[MC] Add .loc_label instruction (#99710)

As discussed in [the RFC](https://discourse.llvm.org/t/rfc-extending-llvm-mc-loc-directive-with-labeling-support/79608) we need a way to create labels in the assembler-generated line section in order to support the future addition of the [DW_AT_LLVM_stmt_sequence](https://discourse.llvm.org/t/rfc-new-dwarf-attribute-for-symbolication-of-merged-functions/79434) attribute. We have a similar precedent for such behavior with the [.cfi_label](#97922) instruction - so we add the `.loc_label THE_LABEL_NAME` instruction which: - Terminates the current line sequence in the line section - Creates a new label with the specified label name in the `.debug_line` section
llvm · Sep 20, 2024 · 28646d0 · 28646d0
1 parent 42b696d
commit 28646d0
Show file tree

Hide file tree

Showing 9 changed files with 206 additions and 16 deletions.
diff --git a/llvm/include/llvm/MC/MCDwarf.h b/llvm/include/llvm/MC/MCDwarf.h
@@ -194,11 +194,23 @@ class MCDwarfLineEntry : public MCDwarfLoc {
 
 public:
   // Constructor to create an MCDwarfLineEntry given a symbol and the dwarf loc.
-  MCDwarfLineEntry(MCSymbol *label, const MCDwarfLoc loc)
-      : MCDwarfLoc(loc), Label(label) {}
+  MCDwarfLineEntry(MCSymbol *label, const MCDwarfLoc loc,
+                   MCSymbol *lineStreamLabel = nullptr,
+                   SMLoc streamLabelDefLoc = {})
+      : MCDwarfLoc(loc), Label(label), LineStreamLabel(lineStreamLabel),
+        StreamLabelDefLoc(streamLabelDefLoc) {}
 
   MCSymbol *getLabel() const { return Label; }
 
+  // This is the label that is to be emitted into the line stream. If this is
+  // non-null and we need to emit a label, also make sure to restart the current
+  // line sequence.
+  MCSymbol *LineStreamLabel;
+
+  // Location where LineStreamLabel was defined. If there is an error emitting
+  // LineStreamLabel, we can use the SMLoc to report an error.
+  SMLoc StreamLabelDefLoc;
+
   // This indicates the line entry is synthesized for an end entry.
   bool IsEndEntry = false;
 
@@ -365,6 +377,9 @@ class MCDwarfLineTable {
   emitOne(MCStreamer *MCOS, MCSection *Section,
           const MCLineSection::MCDwarfLineEntryCollection &LineEntries);
 
+  void endCurrentSeqAndEmitLineStreamLabel(MCStreamer *MCOS, SMLoc DefLoc,
+                                           StringRef Name);
+
   Expected<unsigned> tryGetFile(StringRef &Directory, StringRef &FileName,
                                 std::optional<MD5::MD5Result> Checksum,
                                 std::optional<StringRef> Source,

diff --git a/llvm/include/llvm/MC/MCObjectStreamer.h b/llvm/include/llvm/MC/MCObjectStreamer.h
@@ -146,7 +146,8 @@ class MCObjectStreamer : public MCStreamer {
   void emitDwarfAdvanceLineAddr(int64_t LineDelta, const MCSymbol *LastLabel,
                                 const MCSymbol *Label,
                                 unsigned PointerSize) override;
-  void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel) override;
+  void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel,
+                             MCSymbol *EndLabel = nullptr) override;
   void emitDwarfAdvanceFrameAddr(const MCSymbol *LastLabel,
                                  const MCSymbol *Label, SMLoc Loc);
   void emitCVLocDirective(unsigned FunctionId, unsigned FileNo, unsigned Line,

diff --git a/llvm/include/llvm/MC/MCStreamer.h b/llvm/include/llvm/MC/MCStreamer.h
@@ -912,6 +912,9 @@ class MCStreamer {
                                      unsigned Isa, unsigned Discriminator,
                                      StringRef FileName);
 
