[llvm] Extract and propagate callee_type metadata

llvm/include/llvm/CodeGen/MachineFunction.h

@@ -26,11 +26,14 @@
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/MachineInstr.h"
 #include "llvm/CodeGen/MachineMemOperand.h"
+#include "llvm/IR/Constants.h"
 #include "llvm/IR/EHPersonalities.h"
+#include "llvm/IR/Instructions.h"
 #include "llvm/Support/Allocator.h"
 #include "llvm/Support/ArrayRecycler.h"
 #include "llvm/Support/AtomicOrdering.h"
 #include "llvm/Support/Compiler.h"
+#include "llvm/Support/MD5.h"
 #include "llvm/Support/Recycler.h"
 #include "llvm/Target/TargetOptions.h"
 #include <bitset>
@@ -517,6 +520,32 @@ class LLVM_ABI MachineFunction {
     SmallVector<ArgRegPair, 1> ArgRegPairs;
     /// Callee type ids.
     SmallVector<ConstantInt *, 4> CalleeTypeIds;
+
+    CallSiteInfo() = default;
+
+    /// Extracts the numeric type id from the CallBase's callee_type Metadata,
+    /// and sets CalleeTypeIds. This is used as type id for the indirect call in
+    /// the call graph section.
+    CallSiteInfo(const CallBase &CB) {
+      // Call graph section needs numeric callee_type id only for indirect
+      // calls.
+      if (!CB.isIndirectCall())
+        return;
+
+      MDNode *CalleeTypeList = CB.getMetadata(LLVMContext::MD_callee_type);
+      if (!CalleeTypeList)
+        return;
+
+      for (const MDOperand &Op : CalleeTypeList->operands()) {
+        MDNode *TypeMD = cast<MDNode>(Op);
+        MDString *TypeIdStr = cast<MDString>(TypeMD->getOperand(1));
+        // Compute numeric type id from generalized type id string
+        uint64_t TypeIdVal = MD5Hash(TypeIdStr->getString());
+        IntegerType *Int64Ty = Type::getInt64Ty(CB.getContext());
+        CalleeTypeIds.push_back(
+            ConstantInt::get(Int64Ty, TypeIdVal, /*IsSigned=*/false));
+      }
+    }
   };
 
   struct CalledGlobalInfo {

llvm/lib/CodeGen/SelectionDAG/ScheduleDAGSDNodes.cpp

@@ -888,7 +888,8 @@ EmitSchedule(MachineBasicBlock::iterator &InsertPos) {
     }
 
     if (MI->isCandidateForAdditionalCallInfo()) {
-      if (DAG->getTarget().Options.EmitCallSiteInfo)
+      if (DAG->getTarget().Options.EmitCallSiteInfo ||
+          DAG->getTarget().Options.EmitCallGraphSection)
         MF.addCallSiteInfo(MI, DAG->getCallSiteInfo(Node));
 
       if (auto CalledGlobal = DAG->getCalledGlobal(Node))

llvm/lib/Target/AArch64/AArch64ISelLowering.cpp

@@ -8952,6 +8952,7 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
   bool &IsTailCall = CLI.IsTailCall;
   CallingConv::ID &CallConv = CLI.CallConv;
   bool IsVarArg = CLI.IsVarArg;
+  const CallBase *CB = CLI.CB;
 
   MachineFunction &MF = DAG.getMachineFunction();
   MachineFunction::CallSiteInfo CSInfo;
@@ -8991,6 +8992,10 @@ AArch64TargetLowering::LowerCall(CallLoweringInfo &CLI,
                     *DAG.getContext());
   RetCCInfo.AnalyzeCallResult(Ins, RetCC);
 
+  // Set type id for call site info.
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    CSInfo = MachineFunction::CallSiteInfo(*CB);
+
   // Check callee args/returns for SVE registers and set calling convention
   // accordingly.
   if (CallConv == CallingConv::C || CallConv == CallingConv::Fast) {

llvm/lib/Target/ARM/ARMISelLowering.cpp

@@ -2423,6 +2423,7 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   CallingConv::ID CallConv              = CLI.CallConv;
   bool doesNotRet                       = CLI.DoesNotReturn;
   bool isVarArg                         = CLI.IsVarArg;
+  const CallBase *CB = CLI.CB;
 
