WinEH for unnamed functions

[• pbrunet]

Summary

dtor and catch symbol are required for windows exception handling. These symboles were defined based on function name which is incorrect in the case of unnamed functions as we may have multiple functions with the same empty name.
It leads to assert if they are available or crash on the exception path.

[paquette]

I feel like the test here could use some simplification to make it clearer what you're testing. If there's any IR in there that you don't *absolutely* need, it'd be good to remove it. This makes it easier for others to debug failures in the future.

test/CodeGen/WinEH/wineh-unnamed.ll

2

Spelling and grammar could use some fixing here.

103

You can remove the ; Function Attrs... comments. Basically any comment that you didn't write for your test can go.

117

You don't need comments like the ; preds... lines in tests

153

Do you need all of these attributes for your test? If you don't, it would simplify it a lot to remove some of them.

[paquette]

I also suggest adding some reviewers to your patch! If you look at the blamelist of WinException.cpp you'll probably find some people that know the code well that can look at your stuff. I'm not a Windows expert, but if I was I'd give you a proper review. :)

[• pbrunet]

Thanks for the review and sorry for the delay.

I Updated this patch with :

• handle new API : replace getRealLinkageName by dropLLVMManglingEscape
• Reduce test case.

[rnk]

The same logic would need to be applied to the X86ISelLowering call site of MCContext::getOrCreateParentFrameOffsetSymbol. I think it would be preferable to manually number and provide names for nameless functions that use funclet personalities in WinEHPrepare.

[majnemer]

In D37434#978881, @rnk wrote:

The same logic would need to be applied to the X86ISelLowering call site of MCContext::getOrCreateParentFrameOffsetSymbol. I think it would be preferable to manually number and provide names for nameless functions that use funclet personalities in WinEHPrepare.

Heh, I was just about to recommend using something like Mangler::getNameWithPrefix (with CannotUsePrivateLabel set to true) into a raw_svector_ostream. This way we get a name which will correspond to what we get in the object file. We would only use Mangler::getNameWithPrefix if the GV had no name.

[rnk]

In D37434#978885, @majnemer wrote:

In D37434#978881, @rnk wrote:

The same logic would need to be applied to the X86ISelLowering call site of MCContext::getOrCreateParentFrameOffsetSymbol. I think it would be preferable to manually number and provide names for nameless functions that use funclet personalities in WinEHPrepare.

Heh, I was just about to recommend using something like Mangler::getNameWithPrefix (with CannotUsePrivateLabel set to true) into a raw_svector_ostream. This way we get a name which will correspond to what we get in the object file. We would only use Mangler::getNameWithPrefix if the GV had no name.

I think part of the trouble here is that I put the logic for computing the label name in MCContext, which cannot use IR constructs. Having this in the Mangler would work better.

lib/CodeGen/AsmPrinter/WinException.cpp

 #include "llvm/Target/TargetOptions.h"
 using namespace llvm;
 
+static std::string GetNameFromFunction(llvm::Function const& F) {
+  std::string name;
+  if(F.hasName())
+    name = F.getName();
+  else 
+    name = std::to_string(
+        std::distance(
+          F.getParent()->begin(),
+          std::find_if(
+            F.getParent()->begin(),
+            F.getParent()->end(),
+            [&F](Function const& Fun){return &Fun == &F;})
+          )
+        );
+  return GlobalValue::dropLLVMManglingEscape(name);
+}
+
 WinException::WinException(AsmPrinter *A) : EHStreamer(A) {
   // MSVC's EH tables are always composed of 32-bit words.  All known 64-bit
   // platforms use an imagerel32 relocation to refer to symbols.

       // make sure we emit the parent offset label. Some unreferenced filter
       // functions may still refer to it.
       const WinEHFuncInfo &FuncInfo = *MF->getWinEHFuncInfo();
-      StringRef FLinkageName =
-          GlobalValue::dropLLVMManglingEscape(MF->getFunction().getName());
+      std::string FLinkageName = GetNameFromFunction(MF->getFunction());
       emitEHRegistrationOffsetLabel(FuncInfo, FLinkageName);
     }
     shouldEmitLSDA = hasEHFunclets;