+  /// This implements the '.loc_label Name' directive.
+  virtual void emitDwarfLocLabelDirective(SMLoc Loc, StringRef Name);
+
   /// Associate a filename with a specified logical file number, and also
   /// specify that file's checksum information.  This implements the '.cv_file 4
   /// "foo.c"' assembler directive. Returns true on success.
@@ -1119,7 +1122,8 @@ class MCStreamer {
   virtual void emitDwarfLineStartLabel(MCSymbol *StartSym);
 
   /// Emit the debug line end entry.
-  virtual void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel) {}
+  virtual void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel,
+                                     MCSymbol *EndLabel = nullptr) {}
 
   /// If targets does not support representing debug line section by .loc/.file
   /// directives in assembly output, we need to populate debug line section with

diff --git a/llvm/lib/MC/MCAsmStreamer.cpp b/llvm/lib/MC/MCAsmStreamer.cpp
@@ -301,6 +301,8 @@ class MCAsmStreamer final : public MCStreamer {
                              unsigned Flags, unsigned Isa,
                              unsigned Discriminator,
                              StringRef FileName) override;
+  virtual void emitDwarfLocLabelDirective(SMLoc Loc, StringRef Name) override;
+
   MCSymbol *getDwarfLineTableSymbol(unsigned CUID) override;
 
   bool emitCVFileDirective(unsigned FileNo, StringRef Filename,
@@ -429,7 +431,8 @@ class MCAsmStreamer final : public MCStreamer {
 
   void emitDwarfLineStartLabel(MCSymbol *StartSym) override;
 
-  void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel) override;
+  void emitDwarfLineEndEntry(MCSection *Section, MCSymbol *LastLabel,
+                             MCSymbol *EndLabel = nullptr) override;
 
   void emitDwarfAdvanceLineAddr(int64_t LineDelta, const MCSymbol *LastLabel,
                                 const MCSymbol *Label,
@@ -1767,6 +1770,12 @@ void MCAsmStreamer::emitDwarfLocDirective(unsigned FileNo, unsigned Line,
                                           Discriminator, FileName);
 }
 
+void MCAsmStreamer::emitDwarfLocLabelDirective(SMLoc Loc, StringRef Name) {
+  MCStreamer::emitDwarfLocLabelDirective(Loc, Name);
+  OS << ".loc_label\t" << Name;
+  EmitEOL();
+}
+
 MCSymbol *MCAsmStreamer::getDwarfLineTableSymbol(unsigned CUID) {
   // Always use the zeroth line table, since asm syntax only supports one line
   // table for now.
@@ -2579,7 +2588,8 @@ void MCAsmStreamer::emitDwarfLineStartLabel(MCSymbol *StartSym) {
 }
 
 void MCAsmStreamer::emitDwarfLineEndEntry(MCSection *Section,
-                                          MCSymbol *LastLabel) {
+                                          MCSymbol *LastLabel,
+                                          MCSymbol *EndLabel) {
   // If the targets write the raw debug line data for assembly output (We can
   // not switch to Section and add the end symbol there for assembly output)
   // we currently use the .text end label as any section end. This will not
@@ -2596,9 +2606,10 @@ void MCAsmStreamer::emitDwarfLineEndEntry(MCSection *Section,
   MCSection *TextSection = Ctx.getObjectFileInfo()->getTextSection();
   assert(TextSection->hasEnded() && ".text section is not end!");
 