   MachineFunction &MF = DAG.getMachineFunction();
   ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
@@ -2446,6 +2447,10 @@ ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
       !Subtarget->noBTIAtReturnTwice())
     GuardWithBTI = AFI->branchTargetEnforcement();
 
+  // Set type id for call site info.
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    CSInfo = MachineFunction::CallSiteInfo(*CB);
+
   // Determine whether this is a non-secure function call.
   if (CLI.CB && CLI.CB->getAttributes().hasFnAttr("cmse_nonsecure_call"))
     isCmseNSCall = true;

llvm/lib/Target/Mips/MipsISelLowering.cpp

@@ -3341,6 +3341,7 @@ MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   bool &IsTailCall                      = CLI.IsTailCall;
   CallingConv::ID CallConv              = CLI.CallConv;
   bool IsVarArg                         = CLI.IsVarArg;
+  const CallBase *CB = CLI.CB;
 
   MachineFunction &MF = DAG.getMachineFunction();
   MachineFrameInfo &MFI = MF.getFrameInfo();
@@ -3397,8 +3398,11 @@ MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   // Get a count of how many bytes are to be pushed on the stack.
   unsigned StackSize = CCInfo.getStackSize();
 
-  // Call site info for function parameters tracking.
+  // Call site info for function parameters tracking and call base type info.
   MachineFunction::CallSiteInfo CSInfo;
+  // Set type id for call site info.
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    CSInfo = MachineFunction::CallSiteInfo(*CB);
 
   // Check if it's really possible to do a tail call. Restrict it to functions
   // that are part of this compilation unit.

llvm/lib/Target/RISCV/RISCVISelLowering.cpp

@@ -22636,8 +22636,14 @@ SDValue RISCVTargetLowering::LowerCall(CallLoweringInfo &CLI,
   bool IsVarArg = CLI.IsVarArg;
   EVT PtrVT = getPointerTy(DAG.getDataLayout());
   MVT XLenVT = Subtarget.getXLenVT();
+  const CallBase *CB = CLI.CB;
 
   MachineFunction &MF = DAG.getMachineFunction();
+  MachineFunction::CallSiteInfo CSInfo;
+
+  // Set type id for call site info.
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    CSInfo = MachineFunction::CallSiteInfo(*CB);
 
   // Analyze the operands of the call, assigning locations to each operand.
   SmallVector<CCValAssign, 16> ArgLocs;
@@ -22895,13 +22901,20 @@ SDValue RISCVTargetLowering::LowerCall(CallLoweringInfo &CLI,
     if (CLI.CFIType)
       Ret.getNode()->setCFIType(CLI.CFIType->getZExtValue());
     DAG.addNoMergeSiteInfo(Ret.getNode(), CLI.NoMerge);
+    if (MF.getTarget().Options.EmitCallGraphSection && CB &&
+        CB->isIndirectCall())
+      DAG.addCallSiteInfo(Ret.getNode(), std::move(CSInfo));
     return Ret;
   }
 
   unsigned CallOpc = NeedSWGuarded ? RISCVISD::SW_GUARDED_CALL : RISCVISD::CALL;
   Chain = DAG.getNode(CallOpc, DL, NodeTys, Ops);
   if (CLI.CFIType)
     Chain.getNode()->setCFIType(CLI.CFIType->getZExtValue());
+
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    DAG.addCallSiteInfo(Chain.getNode(), std::move(CSInfo));
+
   DAG.addNoMergeSiteInfo(Chain.getNode(), CLI.NoMerge);
   Glue = Chain.getValue(1);
 

llvm/lib/Target/X86/X86FastISel.cpp

@@ -3673,6 +3673,12 @@ bool X86FastISel::fastLowerCall(CallLoweringInfo &CLI) {
   CLI.NumResultRegs = RVLocs.size();
   CLI.Call = MIB;
 
+  // Add call site info for call graph section.
+  if (TM.Options.EmitCallGraphSection && CB && CB->isIndirectCall()) {
+    MachineFunction::CallSiteInfo CSInfo(*CB);
+    MF->addCallSiteInfo(CLI.Call, std::move(CSInfo));
+  }
+
   return true;
 }
 