   // their funclet entry block's number.
   const MachineFunction *MF = MBB->getParent();
   const Function &F = MF->getFunction();
-  StringRef FuncLinkageName = GlobalValue::dropLLVMManglingEscape(F.getName());
+  std::string FuncLinkageName = GetNameFromFunction(F);
   MCContext &Ctx = MF->getContext();
   StringRef HandlerPrefix = MBB->isCleanupFuncletEntry() ? "dtor" : "catch";
   return Ctx.getOrCreateSymbol("?" + HandlerPrefix + "$" +

         !CurrentFuncletEntry->isCleanupFuncletEntry()) {
       // If this is a C++ catch funclet (or the parent function),
       // emit a reference to the LSDA for the parent function.
-      StringRef FuncLinkageName = GlobalValue::dropLLVMManglingEscape(F.getName());
+      std::string FuncLinkageName = GetNameFromFunction(F);
       MCSymbol *FuncInfoXData = Asm->OutContext.getOrCreateSymbol(
           Twine("$cppxdata$", FuncLinkageName));
       Asm->OutStreamer->EmitValue(create32bitRef(FuncInfoXData), 4);

 
   // Emit a label assignment with the SEH frame offset so we can use it for
   // llvm.x86.seh.recoverfp.
-  StringRef FLinkageName =
-      GlobalValue::dropLLVMManglingEscape(MF->getFunction().getName());
+  std::string FLinkageName = GetNameFromFunction(MF->getFunction());
   MCSymbol *ParentFrameOffset =
       Ctx.getOrCreateParentFrameOffsetSymbol(FLinkageName);
   const MCExpr *MCOffset =

   auto &OS = *Asm->OutStreamer;
   const WinEHFuncInfo &FuncInfo = *MF->getWinEHFuncInfo();
 
-  StringRef FuncLinkageName = GlobalValue::dropLLVMManglingEscape(F.getName());
+  std::string FuncLinkageName = GetNameFromFunction(F);
 
   SmallVector<std::pair<const MCExpr *, int>, 4> IPToStateTable;
   MCSymbol *FuncInfoXData = nullptr;

 void WinException::emitExceptHandlerTable(const MachineFunction *MF) {
   MCStreamer &OS = *Asm->OutStreamer;
   const Function &F = MF->getFunction();
-  StringRef FLinkageName = GlobalValue::dropLLVMManglingEscape(F.getName());
+  std::string FLinkageName = GetNameFromFunction(F);
 
   bool VerboseAsm = OS.isVerboseAsm();
   auto AddComment = [&](const Twine &Comment) {

