[ELF] --gdb-index: support .debug_loclists

[MaskRay]

Summary

--gdb-index currently crashes when reading a translation unit with
DWARF v5 .debug_loclists . Call stack:

SyntheticSections.cpp GdbIndexSection::create
SyntheticSections.cpp readAddressAreas
DWARFUnit.cpp DWARFUnit::tryExtractDIEsIfNeeded
DWARFListTable.cpp DWARFListTableHeader::extract
...
DWARFDataExtractor.cpp DWARFDataExtractor::getRelocatedValue
lld/ELF/DWARF.cpp LLDDwarfObj<ELFT>::find (sec.sec is nullptr)
...

This patch adds support for .debug_loclists to make DWARFUnit::tryExtractDIEsIfNeeded happy.
Building --gdb-index does not need .debug_loclists

[dblaikie]

I'm surprised that generating gdb-index requires reading any debug locations - is there any chance of refactoring the APIs to make this not needed at all? (I mean, the bug fix is probably still a good thing - but maybe if the libDebugInfoDWARF APIs were changed the StringSwitch could be selective rather than exhaustive, just about the sections needed for gdb-index)

[MaskRay]

In D79061#2009323, @dblaikie wrote:

I'm surprised that generating gdb-index requires reading any debug locations - is there any chance of refactoring the APIs to make this not needed at all? (I mean, the bug fix is probably still a good thing - but maybe if the libDebugInfoDWARF APIs were changed the StringSwitch could be selective rather than exhaustive, just about the sections needed for gdb-index)

An alternative fix is to change https://github.com/llvm/llvm-project/blob/master/llvm/lib/DebugInfo/DWARF/DWARFUnit.cpp#L543 (DWARFUnit::tryExtractDIEsIfNeeded) to optionally decode .debug_loclists, and let --gdb-index not decode .debug_loclists, but I think the complexity may be larger...

[grimar]

LGTM

[dblaikie]

Fair enough - seems like a reasonable fix/consistent with the existing code (are there any other missing sections?) though might be nice to revisit the API at some point. I don't think there should be any reading/need for loc/loclists in LLD's use of DWARF (except the full DWARF-aware linking stuff being prototyped)

[MaskRay]

In D79061#2011271, @dblaikie wrote:

Fair enough - seems like a reasonable fix/consistent with the existing code (are there any other missing sections?) though might be nice to revisit the API at some point. I don't think there should be any reading/need for loc/loclists in LLD's use of DWARF (except the full DWARF-aware linking stuff being prototyped)

Thanks! --gdb-index does not need .debug_loclists . Only the full DWARF aware linking may need it.

I'll update the comment in lld/test/ELF/gdb-index-loclists.s to mention that this is just a regression test.

are there any other missing sections?

Hopefully no :)

[MaskRay]

Update comment

lld/ELF/DWARF.h

 public:
   explicit LLDDwarfObj(ObjFile<ELFT> *obj);
 
   void forEachInfoSections(
       llvm::function_ref<void(const llvm::DWARFSection &)> f) const override {
     f(infoSection);
   }
 
+  const llvm::DWARFSection &getLoclistsSection() const override {
+    return loclistsSection;
+  }
+
   const llvm::DWARFSection &getRangesSection() const override {
     return rangesSection;
   }
 
   const llvm::DWARFSection &getRnglistsSection() const override {
     return rnglistsSection;
   }
 
private:
   template <class RelTy>
   llvm::Optional<llvm::RelocAddrEntry> findAux(const InputSectionBase &sec,
                                                uint64_t pos,
                                                ArrayRef<RelTy> rels) const;
 
   LLDDWARFSection gnuPubnamesSection;
   LLDDWARFSection gnuPubtypesSection;
   LLDDWARFSection infoSection;
+  LLDDWARFSection loclistsSection;
   LLDDWARFSection rangesSection;
   LLDDWARFSection rnglistsSection;
   LLDDWARFSection strOffsetsSection;
   LLDDWARFSection lineSection;
   LLDDWARFSection addrSection;
   StringRef abbrevSection;
   StringRef strSection;
   StringRef lineStrSection;
 };
 
 } // namespace elf
 } // namespace lld
 
 #endif

lld/ELF/DWARF.cpp

if (!sec)
       continue;
 
     if (LLDDWARFSection *m =
             StringSwitch<LLDDWARFSection *>(sec->name)
                 .Case(".debug_addr", &addrSection)
                 .Case(".debug_gnu_pubnames", &gnuPubnamesSection)
                 .Case(".debug_gnu_pubtypes", &gnuPubtypesSection)
                 .Case(".debug_info", &infoSection)
+                .Case(".debug_loclists", &loclistsSection)
                 .Case(".debug_ranges", &rangesSection)
                 .Case(".debug_rnglists", &rnglistsSection)
                 .Case(".debug_str_offsets", &strOffsetsSection)
                 .Case(".debug_line", &lineSection)
                 .Default(nullptr)) {
       m->Data = toStringRef(sec->data());
       m->sec = sec;
       continue;

lld/test/ELF/gdb-index-loclists.s

+# REQUIRES: x86
+## Regression test that we don't crash on DWARF v5 .debug_loclists
+
+# RUN: llvm-mc -filetype=obj -triple=x86_64 %s -o %t.o
+# RUN: ld.lld --gdb-index %t.o -o /dev/null
+
+.section .debug_abbrev,"",@progbits
+  .byte 1            # Abbreviation Code
+  .byte 17           # DW_TAG_compile_unit
+  .byte 0            # DW_CHILDREN_no
+  .ascii "\214\001"  # DW_AT_loclists_base
+  .byte 23           # DW_FORM_sec_offset
+  .byte 0            # EOM(1)
+  .byte 0            # EOM(2)
+  .byte 0
+
+.section .debug_info,"",@progbits
+.Lcu_begin0:
+  .long .Lcu_end0-.Lcu_begin0-4  # Length of Unit
+  .short 5                       # DWARF version number
+  .byte  1                       # DWARF Unit Type
+  .byte  8                       # Address Size
+  .long  0                       # Offset Into Abbrev. Section
+  .byte  1                       # Abbrev [1] DW_TAG_compile_unit
+  .long  .Lloclists_table_base0  # DW_AT_loclists_base
+.Lcu_end0:
+
+.section .debug_loclists,"",@progbits
+  .long .Ldebug_loclist_table_end0-.Ldebug_loclist_table_start0 # Length
+.Ldebug_loclist_table_start0:
+  .short 5                # Version
+  .byte  8                # Address size
+  .byte  0                # Segment selector size
+  .long  0                # Offset entry count
+.Lloclists_table_base0:
+  .byte  0                # DW_LLE_end_of_list
+.Ldebug_loclist_table_end0: