Add ehcont section support

[arlosi]

Summary

In the future Windows will enable Control-flow Enforcement Technology (CET aka shadow stacks). To protect the path where the context is updated during exception handling, the binary is required to enumerate valid unwind entrypoints in a dedicated section which is validated when the context is being set during exception handling.

This change allows llvm to generate the section that contains the appropriate symbol references in the form expected by the msvc linker.

This feature is enabled through a new module flag, ehcontguard, which was modelled on the cfguard flag.

The change includes a test that when the module flag is enabled the section is correctly generated.

The set of exception continuation information includes returns from exceptional control flow (catchret in llvm).

In order to collect catchret we:

1. Includes an additional flag on machine basic blocks to indicate that the given block is the target of a catchret operation,
2. Introduces a new machine function pass to insert and collect symbols at the start of each of each block, and
3. Combines these targets with the other EHCont targets that were already being collected.

Change originally authored by Daniel Frampton <[e-mail]>

For more details, see MSVC documentation for /guard:ehcont

https://docs.microsoft.com/en-us/cpp/build/reference/guard-enable-eh-continuation-metadata

[rnk]

llvm/lib/CodeGen/EHContGuardCatchret.cpp

80

If we do use this name, I would prefer to use as short a name as possible. C++ names are long enough that they significantly contribute to object file size. Let's use getFunctionNumber() here instead, so we get names like $ehgcr_2_42. You'll need a non-digit separator to ensure the names are unique (consider function 5 bb 55 and function 55 bb 5).

This wasn't as much of a concern for longjmp tables, since setjmp is less common and mostly appears in C code.

82

This could be unreliable if some later pass reorders instructions or deletes the first instruction. I think it would be more reliable to take the address of the basic block. You can do this with MBB->setHasAddressTaken and MachineModuleInfo::getAddrLabelSymbol.

[arlosi]

llvm/lib/CodeGen/EHContGuardCatchret.cpp

82

MachineModuleInfo::getAddrLabelSymbol works on BasicBlocks rather than MachineBasicBlocks. So even if I setHasAddressTaken on the MBB, the associated BB might still not have it's address taken, which causes an assert failure in getAddrLabelSymbolToEmit.

I also tried using MBB.setLabelMustBeEmitted() and MBB.getSymbol(), but that produces a symbol name that doesn't end up in the symbol table.

We could add an additional field to MachineBasicBlock to store this special symbol, then emit it in the AsmPrinter in emitBasicBlockStart.

Do you have a preference, or another better solution?

[arlosi]

Use a new field in MachineBasicBlock to store the symbol for ehcont; then emit it in the AsmPrinter.

[arlosi]

Run clang-format

[arlosi]

Include both commits.

[arlosi]

Ping

[pengfei]

The changes LGTM, but I'd let @rnk sign off as well.

[arlosi]

@rnk if you could take a look that would be great.

Also, I do not have commit access.

[pengfei]

In D94835#2549120, @arlosi wrote:

@rnk if you could take a look that would be great.

Also, I do not have commit access.

I can commit it, but let's wait one or two days to see if there're objections from other people.

[arlosi]

I'm looking through this again today, and I don't think we need both IsEHCatchRetTarget and isEHFuncletEntry to be added to MBB. I'll simplify this and post an update later today.

[arlosi]

Remove redundant field IsEHContTarget on MachineBasicBlock

[rnk]

Sorry for the delay, thanks for implementing this.

llvm/include/llvm/CodeGen/MachineBasicBlock.h

private:
   bool IsEHCatchRetTarget = false;
 
   /// Indicate that this basic block is the entry block of an EH funclet.
   bool IsEHFuncletEntry = false;
 
   /// Indicate that this basic block is the entry block of a cleanup funclet.
   bool IsCleanupFuncletEntry = false;
 
+  /// Indicates that this block is an EHCont target.
+  bool IsEHContTarget = false;
+
   /// With basic block sections, this stores the Section ID of the basic block.
   MBBSectionID SectionID{0};
 
   // Indicate that this basic block begins a section.
   bool IsBeginSection = false;
 
   // Indicate that this basic block ends a section.
   bool IsEndSection = false;
 
