DWARFDebugLoclists: Move to a incremental parsing model

Summary:
This patch stems from the discussion D68270 (including some offline
talks). The idea is to provide an "incremental" api for parsing ___location
lists, which will avoid caching or materializing parsed data. An
additional goal is to provide a high level ___location list api, which
abstracts the differences between different encoding schemes, and can be
used by users which don't care about those (such as LLDB).

This patch implements the first part. It implements a call-back based
"visitLocationList" api. This function parses a single ___location list,
calling a user-specified callback for each entry. This is going to be
the base api, which other ___location list functions (right now, just the
dumping code) are going to be based on.

Future patches will do something similar for the v4 ___location lists, and
add a mechanism to translate raw entries into concrete address ranges.

Reviewers: dblaikie, probinson, JDevlieghere, aprantl, SouraVX

Subscribers: hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D69672

Author: labath

llvm/include/llvm/DebugInfo/DWARF/DWARFContext.h

@@ -75,7 +75,6 @@ class DWARFContext : public DIContext {
 
   DWARFUnitVector DWOUnits;
   std::unique_ptr<DWARFDebugAbbrev> AbbrevDWO;
-  std::unique_ptr<DWARFDebugLoclists> LocDWO;
 
   /// The maximum DWARF version of all units.
   unsigned MaxVersion = 0;
@@ -260,9 +259,6 @@ class DWARFContext : public DIContext {
   /// Get a pointer to the parsed dwo abbreviations object.
   const DWARFDebugAbbrev *getDebugAbbrevDWO();
 
-  /// Get a pointer to the parsed DebugLoc object.
-  const DWARFDebugLoclists *getDebugLocDWO();
-
   /// Get a pointer to the parsed DebugAranges object.
   const DWARFDebugAranges *getDebugAranges();
 

llvm/include/llvm/DebugInfo/DWARF/DWARFDebugLoc.h

@@ -76,6 +76,9 @@ class DWARFDebugLoc {
 
 class DWARFDebugLoclists {
 public:
+  // Unconstructible.
+  DWARFDebugLoclists() = delete;
+
   struct Entry {
     uint8_t Kind;
     uint64_t Offset;
@@ -87,35 +90,30 @@ class DWARFDebugLoclists {
               DIDumpOptions DumpOpts, unsigned Indent, size_t MaxEncodingStringLength) const;
   };
 
-  struct LocationList {
-    uint64_t Offset;
-    SmallVector<Entry, 2> Entries;
-    void dump(raw_ostream &OS, uint64_t BaseAddr, bool IsLittleEndian,
-              unsigned AddressSize, const MCRegisterInfo *RegInfo,
-              DWARFUnit *U, DIDumpOptions DumpOpts, unsigned Indent) const;
-  };
-
-private:
-  using LocationLists = SmallVector<LocationList, 4>;
-
-  LocationLists Locations;
-
-  unsigned AddressSize;
-
-  bool IsLittleEndian;
-
-public:
-  void parse(const DWARFDataExtractor &data, uint64_t Offset,
-             uint64_t EndOffset, uint16_t Version);
-  void dump(raw_ostream &OS, uint64_t BaseAddr, const MCRegisterInfo *RegInfo,
-            DIDumpOptions DumpOpts, Optional<uint64_t> Offset) const;
-
-  /// Return the ___location list at the given offset or nullptr.
-  LocationList const *getLocationListAtOffset(uint64_t Offset) const;
-
-  static Expected<LocationList>
-  parseOneLocationList(const DWARFDataExtractor &Data, uint64_t *Offset,
-                       unsigned Version);
+  /// Call the user-provided callback for each entry (including the end-of-list
+  /// entry) in the ___location list starting at \p Offset. The callback can return
+  /// false to terminate the iteration early. Returns an error if it was unable
+  /// to parse the entire ___location list correctly. Upon successful termination
+  /// \p Offset will be updated point past the end of the list.
+  static Error visitLocationList(const DWARFDataExtractor &Data,
+                                 uint64_t *Offset, uint16_t Version,
+                                 llvm::function_ref<bool(const Entry &)> F);
+
+  /// Dump the ___location list at the given \p Offset. The function returns true
+  /// iff it has successfully reched the end of the list. This means that one
+  /// can attempt to parse another list after the current one (\p Offset will be
+  /// updated to point past the end of the current list).
+  static bool dumpLocationList(const DWARFDataExtractor &Data, uint64_t *Offset,
+                               uint16_t Version, raw_ostream &OS,
+                               uint64_t BaseAddr, const MCRegisterInfo *MRI,
+                               DWARFUnit *U, DIDumpOptions DumpOpts,
+                               unsigned Indent);
+
+  /// Dump all ___location lists within the given range.
+  static void dumpRange(const DWARFDataExtractor &Data, uint64_t StartOffset,
+                        uint64_t Size, uint16_t Version, raw_ostream &OS,
+                        uint64_t BaseAddr, const MCRegisterInfo *MRI,
+                        DIDumpOptions DumpOpts);
 };
 
