[DebugInfo] Generate fixups as emitting DWARF .debug_line.

[HsiangKai]

Summary

It is necessary to generate fixups in .debug_line as relaxation is
enabled due to the address delta may be changed after relaxation.

DWARF will record the mappings of lines and addresses in
.debug_line section. It will encode the information using special
opcodes, standard opcodes and extended opcodes in 'Line Number
Program'. I use DW_LNS_fixed_advance_pc to encode fixed length
address delta and DW_LNE_set_address to encode absolute address
to make it possible to generate fixups in .debug_line section.

[probinson]

lib/MC/MCAssembler.cpp

951

Why not cast<MCBinaryExpr> and skip the assertion?

[asb]

@echristo: Do you think you might be able to help review this?

By way of background: RISC-V compilers are strongly encouraged to make use of linker relaxation. The compiler might generate a pair of instructions to call a function, but the linker may be able to delete one once the final code layout is known. In order for the linker to do this, relocations must be generated for everything - local control flow, symbol differences, debug info.

[HsiangKai]

Create a simple slide to describe the problem in current LLVM implementation.
If there is no relocation in .debug_line, the program will be undebuggable after relaxation.

https://www.slideshare.net/chimerawang/debug-line-issues-after-relaxation-98620696

[probinson]

It looks to me like the line-table emitted by the new path would be valid. But I am not comfortable reviewing things related to MCExprs and fragments.

[HsiangKai]

In D46850#1112452, @probinson wrote:

It looks to me like the line-table emitted by the new path would be valid. But I am not comfortable reviewing things related to MCExprs and fragments.

In original implementation, it is not expected to have fixups in MCDwarfLineAddrFragment, so it is derived from MCFragment. However, I want to put fixups in MCDwarfLineAddrFragment, so I change the parent of it to MCEncodedFragmentWithFixups.

Is there any suggestion for the implementation?

[HsiangKai]

Ping.

[probinson]

I have spent a little time familiarizing myself with the fragment stuff, which prompts the inline questions.
If you agree with my point about not having two places to store the fragment contents, then I think the first step would be to do that NFC refactoring to change the base class, and follow that with the change to handle the fixups for fixed advance.

include/llvm/MC/MCFragment.h

510

These template parameters will reserve 32 bytes for the content of the fragment and 4 MCFixups. Are these reasonable values? I see there is a SmallString<8> Contents; field in this class already, and it would appear that each MCDwarfLineAddrFragment would have at most one fixup.
If the new parent class is going to reserve memory for the contents, it seems like there is some additional refactoring needed to avoid a pointless increase in memory consumption.

[HsiangKai]

Refactoring done.

[probinson]

LGTM.

include/llvm/MC/MCDwarf.h

 };
 
 class MCDwarfLineAddr {
 public:
   /// Utility function to encode a Dwarf pair of LineDelta and AddrDeltas.
   static void Encode(MCContext &Context, MCDwarfLineTableParams Params,
                      int64_t LineDelta, uint64_t AddrDelta, raw_ostream &OS);
 
+  /// Utility function to encode a Dwarf pair of LineDelta and AddrDeltas using
+  /// fixed length operands.
+  static bool FixedEncode(MCContext &Context,
+                          MCDwarfLineTableParams Params,
+                          int64_t LineDelta, uint64_t AddrDelta,
+                          raw_ostream &OS, uint32_t *Offset, uint32_t *Size);
+
   /// Utility function to emit the encoding to a streamer.
   static void Emit(MCStreamer *MCOS, MCDwarfLineTableParams Params,
                    int64_t LineDelta, uint64_t AddrDelta);
 };
 
 class MCGenDwarfInfo {
 public:
   //

include/llvm/MC/MCFragment.h

public:
   static bool classof(const MCFragment *F) {
     MCFragment::FragmentType Kind = F->getKind();
     switch (Kind) {
     default:
       return false;
     case MCFragment::FT_Relaxable:
     case MCFragment::FT_CompactEncodedInst:
     case MCFragment::FT_Data:
+    case MCFragment::FT_Dwarf:
       return true;
     }
   }
 };
 