@@ -4042,6 +4048,8 @@ bool X86FastISel::tryToFoldLoadIntoMI(MachineInstr *MI, unsigned OpNo,
     MO.setReg(IndexReg);
   }
 
+  if (MI->isCall())
+    FuncInfo.MF->moveAdditionalCallInfo(MI, Result);
   Result->addMemOperand(*FuncInfo.MF, createMachineMemOperandFor(LI));
   Result->cloneInstrSymbols(*FuncInfo.MF, *MI);
   MachineBasicBlock::iterator I(MI);

llvm/lib/Target/X86/X86ISelLoweringCall.cpp

@@ -2060,6 +2060,10 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   if (CallConv == CallingConv::X86_INTR)
     report_fatal_error("X86 interrupts may not be called directly");
 
+  // Set type id for call site info.
+  if (MF.getTarget().Options.EmitCallGraphSection && CB && CB->isIndirectCall())
+    CSInfo = MachineFunction::CallSiteInfo(*CB);
+
   if (IsIndirectCall && !IsWin64 &&
       M->getModuleFlag("import-call-optimization"))
     errorUnsupported(DAG, dl,

llvm/test/CodeGen/AArch64/calleetypeid-directcall-mismatched.ll

@@ -0,0 +1,32 @@
+;; Tests that callee_type metadata attached to direct call sites are safely ignored.
+
+; RUN: llc --call-graph-section -mtriple aarch64-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+;; Test that `calleeTypeIds` field is not present in `callSites`
+; CHECK-LABEL: callSites:
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+define i32 @foo(i32 %x, i32 %y) !type !0 {
+entry:
+  ;; Call instruction with accurate callee_type.
+  ;; callee_type should be dropped seemlessly.
+  %call = call i32 @fizz(i32 %x, i32 %y), !callee_type !1
+  ;; Call instruction with mismatched callee_type.
+  ;; callee_type should be dropped seemlessly without errors.
+  %call1 = call i32 @fizz(i32 %x, i32 %y), !callee_type !3
+  %add = add nsw i32 %call, %call1
+  ;; Call instruction with mismatched callee_type.
+  ;; callee_type should be dropped seemlessly without errors.
+  %call2 = call i32 @fizz(i32 %add, i32 %y), !callee_type !3
+  %sub = sub nsw i32 %add, %call2
+  ret i32 %sub
+}
+
+declare !type !2 i32 @fizz(i32, i32)
+
+!0 = !{i64 0, !"_ZTSFiiiiE.generalized"}
+!1 = !{!2}
+!2 = !{i64 0, !"_ZTSFiiiE.generalized"}
+!3 = !{!4}
+!4 = !{i64 0, !"_ZTSFicE.generalized"}

llvm/test/CodeGen/AArch64/callsite-emit-calleetypeid-tailcall.ll

@@ -0,0 +1,19 @@
+;; Tests that call site callee type ids can be extracted and set from
+;; callee_type metadata for indirect tail calls.
+
+;; Verify the exact calleeTypeId value to ensure it is not garbage but the value
+;; computed as the type id from the callee_type metadata.
+; RUN: llc --call-graph-section -mtriple aarch64-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+define i32 @check_tailcall(ptr %func, i8 %x) !type !0 {
+entry:
+  ; CHECK: callSites:
+  ; CHECK-NEXT: - { bb: {{.*}}, offset: {{.*}}, fwdArgRegs: [], calleeTypeIds:
+  ; CHECK-NEXT: [ 3498816979441845844 ] }
+  %call = tail call i32 %func(i8 signext %x), !callee_type !1
+  ret i32 %call
+}
+
+!0 = !{i64 0, !"_ZTSFiPvcE.generalized"}
+!1 = !{!2}
+!2 = !{i64 0, !"_ZTSFicE.generalized"}

llvm/test/CodeGen/AArch64/callsite-emit-calleetypeid.ll

@@ -0,0 +1,20 @@
+;; Tests that call site callee type ids can be extracted and set from
+;; callee_type metadata.
+
+;; Verify the exact calleeTypeIds value to ensure it is not garbage but the value
+;; computed as the type id from the callee_type metadata.
+; RUN: llc --call-graph-section -mtriple aarch64-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+; CHECK: name: main
+; CHECK: callSites:
+; CHECK-NEXT: - { bb: {{.*}}, offset: {{.*}}, fwdArgRegs: [], calleeTypeIds:
+; CHECK-NEXT: [ 7854600665770582568 ] }
+define i32 @main() {
+entry:
+  %fn = load ptr, ptr null, align 8
+  call void %fn(i8 0), !callee_type !0
+  ret i32 0
+}
+
+!0 = !{!1}
+!1 = !{i64 0, !"_ZTSFvcE.generalized"}