 } // end namespace llvm

llvm/lib/DebugInfo/DWARF/DWARFContext.cpp

@@ -300,11 +300,21 @@ static void dumpLoclistsSection(raw_ostream &OS, DIDumpOptions DumpOpts,
 
     Header.dump(OS, DumpOpts);
 
-    DWARFDebugLoclists Loclists;
     uint64_t EndOffset = Header.length() + Header.getHeaderOffset();
     Data.setAddressSize(Header.getAddrSize());
-    Loclists.parse(Data, Offset, EndOffset, Header.getVersion());
-    Loclists.dump(OS, 0, MRI, DumpOpts, DumpOffset);
+    if (DumpOffset) {
+      if (DumpOffset >= Offset && DumpOffset < EndOffset) {
+        Offset = *DumpOffset;
+        DWARFDebugLoclists::dumpLocationList(Data, &Offset, Header.getVersion(),
+                                             OS, /*BaseAddr=*/0, MRI, nullptr,
+                                             DumpOpts, /*Indent=*/0);
+        OS << "\n";
+        return;
+      }
+    } else {
+      DWARFDebugLoclists::dumpRange(Data, Offset, EndOffset - Offset,
+                                    Header.getVersion(), OS, 0, MRI, DumpOpts);
+    }
     Offset = EndOffset;
   }
 }
@@ -393,7 +403,19 @@ void DWARFContext::dump(
   if (const auto *Off =
           shouldDump(ExplicitDWO, ".debug_loc.dwo", DIDT_ID_DebugLoc,
                      DObj->getLocDWOSection().Data)) {
-    getDebugLocDWO()->dump(OS, 0, getRegisterInfo(), DumpOpts, *Off);
+    DWARFDataExtractor Data(*DObj, DObj->getLocDWOSection(), isLittleEndian(),
+                            4);
+    if (*Off) {
+      uint64_t Offset = **Off;
+      DWARFDebugLoclists::dumpLocationList(Data, &Offset, /*Version=*/4, OS,
+                                           /*BaseAddr=*/0, getRegisterInfo(),
+                                           nullptr, DumpOpts, /*Indent=*/0);
+      OS << "\n";
+    } else {
+      DWARFDebugLoclists::dumpRange(Data, 0, Data.getData().size(),
+                                    /*Version=*/4, OS, /*BaseAddr=*/0,
+                                    getRegisterInfo(), DumpOpts);
+    }
   }
 
   if (const auto *Off = shouldDump(Explicit, ".debug_frame", DIDT_ID_DebugFrame,
@@ -724,23 +746,6 @@ const DWARFDebugLoc *DWARFContext::getDebugLoc() {
   return Loc.get();
 }
 
-const DWARFDebugLoclists *DWARFContext::getDebugLocDWO() {
-  if (LocDWO)
-    return LocDWO.get();
-
-  LocDWO.reset(new DWARFDebugLoclists());
-  // Assume all compile units have the same address byte size.
-  // FIXME: We don't need AddressSize for split DWARF since relocatable
-  // addresses cannot appear there. At the moment DWARFExpression requires it.
-  DWARFDataExtractor LocData(*DObj, DObj->getLocDWOSection(), isLittleEndian(),
-                             4);
-  // Use version 4. DWO does not support the DWARF v5 .debug_loclists yet and
-  // that means we are parsing the new style .debug_loc (pre-standatized version
-  // of the .debug_loclists).
-  LocDWO->parse(LocData, 0, LocData.getData().size(), 4 /* Version */);
-  return LocDWO.get();
-}
-
 const DWARFDebugAranges *DWARFContext::getDebugAranges() {
   if (Aranges)
     return Aranges.get();

llvm/lib/DebugInfo/DWARF/DWARFDebugLoc.cpp

@@ -139,19 +139,19 @@ void DWARFDebugLoc::parse(const DWARFDataExtractor &data) {
   }
 }
 