-  MCSymbol *SectionEnd = TextSection->getEndSymbol(Ctx);
+  if (!EndLabel)
+    EndLabel = TextSection->getEndSymbol(Ctx);
   const MCAsmInfo *AsmInfo = Ctx.getAsmInfo();
-  emitDwarfAdvanceLineAddr(INT64_MAX, LastLabel, SectionEnd,
+  emitDwarfAdvanceLineAddr(INT64_MAX, LastLabel, EndLabel,
                            AsmInfo->getCodePointerSize());
 }
 

diff --git a/llvm/lib/MC/MCDwarf.cpp b/llvm/lib/MC/MCDwarf.cpp
@@ -172,6 +172,7 @@ void MCDwarfLineTable::emitOne(
     const MCLineSection::MCDwarfLineEntryCollection &LineEntries) {
 
   unsigned FileNum, LastLine, Column, Flags, Isa, Discriminator;
+  bool IsAtStartSeq;
   MCSymbol *LastLabel;
   auto init = [&]() {
     FileNum = 1;
@@ -181,6 +182,7 @@ void MCDwarfLineTable::emitOne(
     Isa = 0;
     Discriminator = 0;
     LastLabel = nullptr;
+    IsAtStartSeq = true;
   };
   init();
 
@@ -189,6 +191,17 @@ void MCDwarfLineTable::emitOne(
   for (const MCDwarfLineEntry &LineEntry : LineEntries) {
     MCSymbol *Label = LineEntry.getLabel();
     const MCAsmInfo *asmInfo = MCOS->getContext().getAsmInfo();
+
+    if (LineEntry.LineStreamLabel) {
+      if (!IsAtStartSeq) {
+        MCOS->emitDwarfLineEndEntry(Section, LastLabel,
+                                    /*EndLabel =*/LastLabel);
+        init();
+      }
+      MCOS->emitLabel(LineEntry.LineStreamLabel, LineEntry.StreamLabelDefLoc);
+      continue;
+    }
+
     if (LineEntry.IsEndEntry) {
       MCOS->emitDwarfAdvanceLineAddr(INT64_MAX, LastLabel, Label,
                                      asmInfo->getCodePointerSize());
@@ -243,17 +256,34 @@ void MCDwarfLineTable::emitOne(
     Discriminator = 0;
     LastLine = LineEntry.getLine();
     LastLabel = Label;
+    IsAtStartSeq = false;
   }
 
   // Generate DWARF line end entry.
   // We do not need this for DwarfDebug that explicitly terminates the line
   // table using ranges whenever CU or section changes. However, the MC path
   // does not track ranges nor terminate the line table. In that case,
   // conservatively use the section end symbol to end the line table.
-  if (!EndEntryEmitted)
+  if (!EndEntryEmitted && !IsAtStartSeq)
     MCOS->emitDwarfLineEndEntry(Section, LastLabel);
 }
 
+void MCDwarfLineTable::endCurrentSeqAndEmitLineStreamLabel(MCStreamer *MCOS,
+                                                           SMLoc DefLoc,
+                                                           StringRef Name) {
+  auto &ctx = MCOS->getContext();
+  auto *LineStreamLabel = ctx.getOrCreateSymbol(Name);
+  auto *LineSym = ctx.createTempSymbol();
+  MCOS->emitLabel(LineSym);
+  const MCDwarfLoc &DwarfLoc = ctx.getCurrentDwarfLoc();
+
+  // Create a 'fake' line entry by having LineStreamLabel be non-null. This
+  // won't actually emit any line information, it will reset the line table
+  // sequence and emit a label at the start of the new line table sequence.
+  MCDwarfLineEntry LineEntry(LineSym, DwarfLoc, LineStreamLabel, DefLoc);
+  getMCLineSections().addLineEntry(LineEntry, MCOS->getCurrentSectionOnly());
+}
+
 //
 // This emits the Dwarf file and the line tables.
 //

diff --git a/llvm/lib/MC/MCObjectStreamer.cpp b/llvm/lib/MC/MCObjectStreamer.cpp
@@ -467,20 +467,24 @@ void MCObjectStreamer::emitDwarfAdvanceLineAddr(int64_t LineDelta,
 }
 
 void MCObjectStreamer::emitDwarfLineEndEntry(MCSection *Section,
-                                             MCSymbol *LastLabel) {
-  // Emit a DW_LNE_end_sequence for the end of the section.
-  // Use the section end label to compute the address delta and use INT64_MAX
-  // as the line delta which is the signal that this is actually a
+                                             MCSymbol *LastLabel,
+                                             MCSymbol *EndLabel) {
+  // Emit a DW_LNE_end_sequence into the line table. When EndLabel is null, it
+  // means we should emit the entry for the end of the section and therefore we
+  // use the section end label for the reference label. After having the
+  // appropriate reference label, we emit the address delta and use INT64_MAX as
+  // the line delta which is the signal that this is actually a
   // DW_LNE_end_sequence.
-  MCSymbol *SectionEnd = endSection(Section);
+  if (!EndLabel)
+    EndLabel = endSection(Section);
 