llvm/test/CodeGen/ARM/calleetypeid-directcall-mismatched.ll

@@ -0,0 +1,32 @@
+;; Tests that callee_type metadata attached to direct call sites are safely ignored.
+
+; RUN: llc --call-graph-section -mtriple arm-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+;; Test that `calleeTypeIds` field is not present in `callSites`
+; CHECK-LABEL: callSites:
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+; CHECK-NEXT: - { bb: {{[0-9]+}}, offset: {{[0-9]+}}, fwdArgRegs: [] }
+define i32 @foo(i32 %x, i32 %y) !type !0 {
+entry:
+  ;; Call instruction with accurate callee_type.
+  ;; callee_type should be dropped seemlessly.
+  %call = call i32 @fizz(i32 %x, i32 %y), !callee_type !1
+  ;; Call instruction with mismatched callee_type.
+  ;; callee_type should be dropped seemlessly without errors.
+  %call1 = call i32 @fizz(i32 %x, i32 %y), !callee_type !3
+  %add = add nsw i32 %call, %call1
+  ;; Call instruction with mismatched callee_type.
+  ;; callee_type should be dropped seemlessly without errors.
+  %call2 = call i32 @fizz(i32 %add, i32 %y), !callee_type !3
+  %sub = sub nsw i32 %add, %call2
+  ret i32 %sub
+}
+
+declare !type !2 i32 @fizz(i32, i32)
+
+!0 = !{i64 0, !"_ZTSFiiiiE.generalized"}
+!1 = !{!2}
+!2 = !{i64 0, !"_ZTSFiiiE.generalized"}
+!3 = !{!4}
+!4 = !{i64 0, !"_ZTSFicE.generalized"}

llvm/test/CodeGen/ARM/callsite-emit-calleetypeid-tailcall.ll

@@ -0,0 +1,19 @@
+;; Tests that call site callee type ids can be extracted and set from
+;; callee_type metadata for indirect tail calls.
+
+;; Verify the exact calleeTypeId value to ensure it is not garbage but the value
+;; computed as the type id from the callee_type metadata.
+; RUN: llc --call-graph-section -mtriple arm-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+define i32 @check_tailcall(ptr %func, i8 %x) !type !0 {
+entry:
+  ; CHECK: callSites:
+  ; CHECK-NEXT: - { bb: {{.*}}, offset: {{.*}}, fwdArgRegs: [], calleeTypeIds:
+  ; CHECK-NEXT: [ 3498816979441845844 ] }
+  %call = tail call i32 %func(i8 signext %x), !callee_type !1
+  ret i32 %call
+}
+
+!0 = !{i64 0, !"_ZTSFiPvcE.generalized"}
+!1 = !{!2}
+!2 = !{i64 0, !"_ZTSFicE.generalized"}

llvm/test/CodeGen/ARM/callsite-emit-calleetypeid.ll

@@ -0,0 +1,20 @@
+;; Tests that call site callee type ids can be extracted and set from
+;; callee_type metadata.
+
+;; Verify the exact calleeTypeIds value to ensure it is not garbage but the value
+;; computed as the type id from the callee_type metadata.
+; RUN: llc --call-graph-section -mtriple arm-linux-gnu < %s -stop-after=finalize-isel -o - | FileCheck --match-full-lines %s
+
+; CHECK: name: main
+; CHECK: callSites:
+; CHECK-NEXT: - { bb: {{.*}}, offset: {{.*}}, fwdArgRegs: [], calleeTypeIds:
+; CHECK-NEXT: [ 7854600665770582568 ] }
+define i32 @main() {
+entry:
+  %fn = load ptr, ptr null, align 8
+  call void %fn(i8 0), !callee_type !0
+  ret i32 0
+}
+
+!0 = !{!1}
+!1 = !{i64 0, !"_ZTSFvcE.generalized"}