-Expected<DWARFDebugLoclists::LocationList>
-DWARFDebugLoclists::parseOneLocationList(const DWARFDataExtractor &Data,
-                                         uint64_t *Offset, unsigned Version) {
-  LocationList LL;
-  LL.Offset = *Offset;
-  DataExtractor::Cursor C(*Offset);
+Error DWARFDebugLoclists::visitLocationList(
+    const DWARFDataExtractor &Data, uint64_t *Offset, uint16_t Version,
+    llvm::function_ref<bool(const Entry &)> F) {
 
-  // dwarf::DW_LLE_end_of_list_entry is 0 and indicates the end of the list.
-  while (auto Kind = Data.getU8(C)) {
+  DataExtractor::Cursor C(*Offset);
+  bool Continue = true;
+  while (Continue) {
     Entry E;
-    E.Kind = Kind;
-    E.Offset = C.tell() - 1;
-    switch (Kind) {
+    E.Offset = C.tell();
+    E.Kind = Data.getU8(C);
+    switch (E.Kind) {
+    case dwarf::DW_LLE_end_of_list:
+      break;
     case dwarf::DW_LLE_base_addressx:
       E.Value0 = Data.getULEB128(C);
       break;
@@ -164,64 +164,69 @@ DWARFDebugLoclists::parseOneLocationList(const DWARFDataExtractor &Data,
       else
         E.Value1 = Data.getULEB128(C);
       break;
-    case dwarf::DW_LLE_start_length:
-      E.Value0 = Data.getRelocatedAddress(C);
-      E.Value1 = Data.getULEB128(C);
-      break;
     case dwarf::DW_LLE_offset_pair:
       E.Value0 = Data.getULEB128(C);
       E.Value1 = Data.getULEB128(C);
       break;
     case dwarf::DW_LLE_base_address:
       E.Value0 = Data.getRelocatedAddress(C);
       break;
+    case dwarf::DW_LLE_start_length:
+      E.Value0 = Data.getRelocatedAddress(C);
+      E.Value1 = Data.getULEB128(C);
+      break;
+    case dwarf::DW_LLE_startx_endx:
+    case dwarf::DW_LLE_default_location:
+    case dwarf::DW_LLE_start_end:
     default:
       cantFail(C.takeError());
       return createStringError(errc::illegal_byte_sequence,
-                               "LLE of kind %x not supported", (int)Kind);
+                               "LLE of kind %x not supported", (int)E.Kind);
     }
 
-    if (Kind != dwarf::DW_LLE_base_address &&
-        Kind != dwarf::DW_LLE_base_addressx) {
+    if (E.Kind != dwarf::DW_LLE_base_address &&
+        E.Kind != dwarf::DW_LLE_base_addressx &&
+        E.Kind != dwarf::DW_LLE_end_of_list) {
       unsigned Bytes = Version >= 5 ? Data.getULEB128(C) : Data.getU16(C);
       // A single ___location description describing the ___location of the object...
       Data.getU8(C, E.Loc, Bytes);
     }
 
-    LL.Entries.push_back(std::move(E));
+    if (!C)
+      return C.takeError();
+    Continue = F(E) && E.Kind != dwarf::DW_LLE_end_of_list;
   }
-  if (Error Err = C.takeError())
-    return std::move(Err);
-  Entry E;
-  E.Kind = dwarf::DW_LLE_end_of_list;
-  E.Offset = C.tell() - 1;
-  LL.Entries.push_back(E);
   *Offset = C.tell();
-  return LL;
+  return Error::success();
 }
 
-void DWARFDebugLoclists::parse(const DWARFDataExtractor &data, uint64_t Offset,
-                               uint64_t EndOffset, uint16_t Version) {
-  IsLittleEndian = data.isLittleEndian();
-  AddressSize = data.getAddressSize();
-
-  while (Offset < EndOffset) {
-    if (auto LL = parseOneLocationList(data, &Offset, Version))
-      Locations.push_back(std::move(*LL));
-    else {
-      logAllUnhandledErrors(LL.takeError(), WithColor::error());
-      return;
-    }
+bool DWARFDebugLoclists::dumpLocationList(const DWARFDataExtractor &Data,
+                                          uint64_t *Offset, uint16_t Version,
+                                          raw_ostream &OS, uint64_t BaseAddr,
+                                          const MCRegisterInfo *MRI,
+                                          DWARFUnit *U, DIDumpOptions DumpOpts,
+                                          unsigned Indent) {
+  size_t MaxEncodingStringLength = 0;
+  if (DumpOpts.Verbose) {
+#define HANDLE_DW_LLE(ID, NAME)                                                \
+  MaxEncodingStringLength = std::max(MaxEncodingStringLength,                  \
+                                     dwarf::LocListEncodingString(ID).size());
+#include "llvm/BinaryFormat/Dwarf.def"
   }
-}
 