test/CodeGen/WinEH/wineh-unnamed.ll

+; Check we can compile winEH file with unnamed functions without collision on 
+; funclet symbol name.
+; We can improve this test and run it to see that generated program crash but
+; it will only work on windows targets
+; RUN: llc < %s
+
+; ModuleID = 'wineh-unnamed.ll'
+source_filename = "C:\5CUsers\5Cvagrant\5CDesktop\5Ctest.cpp"
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+%rtti.TypeDescriptor5 = type { i8**, i8*, [6 x i8] }
+%eh.CatchableType = type { i32, i32, i32, i32, i32, i32, i32 }
+%eh.CatchableTypeArray.2 = type { i32, [2 x i32] }
+%eh.ThrowInfo = type { i32, i32, i32, i32 }
+
+$"\01??_C@_03JIDPIAJI@err?$AA@" = comdat any
+
+$"\01??_R0PEAD@8" = comdat any
+
+$"_CT??_R0PEAD@88" = comdat any
+
+$"\01??_R0PEAX@8" = comdat any
+
+$"_CT??_R0PEAX@88" = comdat any
+
+$_CTA2PEAD = comdat any
+
+$_TIC2PEAD = comdat any
+
+@"\01??_C@_03JIDPIAJI@err?$AA@" = linkonce_odr unnamed_addr constant [4 x i8] c"err\00", comdat, align 1
+@"\01??_7type_info@@6B@" = external constant i8*
+@"\01??_R0PEAD@8" = linkonce_odr global %rtti.TypeDescriptor5 { i8** @"\01??_7type_info@@6B@", i8* null, [6 x i8] c".PEAD\00" }, comdat
+@__ImageBase = external constant i8
+@"_CT??_R0PEAD@88" = linkonce_odr unnamed_addr constant %eh.CatchableType { i32 1, i32 trunc (i64 sub nuw nsw (i64 ptrtoint (%rtti.TypeDescriptor5* @"\01??_R0PEAD@8" to i64), i64 ptrtoint (i8* @__ImageBase to i64)) to i32), i32 0, i32 -1, i32 0, i32 8, i32 0 }, section ".xdata", comdat
+@"\01??_R0PEAX@8" = linkonce_odr global %rtti.TypeDescriptor5 { i8** @"\01??_7type_info@@6B@", i8* null, [6 x i8] c".PEAX\00" }, comdat
+@"_CT??_R0PEAX@88" = linkonce_odr unnamed_addr constant %eh.CatchableType { i32 1, i32 trunc (i64 sub nuw nsw (i64 ptrtoint (%rtti.TypeDescriptor5* @"\01??_R0PEAX@8" to i64), i64 ptrtoint (i8* @__ImageBase to i64)) to i32), i32 0, i32 -1, i32 0, i32 8, i32 0 }, section ".xdata", comdat
+@_CTA2PEAD = linkonce_odr unnamed_addr constant %eh.CatchableTypeArray.2 { i32 2, [2 x i32] [i32 trunc (i64 sub nuw nsw (i64 ptrtoint (%eh.CatchableType* @"_CT??_R0PEAD@88" to i64), i64 ptrtoint (i8* @__ImageBase to i64)) to i32), i32 trunc (i64 sub nuw nsw (i64 ptrtoint (%eh.CatchableType* @"_CT??_R0PEAX@88" to i64), i64 ptrtoint (i8* @__ImageBase to i64)) to i32)] }, section ".xdata", comdat
+@_TIC2PEAD = linkonce_odr unnamed_addr constant %eh.ThrowInfo { i32 1, i32 0, i32 0, i32 trunc (i64 sub nuw nsw (i64 ptrtoint (%eh.CatchableTypeArray.2* @_CTA2PEAD to i64), i64 ptrtoint (i8* @__ImageBase to i64)) to i32) }, section ".xdata", comdat
+
+define void @"\01?raise@@YAXH@Z"(i32 %i) local_unnamed_addr {
+entry:
+  %tmp = alloca i8*, align 8
+  %cmp = icmp eq i32 %i, 1
+  br i1 %cmp, label %if.then, label %if.end
+
+if.then:
+  store i8* getelementptr inbounds ([4 x i8], [4 x i8]* @"\01??_C@_03JIDPIAJI@err?$AA@", i64 0, i64 0), i8** %tmp, align 8
+  %0 = bitcast i8** %tmp to i8*
+  call void @_CxxThrowException(i8* nonnull %0, %eh.ThrowInfo* nonnull @_TIC2PEAD)
+  unreachable
+
+if.end:
+  ret void
+}
+
+declare dllimport void @_CxxThrowException(i8*, %eh.ThrowInfo*) local_unnamed_addr
+
+define internal fastcc i32 @0(i32 %i) unnamed_addr personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
+entry:
+  invoke void @"\01?raise@@YAXH@Z"(i32 %i)
+          to label %return unwind label %catch.dispatch
+
+catch.dispatch:
+  %0 = catchswitch within none [label %catch] unwind to caller
+
+catch:
+  %1 = catchpad within %0 [i8* null, i32 64, i8* null]
+  catchret from %1 to label %return
+
+return:
+  %retval.0 = phi i32 [ 1, %catch ], [ 0, %entry ]
+  ret i32 %retval.0
+}
+
+declare i32 @__CxxFrameHandler3(...)
+
+define internal fastcc i32 @1(i32 %i) unnamed_addr personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
+entry:
+  %add = add nsw i32 %i, -1
+  invoke void @"\01?raise@@YAXH@Z"(i32 %add)
+          to label %return unwind label %catch.dispatch
+
+catch.dispatch:
+  %0 = catchswitch within none [label %catch] unwind to caller
+
+catch:
+  %1 = catchpad within %0 [i8* null, i32 64, i8* null]
+  catchret from %1 to label %return
+
+return:
+  %retval.0 = phi i32 [ 4, %catch ], [ 2, %entry ]
+  ret i32 %retval.0
+}
+
+define i32 @main(i32 %argc) local_unnamed_addr {
+entry:
+  %call = tail call fastcc i32 @0(i32 %argc)
+  %call1 = tail call fastcc i32 @1(i32 %argc)
+  %add = add nsw i32 %call1, %call
+  ret i32 %add
+}