   // Switch back the dwarf line section, in case endSection had to switch the
   // section.
   MCContext &Ctx = getContext();
   switchSection(Ctx.getObjectFileInfo()->getDwarfLineSection());
 
   const MCAsmInfo *AsmInfo = Ctx.getAsmInfo();
-  emitDwarfAdvanceLineAddr(INT64_MAX, LastLabel, SectionEnd,
+  emitDwarfAdvanceLineAddr(INT64_MAX, LastLabel, EndLabel,
                            AsmInfo->getCodePointerSize());
 }
 

diff --git a/llvm/lib/MC/MCParser/AsmParser.cpp b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -485,6 +485,7 @@ class AsmParser : public MCAsmParser {
     DK_FILE,
     DK_LINE,
     DK_LOC,
+    DK_LOC_LABEL,
     DK_STABS,
     DK_CV_FILE,
     DK_CV_FUNC_ID,
@@ -580,10 +581,11 @@ class AsmParser : public MCAsmParser {
   // ".align{,32}", ".p2align{,w,l}"
   bool parseDirectiveAlign(bool IsPow2, unsigned ValueSize);
 
-  // ".file", ".line", ".loc", ".stabs"
+  // ".file", ".line", ".loc", ".loc_label", ".stabs"
   bool parseDirectiveFile(SMLoc DirectiveLoc);
   bool parseDirectiveLine();
   bool parseDirectiveLoc();
+  bool parseDirectiveLocLabel(SMLoc DirectiveLoc);
   bool parseDirectiveStabs();
 
   // ".cv_file", ".cv_func_id", ".cv_inline_site_id", ".cv_loc", ".cv_linetable",
@@ -2156,6 +2158,8 @@ bool AsmParser::parseStatement(ParseStatementInfo &Info,
       return parseDirectiveLine();
     case DK_LOC:
       return parseDirectiveLoc();
+    case DK_LOC_LABEL:
+      return parseDirectiveLocLabel(IDLoc);
     case DK_STABS:
       return parseDirectiveStabs();
     case DK_CV_FILE:
@@ -3733,6 +3737,19 @@ bool AsmParser::parseDirectiveLoc() {
   return false;
 }
 
+/// parseDirectiveLoc
+/// ::= .loc_label label
+bool AsmParser::parseDirectiveLocLabel(SMLoc DirectiveLoc) {
+  StringRef Name;
+  DirectiveLoc = Lexer.getLoc();
+  if (parseIdentifier(Name))
+    return TokError("expected identifier");
+  if (parseEOL())
+    return true;
+  getStreamer().emitDwarfLocLabelDirective(DirectiveLoc, Name);
+  return false;
+}
+
 /// parseDirectiveStabs
 /// ::= .stabs string, number, number, number
 bool AsmParser::parseDirectiveStabs() {
@@ -5541,6 +5558,7 @@ void AsmParser::initializeDirectiveKindMap() {
   DirectiveKindMap[".file"] = DK_FILE;
   DirectiveKindMap[".line"] = DK_LINE;
   DirectiveKindMap[".loc"] = DK_LOC;
+  DirectiveKindMap[".loc_label"] = DK_LOC_LABEL;
   DirectiveKindMap[".stabs"] = DK_STABS;
   DirectiveKindMap[".cv_file"] = DK_CV_FILE;
   DirectiveKindMap[".cv_func_id"] = DK_CV_FUNC_ID;

diff --git a/llvm/lib/MC/MCStreamer.cpp b/llvm/lib/MC/MCStreamer.cpp
@@ -267,6 +267,12 @@ void MCStreamer::emitDwarfLocDirective(unsigned FileNo, unsigned Line,
                                   Discriminator);
 }
 