   /// Indicate that this basic block is the indirect dest of an INLINEASM_BR.
   bool IsInlineAsmBrIndirectTarget = false;
 
   /// since getSymbol is a relatively heavy-weight operation, the symbol
   /// is only computed once and is cached.
   mutable MCSymbol *CachedMCSymbol = nullptr;
 
+  /// Cached MCSymbol for this block (used if IsEHContTarget).
+  mutable MCSymbol *CachedEHContMCSymbol = nullptr;
+
   /// Marks the end of the basic block. Used during basic block sections to
   /// calculate the size of the basic block, or the BB section ending with it.
   mutable MCSymbol *CachedEndMCSymbol = nullptr;
 
   // Intrusive list support
   MachineBasicBlock() = default;
 
   explicit MachineBasicBlock(MachineFunction &MF, const BasicBlock *BB);
#endif
   void setIsEHFuncletEntry(bool V = true) { IsEHFuncletEntry = V; }
 
   /// Returns true if this is the entry block of a cleanup funclet.
   bool isCleanupFuncletEntry() const { return IsCleanupFuncletEntry; }
 
   /// Indicates if this is the entry block of a cleanup funclet.
   void setIsCleanupFuncletEntry(bool V = true) { IsCleanupFuncletEntry = V; }
 
+  /// Returns true if this block is an EHCont target.
+  bool isEHContTarget() const { return IsEHContTarget; }
+
+  /// Indicates if this is an EHCont target.
+  void setIsEHContTarget(bool V = true) { IsEHContTarget = V; }
+
   /// Returns true if this block begins any section.
   bool isBeginSection() const { return IsBeginSection; }
 
   /// Returns true if this block ends any section.
   bool isEndSection() const { return IsEndSection; }
 
   void setIsBeginSection(bool V = true) { IsBeginSection = V; }
 
#endif
   /// MachineBasicBlocks are uniquely numbered at the function level, unless
   /// they're not in a MachineFunction yet, in which case this will return -1.
   int getNumber() const { return Number; }
   void setNumber(int N) { Number = N; }
 
   /// Return the MCSymbol for this basic block.
   MCSymbol *getSymbol() const;
 
+  /// Return the EHCont Symbol for this basic block.
+  MCSymbol *getEHContSymbol() const;
+
   Optional<uint64_t> getIrrLoopHeaderWeight() const {
     return IrrLoopHeaderWeight;
   }
 
   void setIrrLoopHeaderWeight(uint64_t Weight) {
     IrrLoopHeaderWeight = Weight;
   }
 

llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp

void AsmPrinter::emitBasicBlockStart(const MachineBasicBlock &MBB) {
   } else {
     if (isVerbose()) {
       // NOTE: Want this comment at start of line, don't emit with AddComment.
       OutStreamer->emitRawComment(" %bb." + Twine(MBB.getNumber()) + ":",
                                   false);
     }
   }
 
+  if (MBB.isEHContTarget()) {
+    OutStreamer->emitLabel(MBB.getEHContSymbol());
+  }
+
   // With BB sections, each basic block must handle CFI information on its own
   // if it begins a section (Entry block is handled separately by
   // AsmPrinterHandler::beginFunction).
   if (MBB.isBeginSection() && !MBB.isEntryBlock())
     for (const HandlerInfo &HI : Handlers)
       HI.Handler->beginBasicBlock(MBB);
 }
 

llvm/lib/CodeGen/AsmPrinter/WinException.cpp

void WinException::beginFunclet(const MachineBasicBlock &MBB,
   }
 
   // Mark 'Sym' as starting our funclet.
   if (shouldEmitMoves || shouldEmitPersonality) {
     CurrentFuncletTextSection = Asm->OutStreamer->getCurrentSectionOnly();
     Asm->OutStreamer->EmitWinCFIStartProc(Sym);
   }
 
-  // Add 'Sym' as an ehcont target
-  if (CurrentFuncletEntry->isEHFuncletEntry())
-    EHContTargets.push_back(Sym);
-
   if (shouldEmitPersonality) {
     const TargetLoweringObjectFile &TLOF = Asm->getObjFileLowering();
     const Function *PerFn = nullptr;
 