-DWARFDebugLoclists::LocationList const *
-DWARFDebugLoclists::getLocationListAtOffset(uint64_t Offset) const {
-  auto It = partition_point(
-      Locations, [=](const LocationList &L) { return L.Offset < Offset; });
-  if (It != Locations.end() && It->Offset == Offset)
-    return &(*It);
-  return nullptr;
+  OS << format("0x%8.8" PRIx64 ": ", *Offset);
+  Error E = visitLocationList(Data, Offset, Version, [&](const Entry &E) {
+    E.dump(OS, BaseAddr, Data.isLittleEndian(), Data.getAddressSize(), MRI, U,
+           DumpOpts, Indent, MaxEncodingStringLength);
+    return true;
+  });
+  if (E) {
+    OS << "\n";
+    OS.indent(Indent);
+    OS << "error: " << toString(std::move(E));
+    return false;
+  }
+  return true;
 }
 
 void DWARFDebugLoclists::Entry::dump(raw_ostream &OS, uint64_t &BaseAddr,
@@ -297,44 +302,25 @@ void DWARFDebugLoclists::Entry::dump(raw_ostream &OS, uint64_t &BaseAddr,
 
   dumpExpression(OS, Loc, IsLittleEndian, AddressSize, MRI, U);
 }
-void DWARFDebugLoclists::LocationList::dump(raw_ostream &OS, uint64_t BaseAddr,
-                                            bool IsLittleEndian,
-                                            unsigned AddressSize,
-                                            const MCRegisterInfo *MRI,
-                                            DWARFUnit *U,
-                                            DIDumpOptions DumpOpts, 
-                                            unsigned Indent) const {
-  size_t MaxEncodingStringLength = 0;
-  if (DumpOpts.Verbose)
-    for (const auto &Entry : Entries)
-      MaxEncodingStringLength =
-          std::max(MaxEncodingStringLength,
-                   dwarf::LocListEncodingString(Entry.Kind).size());
-
-  for (const Entry &E : Entries)
-    E.dump(OS, BaseAddr, IsLittleEndian, AddressSize, MRI, U, DumpOpts, Indent,
-           MaxEncodingStringLength);
-}
-
-void DWARFDebugLoclists::dump(raw_ostream &OS, uint64_t BaseAddr,
-                              const MCRegisterInfo *MRI, DIDumpOptions DumpOpts, 
-                              Optional<uint64_t> Offset) const {
-  auto DumpLocationList = [&](const LocationList &L) {
-    OS << format("0x%8.8" PRIx64 ": ", L.Offset);
-    L.dump(OS, BaseAddr, IsLittleEndian, AddressSize, MRI, nullptr, DumpOpts,
-           /*Indent=*/12);
-    OS << "\n";
-  };
 
-  if (Offset) {
-    if (auto *L = getLocationListAtOffset(*Offset))
-      DumpLocationList(*L);
+void DWARFDebugLoclists::dumpRange(const DWARFDataExtractor &Data,
+                                   uint64_t StartOffset, uint64_t Size,
+                                   uint16_t Version, raw_ostream &OS,
+                                   uint64_t BaseAddr, const MCRegisterInfo *MRI,
+                                   DIDumpOptions DumpOpts) {
+  if (!Data.isValidOffsetForDataOfSize(StartOffset, Size))  {
+    OS << "Invalid dump range\n";
     return;
   }
-
-  for (const LocationList &L : Locations) {
-    DumpLocationList(L);
-    if (&L != &Locations.back())
-      OS << '\n';
+  uint64_t Offset = StartOffset;
+  StringRef Separator;
+  bool CanContinue = true;
+  while (CanContinue && Offset < StartOffset + Size) {
+    OS << Separator;
+    Separator = "\n";
+
+    CanContinue = dumpLocationList(Data, &Offset, Version, OS, BaseAddr, MRI,
+                                   nullptr, DumpOpts, /*Indent=*/12);
+    OS << '\n';
   }
 }