+void MCStreamer::emitDwarfLocLabelDirective(SMLoc Loc, StringRef Name) {
+  getContext()
+      .getMCDwarfLineTable(getContext().getDwarfCompileUnitID())
+      .endCurrentSeqAndEmitLineStreamLabel(this, Loc, Name);
+}
+
 MCSymbol *MCStreamer::getDwarfLineTableSymbol(unsigned CUID) {
   MCDwarfLineTable &Table = getContext().getMCDwarfLineTable(CUID);
   if (!Table.getLabel()) {

diff --git a/llvm/test/MC/ELF/debug-loc-label.s b/llvm/test/MC/ELF/debug-loc-label.s
@@ -0,0 +1,101 @@
+// Verify that the .loc_label instruction resets the line sequence and generates
+// the requested label at the correct position in the line stream
+
+// RUN: llvm-mc -filetype obj -triple x86_64 %s -o %t.o
+// RUN: llvm-dwarfdump -v --debug-line %t.o | FileCheck %s --check-prefix=CHECK-LINE-TABLE
+// RUN: llvm-readelf -s %t.o | FileCheck %s --check-prefix=CHECK-SYM
+// RUN: llvm-objdump -s -j .offsets %t.o | FileCheck %s --check-prefix=CHECK-OFFSETS
+
+// RUN: not llvm-mc -filetype obj -triple x86_64 --defsym ERR=1 %s -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR --implicit-check-not=error:
+// RUN: not llvm-mc -filetype obj -triple x86_64 --defsym ERR2=1 %s -o /dev/null 2>&1 | FileCheck %s --check-prefix=ERR2 --implicit-check-not=error:
+
+
+
+# CHECK-LINE-TABLE:                  Address            Line   Column File   ISA Discriminator OpIndex Flags
+# CHECK-LINE-TABLE-NEXT:             ------------------ ------ ------ ------ --- ------------- ------- -------------
+# CHECK-LINE-TABLE-NEXT: 0x00000028: 05 DW_LNS_set_column (1)
+# CHECK-LINE-TABLE-NEXT: 0x0000002a: 00 DW_LNE_set_address (0x0000000000000000)
+# CHECK-LINE-TABLE-NEXT: 0x00000035: 01 DW_LNS_copy
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000000      1      1      1   0             0       0  is_stmt
+# CHECK-LINE-TABLE-NEXT: 0x00000036: 00 DW_LNE_end_sequence
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000000      1      1      1   0             0       0  is_stmt end_sequence
+# CHECK-LINE-TABLE-NEXT: 0x00000039: 05 DW_LNS_set_column (2)
+# CHECK-LINE-TABLE-NEXT: 0x0000003b: 00 DW_LNE_set_address (0x0000000000000008)
+# CHECK-LINE-TABLE-NEXT: 0x00000046: 01 DW_LNS_copy
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000008      1      2      1   0             0       0  is_stmt
+# CHECK-LINE-TABLE-NEXT: 0x00000047: 00 DW_LNE_end_sequence
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000008      1      2      1   0             0       0  is_stmt end_sequence
+# CHECK-LINE-TABLE-NEXT: 0x0000004a: 05 DW_LNS_set_column (3)
+# CHECK-LINE-TABLE-NEXT: 0x0000004c: 00 DW_LNE_set_address (0x0000000000000010)
+# CHECK-LINE-TABLE-NEXT: 0x00000057: 01 DW_LNS_copy
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000010      1      3      1   0             0       0  is_stmt
+# CHECK-LINE-TABLE-NEXT: 0x00000058: 00 DW_LNE_end_sequence
