPreserveFieldInfoArg(Expr *Arg) {
+  if (Arg->getType()->getAsPlaceholderType())
+    return false;
+
+  // The first argument needs to be a record field access.
+  // If it is an array element access, we delay decision
+  // to BPF backend to check whether the access is a
+  // field access or not.
+  return (Arg->IgnoreParens()->getObjectKind() == OK_BitField ||
+          dyn_cast<MemberExpr>(Arg->IgnoreParens()) ||
+          dyn_cast<ArraySubscriptExpr>(Arg->IgnoreParens()));
+}
+
+static bool isValidBPFPreserveTypeInfoArg(Expr *Arg) {
+  QualType ArgType = Arg->getType();
+  if (ArgType->getAsPlaceholderType())
+    return false;
+
+  // for TYPE_EXISTENCE/TYPE_SIZEOF reloc type
+  // format:
+  //   1. __builtin_preserve_type_info(*(<type> *)0, flag);
+  //   2. <type> var;
+  //      __builtin_preserve_type_info(var, flag);
+  if (!dyn_cast<DeclRefExpr>(Arg->IgnoreParens()) &&
+      !dyn_cast<UnaryOperator>(Arg->IgnoreParens()))
+    return false;
+
+  // Typedef type.
+  if (ArgType->getAs<TypedefType>())
+    return true;
+
+  // Record type or Enum type.
+  const Type *Ty = ArgType->getUnqualifiedDesugaredType();
+  if (const auto *RT = Ty->getAs<RecordType>()) {
+    if (!RT->getDecl()->getDeclName().isEmpty())
+      return true;
+  } else if (const auto *ET = Ty->getAs<EnumType>()) {
+    if (!ET->getDecl()->getDeclName().isEmpty())
+      return true;
+  }
+
+  return false;
+}
+
+static bool isValidBPFPreserveEnumValueArg(Expr *Arg) {
+  QualType ArgType = Arg->getType();
+  if (ArgType->getAsPlaceholderType())
+    return false;
+
+  // for ENUM_VALUE_EXISTENCE/ENUM_VALUE reloc type
+  // format:
+  //   __builtin_preserve_enum_value(*(<enum_type> *)<enum_value>,
+  //                                 flag);
+  const auto *UO = dyn_cast<UnaryOperator>(Arg->IgnoreParens());
+  if (!UO)
+    return false;
+
+  const auto *CE = dyn_cast<CStyleCastExpr>(UO->getSubExpr());
+  if (!CE || CE->getCastKind() != CK_IntegralToPointer)
+    return false;
+
+  // The integer must be from an EnumConstantDecl.
+  const auto *DR = dyn_cast<DeclRefExpr>(CE->getSubExpr());
+  if (!DR)
+    return false;
+
+  const EnumConstantDecl *Enumerator =
+      dyn_cast<EnumConstantDecl>(DR->getDecl());
+  if (!Enumerator)
+    return false;
+
+  // The type must be EnumType.
+  const Type *Ty = ArgType->getUnqualifiedDesugaredType();
+  const auto *ET = Ty->getAs<EnumType>();
+  if (!ET)
+    return false;
+
+  // The enum value must be supported.
+  for (auto *EDI : ET->getDecl()->enumerators()) {
+    if (EDI == Enumerator)
+      return true;
+  }
+
+  return false;
+}
+
 bool Sema::CheckBPFBuiltinFunctionCall(unsigned BuiltinID,
                                        CallExpr *TheCall) {
   assert((BuiltinID == BPF::BI__builtin_preserve_field_info ||
-          BuiltinID == BPF::BI__builtin_btf_type_id) &&
-         "unexpected ARM builtin");
+          BuiltinID == BPF::BI__builtin_btf_type_id ||
+          BuiltinID == BPF::BI__builtin_preserve_type_info ||
+          BuiltinID == BPF::BI__builtin_preserve_enum_value) &&
+         "unexpected BPF builtin");
 
   if (checkArgCount(*this, TheCall, 2))
     return true;
 
