[llvm] r237733 - [DWARF parser] Make DWARF parser more robust against missing compile/type units.

Frédéric Riss friss at apple.com
Thu May 28 17:35:33 PDT 2015


> On May 28, 2015, at 5:01 PM, Alexey Samsonov <vonosmas at gmail.com> wrote:
> 
> On Tue, May 26, 2015 at 1:35 PM, Filipe Cabecinhas <filcab at gmail.com <mailto:filcab at gmail.com>> wrote:
> Throwing afl (or libFuzzer) at llvm-dwarfdump might do a lot. :)
> 
> Sure it does. We should first make significant improvements to LLVMObject library so that it doesn't crash with report_fatal_error() if it's given a
> broken executable. Luckily, we have all this std::error_code machinery. However, we use it far less frequently than we should. Trying to tackle this task
> resulted in http://reviews.llvm.org/D10111 <https://urldefense.proofpoint.com/v2/url?u=http-3A__reviews.llvm.org_D10111&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=wDepzKWNQSGQzVSKQ_yREiaBuagKyqDqMiMU_P8ybDw&s=scjjp6gwisgoQ8d0hGdM1Fz7HZ-O9fxXpOknXN0kXEY&e=>.

Hey, this is great! Where libObject might call report_fatal_error too much, I’m pretty sure libDebugInfoDWARF make use of too much asserts driven by input data rather than returning errors. It’s great that you tackle this. Do not hesitate to file bugs for the issues you find, I’m willing to help in fixing these!

Fred

> 
> 
> On Tuesday, May 26, 2015, Alexey Samsonov <vonosmas at gmail.com <mailto:vonosmas at gmail.com>> wrote:
> Fair enough. I will add such an input then. I guess there will be even more once of those once/if I write llvm-dwarfdump-fuzzer :)
> 
> On Tue, May 26, 2015 at 1:30 PM, Filipe Cabecinhas <filcab at gmail.com <>> wrote:
> If we want to be sure we're handling the errors correctly, we want invalid input. There's a bunch of them for macho (tests/Object/Inputs/macho{,64}-invalid-*) and bitcode (tests/Bitcode/Inputs/invalid-*).
> 
>   Filipe
> 
> 
> On Tuesday, May 26, 2015, Eric Christopher <echristo at gmail.com <>> wrote:
> Could have a binary input?
> 
> -eric
> 
> On Tue, May 26, 2015 at 12:46 PM Alexey Samsonov <vonosmas at gmail.com <>> wrote:
> At this stage that would require producing a broken DWARF file, possibly by manually flipping bits in the output of clang/gcc. Do we really want this in the tree?
> 
> On Tue, May 26, 2015 at 7:06 AM, Rafael Espíndola <rafael.espindola at gmail.com <>> wrote:
> testcase?
> 
> On 19 May 2015 at 17:54, Alexey Samsonov <vonosmas at gmail.com <>> wrote:
> > Author: samsonov
> > Date: Tue May 19 16:54:32 2015
> > New Revision: 237733
> >
> > URL: http://llvm.org/viewvc/llvm-project?rev=237733&view=rev <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject-3Frev-3D237733-26view-3Drev&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=WJCv3-sOyU-kNy7q14dfet81O5FN3uwSBCHo7z8kJxI&e=>
> > Log:
> > [DWARF parser] Make DWARF parser more robust against missing compile/type units.
> >
> > DWARF standard claims that each compilation/type unit header in
> > .debug_info/.debug_types section must be followed by corresponding
> > compile/type unit DIE, possibly with its children. Two situations
> > are possible:
> >
> >  * compile/type unit DIE is missing because DWARF producer failed to
> >    emit it.
> >  * DWARF parser failed to parse unit DIE correctly, for instance if it
> >    contains some unsupported attributes (see r237721, for instance).
> >
> > In either of these cases, the library, and the tools that use it
> > (llvm-dwarfdump, llvm-symbolizer) should not crash. Insert appropriate
> > checks to protect against this.