 /// Interface implemented by fragments that contain encoded instructions and/or
 /// data.
 ///
public:
   const_fixup_iterator fixup_begin() const { return Fixups.begin(); }
 
   fixup_iterator fixup_end() { return Fixups.end(); }
   const_fixup_iterator fixup_end() const { return Fixups.end(); }
 
   static bool classof(const MCFragment *F) {
     MCFragment::FragmentType Kind = F->getKind();
     return Kind == MCFragment::FT_Relaxable || Kind == MCFragment::FT_Data ||
-           Kind == MCFragment::FT_CVDefRange;
+           Kind == MCFragment::FT_CVDefRange || Kind == MCFragment::FT_Dwarf;;
   }
 };
 
 /// Fragment for data and encoded instructions.
 ///
 class MCDataFragment : public MCEncodedFragmentWithFixups<32, 4> {
 public:
   MCDataFragment(MCSection *Sec = nullptr)
public:
 
   /// @}
 
   static bool classof(const MCFragment *F) {
     return F->getKind() == MCFragment::FT_LEB;
   }
 };
 
-class MCDwarfLineAddrFragment : public MCFragment {
+class MCDwarfLineAddrFragment : public MCEncodedFragmentWithFixups<32, 4> {
   /// LineDelta - the value of the difference between the two line numbers
   /// between two .loc dwarf directives.
   int64_t LineDelta;
 
   /// AddrDelta - The expression for the difference of the two symbols that
   /// make up the address delta between two .loc dwarf directives.
   const MCExpr *AddrDelta;
 
   SmallString<8> Contents;
 
 public:
   MCDwarfLineAddrFragment(int64_t LineDelta, const MCExpr &AddrDelta,
                           MCSection *Sec = nullptr)
-      : MCFragment(FT_Dwarf, false, 0, Sec), LineDelta(LineDelta),
-        AddrDelta(&AddrDelta) {
+      : MCEncodedFragmentWithFixups<32, 4>(FT_Dwarf, false, Sec),
+        LineDelta(LineDelta), AddrDelta(&AddrDelta) {
     Contents.push_back(0);
   }
 
   /// \name Accessors
   /// @{
 
   int64_t getLineDelta() const { return LineDelta; }
 

lib/MC/MCAssembler.cpp

for (MCFragment &Frag : Sec) {
         Fixups = FragWithFixups->getFixups();
         Contents = FragWithFixups->getContents();
       } else if (auto *FragWithFixups = dyn_cast<MCRelaxableFragment>(&Frag)) {
         Fixups = FragWithFixups->getFixups();
         Contents = FragWithFixups->getContents();
       } else if (auto *FragWithFixups = dyn_cast<MCCVDefRangeFragment>(&Frag)) {
         Fixups = FragWithFixups->getFixups();
         Contents = FragWithFixups->getContents();
+      } else if (auto *FragWithFixups = dyn_cast<MCDwarfLineAddrFragment>(&Frag)) {
+        Fixups = FragWithFixups->getFixups();
+        Contents = FragWithFixups->getContents();
       } else
         llvm_unreachable("Unknown fragment with fixups!");
       for (const MCFixup &Fixup : Fixups) {
         uint64_t FixedValue;
         bool IsResolved;
         MCValue Target;
         std::tie(Target, FixedValue, IsResolved) =
             handleFixup(Layout, Frag, Fixup);
bool MCAssembler::relaxLEB(MCAsmLayout &Layout, MCLEBFragment &LF) {
   return OldSize != LF.getContents().size();
 }
 