-  Expr *Arg;
-  if (BuiltinID == BPF::BI__builtin_btf_type_id) {
-    // The second argument needs to be a constant int
-    Arg = TheCall->getArg(1);
-    if (!Arg->isIntegerConstantExpr(Context)) {
-      Diag(Arg->getBeginLoc(), diag::err_btf_type_id_not_const)
-          << 2 << Arg->getSourceRange();
-      return true;
-    }
-
-    TheCall->setType(Context.UnsignedIntTy);
-    return false;
+  // The second argument needs to be a constant int
+  Expr *Arg = TheCall->getArg(1);
+  Optional<llvm::APSInt> Value = Arg->getIntegerConstantExpr(Context);
+  diag::kind kind;
+  if (!Value) {
+    if (BuiltinID == BPF::BI__builtin_preserve_field_info)
+      kind = diag::err_preserve_field_info_not_const;
+    else if (BuiltinID == BPF::BI__builtin_btf_type_id)
+      kind = diag::err_btf_type_id_not_const;
+    else if (BuiltinID == BPF::BI__builtin_preserve_type_info)
+      kind = diag::err_preserve_type_info_not_const;
+    else
+      kind = diag::err_preserve_enum_value_not_const;
+    Diag(Arg->getBeginLoc(), kind) << 2 << Arg->getSourceRange();
+    return true;
   }
 
-  // The first argument needs to be a record field access.
-  // If it is an array element access, we delay decision
-  // to BPF backend to check whether the access is a
-  // field access or not.
+  // The first argument
   Arg = TheCall->getArg(0);
-  if (Arg->getType()->getAsPlaceholderType() ||
-      (Arg->IgnoreParens()->getObjectKind() != OK_BitField &&
-       !dyn_cast<MemberExpr>(Arg->IgnoreParens()) &&
-       !dyn_cast<ArraySubscriptExpr>(Arg->IgnoreParens()))) {
-    Diag(Arg->getBeginLoc(), diag::err_preserve_field_info_not_field)
-        << 1 << Arg->getSourceRange();
-    return true;
+  bool InvalidArg = false;
+  bool ReturnUnsignedInt = true;
+  if (BuiltinID == BPF::BI__builtin_preserve_field_info) {
+    if (!isValidBPFPreserveFieldInfoArg(Arg)) {
+      InvalidArg = true;
+      kind = diag::err_preserve_field_info_not_field;
+    }
+  } else if (BuiltinID == BPF::BI__builtin_preserve_type_info) {
+    if (!isValidBPFPreserveTypeInfoArg(Arg)) {
+      InvalidArg = true;
+      kind = diag::err_preserve_type_info_invalid;
+    }
+  } else if (BuiltinID == BPF::BI__builtin_preserve_enum_value) {
+    if (!isValidBPFPreserveEnumValueArg(Arg)) {
+      InvalidArg = true;
+      kind = diag::err_preserve_enum_value_invalid;
+    }
+    ReturnUnsignedInt = false;
   }
 
-  // The second argument needs to be a constant int
-  Arg = TheCall->getArg(1);
-  if (!Arg->isIntegerConstantExpr(Context)) {
-    Diag(Arg->getBeginLoc(), diag::err_preserve_field_info_not_const)
-        << 2 << Arg->getSourceRange();
+  if (InvalidArg) {
+    Diag(Arg->getBeginLoc(), kind) << 1 << Arg->getSourceRange();
     return true;
   }
 
-  TheCall->setType(Context.UnsignedIntTy);
+  if (ReturnUnsignedInt)
+    TheCall->setType(Context.UnsignedIntTy);
+  else
+    TheCall->setType(Context.UnsignedLongTy);
   return false;
 }
 