> >
> > Modified:
> >     llvm/trunk/include/llvm/DebugInfo/DWARF/DWARFUnit.h
> >     llvm/trunk/lib/DebugInfo/DWARF/DWARFCompileUnit.cpp
> >     llvm/trunk/lib/DebugInfo/DWARF/DWARFContext.cpp
> >     llvm/trunk/lib/DebugInfo/DWARF/DWARFTypeUnit.cpp
> >     llvm/trunk/lib/DebugInfo/DWARF/DWARFUnit.cpp
> >     llvm/trunk/tools/dsymutil/DwarfLinker.cpp
> >
> > Modified: llvm/trunk/include/llvm/DebugInfo/DWARF/DWARFUnit.h
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/DebugInfo/DWARF/DWARFUnit.h?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_include_llvm_DebugInfo_DWARF_DWARFUnit.h-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=UdM4yDUJGI81iOHT3gWMDhnYdYi5sDeMyoD249yK-9Y&e=>
> > ==============================================================================
> > --- llvm/trunk/include/llvm/DebugInfo/DWARF/DWARFUnit.h (original)
> > +++ llvm/trunk/include/llvm/DebugInfo/DWARF/DWARFUnit.h Tue May 19 16:54:32 2015
> > @@ -195,9 +195,8 @@ public:
> >      BaseAddr = base_addr;
> >    }
> >
> > -  const DWARFDebugInfoEntryMinimal *
> > -  getCompileUnitDIE(bool extract_cu_die_only = true) {
> > -    extractDIEsIfNeeded(extract_cu_die_only);
> > +  const DWARFDebugInfoEntryMinimal *getUnitDIE(bool ExtractUnitDIEOnly = true) {
> > +    extractDIEsIfNeeded(ExtractUnitDIEOnly);
> >      return DieArray.empty() ? nullptr : &DieArray[0];
> >    }
> >
> > @@ -226,8 +225,7 @@ public:
> >    /// It is illegal to call this method with a DIE that hasn't be
> >    /// created by this unit. In other word, it's illegal to call this
> >    /// method on a DIE that isn't accessible by following
> > -  /// children/sibling links starting from this unit's
> > -  /// getCompileUnitDIE().
> > +  /// children/sibling links starting from this unit's getUnitDIE().
> >    uint32_t getDIEIndex(const DWARFDebugInfoEntryMinimal *DIE) {
> >      assert(!DieArray.empty() && DIE >= &DieArray[0] &&
> >             DIE < &DieArray[0] + DieArray.size());
> >
> > Modified: llvm/trunk/lib/DebugInfo/DWARF/DWARFCompileUnit.cpp
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/DebugInfo/DWARF/DWARFCompileUnit.cpp?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_lib_DebugInfo_DWARF_DWARFCompileUnit.cpp-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=q9Qw5_oJ_WYHkzMoXhKy_3DjR5vajUv8gQ4HRKAbXj8&e=>
> > ==============================================================================
> > --- llvm/trunk/lib/DebugInfo/DWARF/DWARFCompileUnit.cpp (original)
> > +++ llvm/trunk/lib/DebugInfo/DWARF/DWARFCompileUnit.cpp Tue May 19 16:54:32 2015
> > @@ -22,9 +22,10 @@ void DWARFCompileUnit::dump(raw_ostream
> >       << " (next unit at " << format("0x%08x", getNextUnitOffset())
> >       << ")\n";
> >
> > -  const DWARFDebugInfoEntryMinimal *CU = getCompileUnitDIE(false);
> > -  assert(CU && "Null Compile Unit?");
> > -  CU->dump(OS, this, -1U);
> > +  if (const DWARFDebugInfoEntryMinimal *CU = getUnitDIE(false))
> > +    CU->dump(OS, this, -1U);
> > +  else
> > +    OS << "<compile unit can't be parsed!>\n\n";
> >  }
> >
> >  // VTable anchor.