     // Determine which personality routine we are using for this funclet.
     if (F.hasPersonalityFn())
       PerFn = dyn_cast<Function>(F.getPersonalityFn()->stripPointerCasts());
     const MCSymbol *PersHandlerSym =

llvm/lib/CodeGen/EHContGuardCatchret.cpp

bool EHContGuardCatchret::runOnMachineFunction(MachineFunction &MF) {
   // Skip modules for which the ehcontguard flag is not set.
   if (!MF.getMMI().getModule()->getModuleFlag("ehcontguard"))
     return false;
 
   // Skip functions that do not have catchret
   if (!MF.hasEHCatchRet())
     return false;
 
-  bool result = false;
+  bool Result = false;
 
   for (MachineBasicBlock &MBB : MF) {
     if (MBB.isEHCatchRetTarget() && !MBB.isEHFuncletEntry()) {
-      auto MBBI = MBB.begin();
-      SmallString<128> SymbolName;
-      raw_svector_ostream(SymbolName)
-          << "$ehgcr_" << MF.getName() << MBB.getNumber();
-      MCSymbol *CrSymbol = MF.getContext().getOrCreateSymbol(SymbolName);
-      MBBI->setPreInstrSymbol(MF, CrSymbol);
-      MF.addCatchretTarget(CrSymbol);
+      MBB.setIsEHContTarget();
+      MF.addCatchretTarget(MBB.getEHContSymbol());
       EHContGuardCatchretTargets++;
-      result = true;
+      Result = true;
     }
   }
 
-  return result;
+  return Result;
 }

llvm/lib/CodeGen/MachineBasicBlock.cpp

if (MF->hasBBSections() && isBeginSection()) {
       CachedMCSymbol = Ctx.getOrCreateSymbol(Twine(Prefix) + "BB" +
                                              Twine(MF->getFunctionNumber()) +
                                              "_" + Twine(getNumber()));
     }
   }
   return CachedMCSymbol;
 }
 
+MCSymbol *MachineBasicBlock::getEHContSymbol() const {
+  if (!CachedEHContMCSymbol) {
+    const MachineFunction *MF = getParent();
+    SmallString<128> SymbolName;
+    raw_svector_ostream(SymbolName)
+        << "$ehgcr_" << MF->getFunctionNumber() << '_' << getNumber();
+    CachedEHContMCSymbol = MF->getContext().getOrCreateSymbol(SymbolName);
+  }
+  return CachedEHContMCSymbol;
+}
+
 MCSymbol *MachineBasicBlock::getEndSymbol() const {
   if (!CachedEndMCSymbol) {
     const MachineFunction *MF = getParent();
     MCContext &Ctx = MF->getContext();
     auto Prefix = Ctx.getAsmInfo()->getPrivateLabelPrefix();
     CachedEndMCSymbol = Ctx.getOrCreateSymbol(Twine(Prefix) + "BB_END" +
                                               Twine(MF->getFunctionNumber()) +
                                               "_" + Twine(getNumber()));

llvm/test/CodeGen/AArch64/ehcontguard.ll

 ; RUN: llc < %s -mtriple=aarch64-windows | FileCheck %s
 ; EHCont Guard is currently only available on Windows
 