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000010      1      3      1   0             0       0  is_stmt end_sequence
+# CHECK-LINE-TABLE-NEXT: 0x0000005b: 05 DW_LNS_set_column (4)
+# CHECK-LINE-TABLE-NEXT: 0x0000005d: 00 DW_LNE_set_address (0x0000000000000018)
+# CHECK-LINE-TABLE-NEXT: 0x00000068: 01 DW_LNS_copy
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000018      1      4      1   0             0       0  is_stmt
+# CHECK-LINE-TABLE-NEXT: 0x00000069: 05 DW_LNS_set_column (5)
+# CHECK-LINE-TABLE-NEXT: 0x0000006b: 01 DW_LNS_copy
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000018      1      5      1   0             0       0  is_stmt
+# CHECK-LINE-TABLE-NEXT: 0x0000006c: 00 DW_LNE_end_sequence
+# CHECK-LINE-TABLE-NEXT:             0x0000000000000018      1      5      1   0             0       0  is_stmt end_sequence
+
+# CHECK-SYM:      Symbol table '.symtab' contains 9 entries:
+# CHECK-SYM-NEXT:    Num:    Value          Size Type    Bind   Vis       Ndx Name
+# CHECK-SYM-NEXT:      0: 0000000000000000     0 NOTYPE  LOCAL  DEFAULT   UND
+# CHECK-SYM-NEXT:      1: 0000000000000000     0 FILE    LOCAL  DEFAULT   ABS test.c
+# CHECK-SYM-NEXT:      2: 0000000000000000     0 SECTION LOCAL  DEFAULT     2 .text
+# CHECK-SYM-NEXT:      3: 0000000000000039     0 NOTYPE  LOCAL  DEFAULT     3 my_label_02
+# CHECK-SYM-NEXT:      4: 000000000000004a     0 NOTYPE  LOCAL  DEFAULT     3 my_label_03
+# CHECK-SYM-NEXT:      5: 000000000000005b     0 NOTYPE  LOCAL  DEFAULT     3 my_label_04
+# CHECK-SYM-NEXT:      6: 000000000000004a     0 NOTYPE  LOCAL  DEFAULT     3 my_label_03.1
+# CHECK-SYM-NEXT:      7: 000000000000006f     0 NOTYPE  LOCAL  DEFAULT     3 my_label_05
+# CHECK-SYM-NEXT:      8: 0000000000000000     0 FUNC    GLOBAL DEFAULT     2 foo
+
+# CHECK-OFFSETS: 0000 39000000 4a000000 5b000000
+
+	.text
+	.file	"test.c"
+	.globl	foo
+	.align	16, 0x90
+	.type	foo,@function
+foo:
+.Lfunc_begin0:
+	.file	1 "test.c"
+	.cfi_startproc
+	.loc	1 1 1
+	mov     %rax, 0x01
+	.loc_label my_label_02
+	.loc	1 1 2
+	mov     %rax, 0x02
+	.loc	1 1 3
+	.loc_label my_label_03
+	.loc_label my_label_03.1
+	mov     %rax, 0x03
+	.loc	1 1 4
+	.loc_label my_label_04
+	.loc	1 1 5
+.ifdef ERR
+  .loc_label my_label_04
+# ERR: [[#@LINE+1]]:13: error: expected identifier
+  .loc_label
+# ERR: [[#@LINE+1]]:19: error: expected newline
+  .loc_label aaaa bbbb
+.endif
+.ifdef ERR2
+# ERR2: [[#@LINE+1]]:14: error: symbol 'my_label_04' is already defined
+  .loc_label my_label_04
+.endif
+	mov     %rax, 0x04
+	.loc_label my_label_05
+	ret
+	.cfi_endproc
+
+	.section	.debug_line,"",@progbits
+.Lline_table_start0:
+
+	.section	.offsets,"",@progbits
+	.long	my_label_02-.Lline_table_start0
+	.long	my_label_03-.Lline_table_start0
+	.long	my_label_04-.Lline_table_start0