> >
> > Modified: llvm/trunk/lib/DebugInfo/DWARF/DWARFContext.cpp
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/DebugInfo/DWARF/DWARFContext.cpp?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_lib_DebugInfo_DWARF_DWARFContext.cpp-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=HJVL60uJ4lybBLyg2Qcq9DDCYOza6L53QTPT0GfYrpU&e=>
> > ==============================================================================
> > --- llvm/trunk/lib/DebugInfo/DWARF/DWARFContext.cpp (original)
> > +++ llvm/trunk/lib/DebugInfo/DWARF/DWARFContext.cpp Tue May 19 16:54:32 2015
> > @@ -140,9 +140,11 @@ void DWARFContext::dump(raw_ostream &OS,
> >      OS << "\n.debug_line contents:\n";
> >      for (const auto &CU : compile_units()) {
> >        savedAddressByteSize = CU->getAddressByteSize();
> > -      unsigned stmtOffset =
> > -          CU->getCompileUnitDIE()->getAttributeValueAsSectionOffset(
> > -              CU.get(), DW_AT_stmt_list, -1U);
> > +      const auto *CUDIE = CU->getUnitDIE();
> > +      if (CUDIE == nullptr)
> > +        continue;
> > +      unsigned stmtOffset = CUDIE->getAttributeValueAsSectionOffset(
> > +          CU.get(), DW_AT_stmt_list, -1U);
> >        if (stmtOffset != -1U) {
> >          DataExtractor lineData(getLineSection().Data, isLittleEndian(),
> >                                 savedAddressByteSize);
> > @@ -321,13 +323,14 @@ const DWARFDebugFrame *DWARFContext::get
> >  }
> >
> >  const DWARFLineTable *
> > -DWARFContext::getLineTableForUnit(DWARFUnit *cu) {
> > +DWARFContext::getLineTableForUnit(DWARFUnit *U) {
> >    if (!Line)
> >      Line.reset(new DWARFDebugLine(&getLineSection().Relocs));
> > -
> > +  const auto *UnitDIE = U->getUnitDIE();
> > +  if (UnitDIE == nullptr)
> > +    return nullptr;
> >    unsigned stmtOffset =
> > -      cu->getCompileUnitDIE()->getAttributeValueAsSectionOffset(
> > -          cu, DW_AT_stmt_list, -1U);
> > +      UnitDIE->getAttributeValueAsSectionOffset(U, DW_AT_stmt_list, -1U);
> >    if (stmtOffset == -1U)
> >      return nullptr; // No line table for this compile unit.
> >
> > @@ -337,7 +340,7 @@ DWARFContext::getLineTableForUnit(DWARFU
> >
> >    // We have to parse it first.
> >    DataExtractor lineData(getLineSection().Data, isLittleEndian(),
> > -                         cu->getAddressByteSize());
> > +                         U->getAddressByteSize());
> >    return Line->getOrParseLineTable(lineData, stmtOffset);
> >  }
> >
> >
> > Modified: llvm/trunk/lib/DebugInfo/DWARF/DWARFTypeUnit.cpp
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/DebugInfo/DWARF/DWARFTypeUnit.cpp?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_lib_DebugInfo_DWARF_DWARFTypeUnit.cpp-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=CKRMfE-kGYuRa3LpYSaggn0lntWaEmnHl811uBLEi58&e=>
> > ==============================================================================
> > --- llvm/trunk/lib/DebugInfo/DWARF/DWARFTypeUnit.cpp (original)
> > +++ llvm/trunk/lib/DebugInfo/DWARF/DWARFTypeUnit.cpp Tue May 19 16:54:32 2015
> > @@ -33,7 +33,8 @@ void DWARFTypeUnit::dump(raw_ostream &OS
> >       << " (next unit at " << format("0x%08x", getNextUnitOffset())