 bool MCAssembler::relaxDwarfLineAddr(MCAsmLayout &Layout,
                                      MCDwarfLineAddrFragment &DF) {
   MCContext &Context = Layout.getAssembler().getContext();
   uint64_t OldSize = DF.getContents().size();
   int64_t AddrDelta;
-  bool Abs = DF.getAddrDelta().evaluateKnownAbsolute(AddrDelta, Layout);
-  assert(Abs && "We created a line delta with an invalid expression");
-  (void) Abs;
+  bool Abs = DF.getAddrDelta().evaluateAsAbsolute(AddrDelta, Layout);
   int64_t LineDelta;
   LineDelta = DF.getLineDelta();
   SmallString<8> &Data = DF.getContents();
   Data.clear();
   raw_svector_ostream OSE(Data);
+  DF.getFixups().clear();
+
+  if (Abs) {
     MCDwarfLineAddr::Encode(Context, getDWARFLinetableParams(), LineDelta,
                             AddrDelta, OSE);
+  } else {
+    uint32_t Offset;
+    uint32_t Size;
+    bool SetDelta = MCDwarfLineAddr::FixedEncode(Context,
+                                                 getDWARFLinetableParams(),
+                                                 LineDelta, AddrDelta,
+                                                 OSE, &Offset, &Size);
+    // Add Fixups for address delta or new address.
+    const MCExpr *FixupExpr;
+    if (SetDelta) {
+      FixupExpr = &DF.getAddrDelta();
+    } else {
+      const MCBinaryExpr *ABE = cast_or_null<MCBinaryExpr>(&DF.getAddrDelta());
+      assert(ABE && "We expect address delta is a binary expression.");
+      FixupExpr = ABE->getLHS();
+    }
+    DF.getFixups().push_back(
+        MCFixup::create(Offset, FixupExpr,
+                        MCFixup::getKindForSize(Size, false /*isPCRel*/)));
+  }
+
   return OldSize != Data.size();
 }
 
 bool MCAssembler::relaxDwarfCallFrameFragment(MCAsmLayout &Layout,
                                               MCDwarfCallFrameFragment &DF) {
   MCContext &Context = Layout.getAssembler().getContext();
   uint64_t OldSize = DF.getContents().size();
   int64_t AddrDelta;

lib/MC/MCDwarf.cpp

void MCDwarfLineAddr::Encode(MCContext &Context, MCDwarfLineTableParams Params,
   if (NeedCopy)
     OS << char(dwarf::DW_LNS_copy);
   else {
     assert(Temp <= 255 && "Buggy special opcode encoding.");
     OS << char(Temp);
   }
 }
 
+bool MCDwarfLineAddr::FixedEncode(MCContext &Context,
+                                  MCDwarfLineTableParams Params,
+                                  int64_t LineDelta, uint64_t AddrDelta,
+                                  raw_ostream &OS,
+                                  uint32_t *Offset, uint32_t *Size) {
+  if (LineDelta != INT64_MAX) {
+    OS << char(dwarf::DW_LNS_advance_line);
+    encodeSLEB128(LineDelta, OS);
+  }
+
+  // Use address delta to adjust address or use absolute address to adjust
+  // address.
+  bool SetDelta;
+  // According to DWARF spec., the DW_LNS_fixed_advance_pc opcode takes a
+  // single uhalf (unencoded) operand. So, the maximum value of AddrDelta
+  // is 65535. We set a conservative upper bound for it for relaxation.
+  if (AddrDelta > 60000) {
+    const MCAsmInfo *asmInfo = Context.getAsmInfo();
+    unsigned AddrSize = asmInfo->getCodePointerSize();
+
+    OS << char(dwarf::DW_LNS_extended_op);
+    encodeULEB128(1 + AddrSize, OS);
+    OS << char(dwarf::DW_LNE_set_address);
+    // Generate fixup for the address.
+    *Offset = OS.tell();
+    *Size = AddrSize;
+    SetDelta = false;
+    std::vector<uint8_t> FillData;
+    FillData.insert(FillData.begin(), AddrSize, 0);
+    OS.write(reinterpret_cast<char *>(FillData.data()), AddrSize);
+  } else {
+    OS << char(dwarf::DW_LNS_fixed_advance_pc);
+    // Generate fixup for 2-bytes address delta.
+    *Offset = OS.tell();
+    *Size = 2;
+    SetDelta = true;
+    OS << char(0);
+    OS << char(0);
+  }
+
+  if (LineDelta == INT64_MAX) {
+    OS << char(dwarf::DW_LNS_extended_op);
+    OS << char(1);
+    OS << char(dwarf::DW_LNE_end_sequence);
+  } else {
+    OS << char(dwarf::DW_LNS_copy);
+  }
+
+  return SetDelta;
+}
+
 // Utility function to write a tuple for .debug_abbrev.
 static void EmitAbbrev(MCStreamer *MCOS, uint64_t Name, uint64_t Form) {
   MCOS->EmitULEB128IntValue(Name);
   MCOS->EmitULEB128IntValue(Form);
 }
 
 // When generating dwarf for assembly source files this emits
 // the data for .debug_abbrev section which contains three DIEs.