 ; CHECK: .set @feat.00, 16384
 
 ; CHECK: .section .gehcont$y
 
-; // struct with dtor
-; struct A
-; {
-;     ~A();
-; };
-;
-; // function we call that might raise an exception
-; void func2();
-;
-; // function which has to call the dtor on unwind
-; void func1()
-; {
-;     A a;
-;     func2();
-; }
-
-%struct.A = type { [4 x i32], [16 x i8] }
-declare dso_local %struct.A* @"??0A@@QEAA@XZ"(%struct.A* returned %0)
-declare dso_local void @"??1A@@QEAA@XZ"(%struct.A* %0)
-declare dso_local i32 @__CxxFrameHandler3(...)
-declare dso_local void @"?func2@@YAXXZ"()
-
-; Function Attrs: noinline optnone uwtable
 define dso_local void @"?func1@@YAXXZ"() #0 personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
-  %1 = alloca %struct.A
-  %2 = call %struct.A* @"??0A@@QEAA@XZ"(%struct.A* %1) #3
+entry:
   invoke void @"?func2@@YAXXZ"()
-          to label %3 unwind label %4
-
-3:                                                ; preds = %0
-  call void @"??1A@@QEAA@XZ"(%struct.A* %1) #3
+          to label %invoke.cont unwind label %catch.dispatch
+catch.dispatch:                                   ; preds = %entry
+  %0 = catchswitch within none [label %catch] unwind to caller
+catch:                                            ; preds = %catch.dispatch
+  %1 = catchpad within %0 [i8* null, i32 64, i8* null]
+  catchret from %1 to label %catchret.dest
+catchret.dest:                                    ; preds = %catch
+  br label %try.cont
+try.cont:                                         ; preds = %catchret.dest, %invoke.cont
   ret void
-
-4:                                                ; preds = %0
-  %5 = cleanuppad within none []
-  call void @"??1A@@QEAA@XZ"(%struct.A* %1) #3 [ "funclet"(token %5) ]
-  cleanupret from %5 unwind to caller
+invoke.cont:                                      ; preds = %entry
+  br label %try.cont
 }
 
+declare dso_local void @"?func2@@YAXXZ"() #1
+declare dso_local i32 @__CxxFrameHandler3(...)
+
 !llvm.module.flags = !{!0}
-!0 = !{i32 1, !"ehcontguard", i32 1}
+!0 = !{i32 1, !"ehcontguard", i32 1}
 

llvm/test/CodeGen/X86/ehcontguard.ll

 ; RUN: llc < %s -mtriple=x86_64-pc-windows-msvc | FileCheck %s
 ; EHCont Guard is currently only available on Windows
 
 ; CHECK: .set @feat.00, 16384
 
 ; CHECK: .section .gehcont$y
 
-; // struct with dtor
-; struct A
-; {
-;     ~A();
-; };
-;
-; // function we call that might raise an exception
-; void func2();
-;
-; // function which has to call the dtor on unwind
-; void func1()
-; {
-;     A a;
-;     func2();
-; }
-
-%struct.A = type { [4 x i32], [16 x i8] }
-declare dso_local %struct.A* @"??0A@@QEAA@XZ"(%struct.A* returned %0)
-declare dso_local void @"??1A@@QEAA@XZ"(%struct.A* %0)
-declare dso_local i32 @__CxxFrameHandler3(...)
-declare dso_local void @"?func2@@YAXXZ"()
-
-; Function Attrs: noinline optnone uwtable
 define dso_local void @"?func1@@YAXXZ"() #0 personality i8* bitcast (i32 (...)* @__CxxFrameHandler3 to i8*) {
-  %1 = alloca %struct.A
-  %2 = call %struct.A* @"??0A@@QEAA@XZ"(%struct.A* %1) #3
+entry:
   invoke void @"?func2@@YAXXZ"()
-          to label %3 unwind label %4
-
-3:                                                ; preds = %0
-  call void @"??1A@@QEAA@XZ"(%struct.A* %1) #3
+          to label %invoke.cont unwind label %catch.dispatch
+catch.dispatch:                                   ; preds = %entry
+  %0 = catchswitch within none [label %catch] unwind to caller
+catch:                                            ; preds = %catch.dispatch
+  %1 = catchpad within %0 [i8* null, i32 64, i8* null]
+  catchret from %1 to label %catchret.dest
+catchret.dest:                                    ; preds = %catch
+  br label %try.cont
+try.cont:                                         ; preds = %catchret.dest, %invoke.cont
   ret void
-
-4:                                                ; preds = %0
-  %5 = cleanuppad within none []
-  call void @"??1A@@QEAA@XZ"(%struct.A* %1) #3 [ "funclet"(token %5) ]
-  cleanupret from %5 unwind to caller
+invoke.cont:                                      ; preds = %entry
+  br label %try.cont
 }
 
+declare dso_local void @"?func2@@YAXXZ"() #1
+declare dso_local i32 @__CxxFrameHandler3(...)
+
 !llvm.module.flags = !{!0}
-!0 = !{i32 1, !"ehcontguard", i32 1}
+!0 = !{i32 1, !"ehcontguard", i32 1}