clang/test/CodeGen/builtins-bpf-preserve-field-info-3.c

@@ -0,0 +1,41 @@
+// REQUIRES: bpf-registered-target
+// RUN: %clang -target bpf -emit-llvm -S -g %s -o - | FileCheck %s
+
+#define _(x, y) (__builtin_preserve_type_info((x), (y)))
+
+struct s {
+  char a;
+};
+typedef int __int;
+enum AA {
+  VAL1 = 1,
+  VAL2 = 2,
+};
+
+unsigned unit1() {
+  struct s v = {};
+  return _(v, 0) + _(*(struct s *)0, 0);
+}
+
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 0, i64 0), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[STRUCT_S:[0-9]+]]
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 1, i64 0), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[STRUCT_S]]
+
+unsigned unit2() {
+  __int n;
+  return _(n, 1) + _(*(__int *)0, 1);
+}
+
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 2, i64 1), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[TYPEDEF_INT:[0-9]+]]
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 3, i64 1), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[TYPEDEF_INT]]
+
+unsigned unit3() {
+  enum AA t;
+  return _(t, 0) + _(*(enum AA *)0, 1);
+}
+
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 4, i64 0), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[ENUM_AA:[0-9]+]]
+// CHECK: call i32 @llvm.bpf.preserve.type.info(i32 5, i64 1), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[ENUM_AA]]
+
+// CHECK: ![[ENUM_AA]] = !DICompositeType(tag: DW_TAG_enumeration_type, name: "AA"
+// CHECK: ![[TYPEDEF_INT]] = !DIDerivedType(tag: DW_TAG_typedef, name: "__int"
+// CHECK: ![[STRUCT_S]] = distinct !DICompositeType(tag: DW_TAG_structure_type, name: "s"

clang/test/CodeGen/builtins-bpf-preserve-field-info-4.c

@@ -0,0 +1,32 @@
+// REQUIRES: bpf-registered-target
+// RUN: %clang -target bpf -emit-llvm -S -g %s -o - | FileCheck %s
+
+#define _(x, y) (__builtin_preserve_enum_value((x), (y)))
+
+enum AA {
+  VAL1 = 2,
+  VAL2 = 0xffffffff80000000UL,
+};
+typedef enum { VAL10 = -2, VAL11 = 0xffff8000, }  __BB;
+
+unsigned unit1() {
+  return _(*(enum AA *)VAL1, 0) + _(*(__BB *)VAL10, 1);
+}
+
+unsigned unit2() {
+  return _(*(enum AA *)VAL2, 0) + _(*(__BB *)VAL11, 1);
+}
+
+// CHECK: @0 = private unnamed_addr constant [7 x i8] c"VAL1:2\00", align 1
+// CHECK: @1 = private unnamed_addr constant [9 x i8] c"VAL10:-2\00", align 1
+// CHECK: @2 = private unnamed_addr constant [17 x i8] c"VAL2:-2147483648\00", align 1
+// CHECK: @3 = private unnamed_addr constant [17 x i8] c"VAL11:4294934528\00", align 1
+
+// CHECK: call i64 @llvm.bpf.preserve.enum.value(i32 0, i8* getelementptr inbounds ([7 x i8], [7 x i8]* @0, i32 0, i32 0), i64 0), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[ENUM_AA:[0-9]+]]
+// CHECK: call i64 @llvm.bpf.preserve.enum.value(i32 1, i8* getelementptr inbounds ([9 x i8], [9 x i8]* @1, i32 0, i32 0), i64 1), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[TYPEDEF_ENUM:[0-9]+]]
+
+// CHECK: call i64 @llvm.bpf.preserve.enum.value(i32 2, i8* getelementptr inbounds ([17 x i8], [17 x i8]* @2, i32 0, i32 0), i64 0), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[ENUM_AA]]
+// CHECK: call i64 @llvm.bpf.preserve.enum.value(i32 3, i8* getelementptr inbounds ([17 x i8], [17 x i8]* @3, i32 0, i32 0), i64 1), !dbg !{{[0-9]+}}, !llvm.preserve.access.index ![[TYPEDEF_ENUM]]
+
+// CHECK: ![[ENUM_AA]] = !DICompositeType(tag: DW_TAG_enumeration_type, name: "AA"
+// CHECK: ![[TYPEDEF_ENUM]] = !DIDerivedType(tag: DW_TAG_typedef, name: "__BB"