test/DebugInfo/RISCV/relax-debug-line.ll

+; RUN: llc -filetype=obj -mtriple=riscv32 -mattr=+relax %s -o - \
+; RUN:     | llvm-readobj -r | FileCheck -check-prefix=RELAX %s
+;
+; RELAX: .rela.debug_line {
+; RELAX: R_RISCV_ADD16
+; RELAX: R_RISCV_SUB16
+source_filename = "line.c"
+
+; Function Attrs: noinline nounwind optnone
+define i32 @init() !dbg !7 {
+entry:
+  ret i32 0, !dbg !11
+}
+
+; Function Attrs: noinline nounwind optnone
+define i32 @foo(i32 signext %value) !dbg !12 {
+entry:
+  %value.addr = alloca i32, align 4
+  store i32 %value, i32* %value.addr, align 4
+  call void @llvm.dbg.declare(metadata i32* %value.addr, metadata !15, metadata !DIExpression()), !dbg !16
+  %0 = load i32, i32* %value.addr, align 4, !dbg !17
+  ret i32 %0, !dbg !18
+}
+
+; Function Attrs: nounwind readnone speculatable
+declare void @llvm.dbg.declare(metadata, metadata, metadata)
+
+; Function Attrs: noinline nounwind optnone
+define i32 @bar() !dbg !19 {
+entry:
+  %result = alloca i32, align 4
+  %v = alloca i32, align 4
+  call void @llvm.dbg.declare(metadata i32* %result, metadata !20, metadata !DIExpression()), !dbg !21
+  call void @llvm.dbg.declare(metadata i32* %v, metadata !22, metadata !DIExpression()), !dbg !23
+  %call = call i32 @init(), !dbg !24
+  store i32 %call, i32* %v, align 4, !dbg !23
+  %0 = load i32, i32* %v, align 4, !dbg !25
+  %call1 = call i32 @foo(i32 signext %0), !dbg !26
+  store i32 %call1, i32* %result, align 4, !dbg !27
+  %1 = load i32, i32* %result, align 4, !dbg !28
+  ret i32 %1, !dbg !29
+}
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2)
+!1 = !DIFile(filename: "line.c", directory: "./")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 2, !"Debug Info Version", i32 3}
+!5 = !{i32 1, !"wchar_size", i32 4}
+!7 = distinct !DISubprogram(name: "init", scope: !1, file: !1, line: 1, type: !8, isLocal: false, isDefinition: true, scopeLine: 2, isOptimized: false, unit: !0, retainedNodes: !2)
+!8 = !DISubroutineType(types: !9)
+!9 = !{!10}
+!10 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
+!11 = !DILocation(line: 3, column: 3, scope: !7)
+!12 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 6, type: !13, isLocal: false, isDefinition: true, scopeLine: 7, flags: DIFlagPrototyped, isOptimized: false, unit: !0, retainedNodes: !2)
+!13 = !DISubroutineType(types: !14)
+!14 = !{!10, !10}
+!15 = !DILocalVariable(name: "value", arg: 1, scope: !12, file: !1, line: 6, type: !10)
+!16 = !DILocation(line: 6, column: 13, scope: !12)
+!17 = !DILocation(line: 8, column: 10, scope: !12)
+!18 = !DILocation(line: 8, column: 3, scope: !12)
+!19 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 11, type: !8, isLocal: false, isDefinition: true, scopeLine: 12, isOptimized: false, unit: !0, retainedNodes: !2)
+!20 = !DILocalVariable(name: "result", scope: !19, file: !1, line: 13, type: !10)
+!21 = !DILocation(line: 13, column: 7, scope: !19)
+!22 = !DILocalVariable(name: "v", scope: !19, file: !1, line: 14, type: !10)
+!23 = !DILocation(line: 14, column: 7, scope: !19)
+!24 = !DILocation(line: 14, column: 11, scope: !19)
+!25 = !DILocation(line: 16, column: 16, scope: !19)
+!26 = !DILocation(line: 16, column: 12, scope: !19)
+!27 = !DILocation(line: 16, column: 10, scope: !19)
+!28 = !DILocation(line: 18, column: 10, scope: !19)
+!29 = !DILocation(line: 18, column: 3, scope: !19)