> >       << ")\n";
> >
> > -  const DWARFDebugInfoEntryMinimal *CU = getCompileUnitDIE(false);
> > -  assert(CU && "Null Compile Unit?");
> > -  CU->dump(OS, this, -1U);
> > +  if (const DWARFDebugInfoEntryMinimal *TU = getUnitDIE(false))
> > +    TU->dump(OS, this, -1U);
> > +  else
> > +    OS << "<type unit can't be parsed!>\n\n";
> >  }
> >
> > Modified: llvm/trunk/lib/DebugInfo/DWARF/DWARFUnit.cpp
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/DebugInfo/DWARF/DWARFUnit.cpp?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_lib_DebugInfo_DWARF_DWARFUnit.cpp-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=ONzxT5tM7jRAOqxX9shQH-CMGyvjBMnqGzlxdK0hAz0&e=>
> > ==============================================================================
> > --- llvm/trunk/lib/DebugInfo/DWARF/DWARFUnit.cpp (original)
> > +++ llvm/trunk/lib/DebugInfo/DWARF/DWARFUnit.cpp Tue May 19 16:54:32 2015
> > @@ -310,8 +310,11 @@ void DWARFUnit::clearDIEs(bool KeepCUDie
> >  }
> >
> >  void DWARFUnit::collectAddressRanges(DWARFAddressRangesVector &CURanges) {
> > -  // First, check if CU DIE describes address ranges for the unit.
> > -  const auto &CUDIERanges = getCompileUnitDIE()->getAddressRanges(this);
> > +  const auto *U = getUnitDIE();
> > +  if (U == nullptr)
> > +    return;
> > +  // First, check if unit DIE describes address ranges for the whole unit.
> > +  const auto &CUDIERanges = U->getAddressRanges(this);
> >    if (!CUDIERanges.empty()) {
> >      CURanges.insert(CURanges.end(), CUDIERanges.begin(), CUDIERanges.end());
> >      return;
> >
> > Modified: llvm/trunk/tools/dsymutil/DwarfLinker.cpp
> > URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/tools/dsymutil/DwarfLinker.cpp?rev=237733&r1=237732&r2=237733&view=diff <https://urldefense.proofpoint.com/v2/url?u=http-3A__llvm.org_viewvc_llvm-2Dproject_llvm_trunk_tools_dsymutil_DwarfLinker.cpp-3Frev-3D237733-26r1-3D237732-26r2-3D237733-26view-3Ddiff&d=AwMFaQ&c=8hUWFZcy2Z-Za5rBPlktOQ&r=mQ4LZ2PUj9hpadE3cDHZnIdEwhEBrbAstXeMaFoB9tg&m=i0RikhEETYZDN7aNgdqzao5KZVbb44wCscGgxV85vOo&s=lthJ-pbHL4ZQvFR0n9YmO2E7a1DFd4N2iMCAXNF-VrM&e=>
> > ==============================================================================
> > --- llvm/trunk/tools/dsymutil/DwarfLinker.cpp (original)
> > +++ llvm/trunk/tools/dsymutil/DwarfLinker.cpp Tue May 19 16:54:32 2015
> > @@ -729,7 +729,7 @@ void DwarfStreamer::emitLocationsForUnit
> >    const DWARFSection &InputSec = Dwarf.getLocSection();
> >    DataExtractor Data(InputSec.Data, Dwarf.isLittleEndian(), AddressSize);
> >    DWARFUnit &OrigUnit = Unit.getOrigUnit();
> > -  const auto *OrigUnitDie = OrigUnit.getCompileUnitDIE(false);
> > +  const auto *OrigUnitDie = OrigUnit.getUnitDIE(false);
> >    int64_t UnitPcOffset = 0;
> >    uint64_t OrigLowPc = OrigUnitDie->getAttributeValueAsAddress(
> >        &OrigUnit, dwarf::DW_AT_low_pc, -1ULL);
> > @@ -2203,7 +2203,7 @@ void DwarfLinker::patchRangesForUnit(con
> >                                 OrigDwarf.isLittleEndian(), AddressSize);
> >    auto InvalidRange = FunctionRanges.end(), CurrRange = InvalidRange;
> >    DWARFUnit &OrigUnit = Unit.getOrigUnit();
> > -  const auto *OrigUnitDie = OrigUnit.getCompileUnitDIE(false);
> > +  const auto *OrigUnitDie = OrigUnit.getUnitDIE(false);
> >    uint64_t OrigLowPc = OrigUnitDie->getAttributeValueAsAddress(
> >        &OrigUnit, dwarf::DW_AT_low_pc, -1ULL);
> >    // Ranges addresses are based on the unit's low_pc. Compute the
> > @@ -2287,7 +2287,7 @@ static void insertLineSequence(std::vect
> >  void DwarfLinker::patchLineTableForUnit(CompileUnit &Unit,
> >                                          DWARFContext &OrigDwarf) {
> >    const DWARFDebugInfoEntryMinimal *CUDie =
> > -      Unit.getOrigUnit().getCompileUnitDIE();
> > +      Unit.getOrigUnit().getUnitDIE();
> >    uint64_t StmtList = CUDie->getAttributeValueAsSectionOffset(
> >        &Unit.getOrigUnit(), dwarf::DW_AT_stmt_list, -1ULL);
> >    if (StmtList == -1ULL)
> > @@ -2461,7 +2461,7 @@ bool DwarfLinker::link(const DebugMap &M
> >      // In a first phase, just read in the debug info and store the DIE
> >      // parent links that we will use during the next phase.
> >      for (const auto &CU : DwarfContext.compile_units()) {
> > -      auto *CUDie = CU->getCompileUnitDIE(false);
> > +      auto *CUDie = CU->getUnitDIE(false);
> >        if (Options.Verbose) {
> >          outs() << "Input compilation unit:";
> >          CUDie->dump(outs(), CU.get(), 0);
> > @@ -2476,7 +2476,7 @@ bool DwarfLinker::link(const DebugMap &M
> >      // references require the ParentIdx to be setup for every CU in
> >      // the object file before calling this.
> >      for (auto &CurrentUnit : Units)
> > -      lookForDIEsToKeep(*CurrentUnit.getOrigUnit().getCompileUnitDIE(), *Obj,
> > +      lookForDIEsToKeep(*CurrentUnit.getOrigUnit().getUnitDIE(), *Obj,
> >                          CurrentUnit, 0);
> >
> >      // The calls to applyValidRelocs inside cloneDIE will walk the
> > @@ -2489,7 +2489,7 @@ bool DwarfLinker::link(const DebugMap &M
> >      // to clone/emit.
> >      if (!ValidRelocs.empty())
> >        for (auto &CurrentUnit : Units) {
> > -        const auto *InputDIE = CurrentUnit.getOrigUnit().getCompileUnitDIE();
> > +        const auto *InputDIE = CurrentUnit.getOrigUnit().getUnitDIE();
> >          CurrentUnit.setStartOffset(OutputDebugInfoSize);
> >          DIE *OutputDIE = cloneDIE(*InputDIE, CurrentUnit, 0 /* PCOffset */,
> >                                    11 /* Unit Header size */);
> >
> >
> > _______________________________________________
> > llvm-commits mailing list
> > llvm-commits at cs.uiuc.edu <>
> > http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
> 
> 
> 
> -- 
> Alexey Samsonov
> vonosmas at gmail.com <>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu <>
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
> 
> 
> -- 
>   F
> 
> 
> 
> 
> -- 
> Alexey Samsonov
> vonosmas at gmail.com <>
> 
> -- 
>   F
> 
> 
> 
> 
> -- 
> Alexey Samsonov
> vonosmas at gmail.com <mailto:vonosmas at gmail.com>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu <mailto:llvm-commits at cs.uiuc.edu>
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150528/8e0021ab/attachment.html>


More information about the llvm-commits mailing list