1//===- DWARFDebugLine.cpp -------------------------------------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8
9#include "llvm/DebugInfo/DWARF/DWARFDebugLine.h"
10#include "llvm/ADT/Optional.h"
11#include "llvm/ADT/SmallString.h"
12#include "llvm/ADT/SmallVector.h"
13#include "llvm/ADT/StringRef.h"
14#include "llvm/BinaryFormat/Dwarf.h"
15#include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
16#include "llvm/DebugInfo/DWARF/DWARFRelocMap.h"
17#include "llvm/Support/Errc.h"
18#include "llvm/Support/Format.h"
19#include "llvm/Support/FormatVariadic.h"
20#include "llvm/Support/WithColor.h"
21#include "llvm/Support/raw_ostream.h"
22#include <algorithm>
23#include <cassert>
24#include <cinttypes>
25#include <cstdint>
26#include <cstdio>
27#include <utility>
28
29using namespace llvm;
30using namespace dwarf;
31
32using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind;
33
34namespace {
35
36struct ContentDescriptor {
37  dwarf::LineNumberEntryFormat Type;
38  dwarf::Form Form;
39};
40
41using ContentDescriptors = SmallVector<ContentDescriptor, 4>;
42
43} // end anonymous namespace
44
45static bool versionIsSupported(uint16_t Version) {
46  return Version >= 2 && Version <= 5;
47}
48
49void DWARFDebugLine::ContentTypeTracker::trackContentType(
50    dwarf::LineNumberEntryFormat ContentType) {
51  switch (ContentType) {
52  case dwarf::DW_LNCT_timestamp:
53    HasModTime = true;
54    break;
55  case dwarf::DW_LNCT_size:
56    HasLength = true;
57    break;
58  case dwarf::DW_LNCT_MD5:
59    HasMD5 = true;
60    break;
61  case dwarf::DW_LNCT_LLVM_source:
62    HasSource = true;
63    break;
64  default:
65    // We only care about values we consider optional, and new values may be
66    // added in the vendor extension range, so we do not match exhaustively.
67    break;
68  }
69}
70
71DWARFDebugLine::Prologue::Prologue() { clear(); }
72
73bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const {
74  uint16_t DwarfVersion = getVersion();
75  assert(DwarfVersion != 0 &&
76         "line table prologue has no dwarf version information");
77  if (DwarfVersion >= 5)
78    return FileIndex < FileNames.size();
79  return FileIndex != 0 && FileIndex <= FileNames.size();
80}
81
82Optional<uint64_t> DWARFDebugLine::Prologue::getLastValidFileIndex() const {
83  if (FileNames.empty())
84    return None;
85  uint16_t DwarfVersion = getVersion();
86  assert(DwarfVersion != 0 &&
87         "line table prologue has no dwarf version information");
88  // In DWARF v5 the file names are 0-indexed.
89  if (DwarfVersion >= 5)
90    return FileNames.size() - 1;
91  return FileNames.size();
92}
93
94const llvm::DWARFDebugLine::FileNameEntry &
95DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const {
96  uint16_t DwarfVersion = getVersion();
97  assert(DwarfVersion != 0 &&
98         "line table prologue has no dwarf version information");
99  // In DWARF v5 the file names are 0-indexed.
100  if (DwarfVersion >= 5)
101    return FileNames[Index];
102  return FileNames[Index - 1];
103}
104
105void DWARFDebugLine::Prologue::clear() {
106  TotalLength = PrologueLength = 0;
107  SegSelectorSize = 0;
108  MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
109  OpcodeBase = 0;
110  FormParams = dwarf::FormParams({0, 0, DWARF32});
111  ContentTypes = ContentTypeTracker();
112  StandardOpcodeLengths.clear();
113  IncludeDirectories.clear();
114  FileNames.clear();
115}
116
117void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
118                                    DIDumpOptions DumpOptions) const {
119  if (!totalLengthIsValid())
120    return;
121  int OffsetDumpWidth = 2 * dwarf::getDwarfOffsetByteSize(FormParams.Format);
122  OS << "Line table prologue:\n"
123     << format("    total_length: 0x%0*" PRIx64 "\n", OffsetDumpWidth,
124               TotalLength)
125     << "          format: " << dwarf::FormatString(FormParams.Format) << "\n"
126     << format("         version: %u\n", getVersion());
127  if (!versionIsSupported(getVersion()))
128    return;
129  if (getVersion() >= 5)
130    OS << format("    address_size: %u\n", getAddressSize())
131       << format(" seg_select_size: %u\n", SegSelectorSize);
132  OS << format(" prologue_length: 0x%0*" PRIx64 "\n", OffsetDumpWidth,
133               PrologueLength)
134     << format(" min_inst_length: %u\n", MinInstLength)
135     << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
136     << format(" default_is_stmt: %u\n", DefaultIsStmt)
137     << format("       line_base: %i\n", LineBase)
138     << format("      line_range: %u\n", LineRange)
139     << format("     opcode_base: %u\n", OpcodeBase);
140
141  for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
142    OS << formatv("standard_opcode_lengths[{0}] = {1}\n",
143                  static_cast<dwarf::LineNumberOps>(I + 1),
144                  StandardOpcodeLengths[I]);
145
146  if (!IncludeDirectories.empty()) {
147    // DWARF v5 starts directory indexes at 0.
148    uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
149    for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
150      OS << format("include_directories[%3u] = ", I + DirBase);
151      IncludeDirectories[I].dump(OS, DumpOptions);
152      OS << '\n';
153    }
154  }
155
156  if (!FileNames.empty()) {
157    // DWARF v5 starts file indexes at 0.
158    uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
159    for (uint32_t I = 0; I != FileNames.size(); ++I) {
160      const FileNameEntry &FileEntry = FileNames[I];
161      OS <<   format("file_names[%3u]:\n", I + FileBase);
162      OS <<          "           name: ";
163      FileEntry.Name.dump(OS, DumpOptions);
164      OS << '\n'
165         <<   format("      dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
166      if (ContentTypes.HasMD5)
167        OS <<        "   md5_checksum: " << FileEntry.Checksum.digest() << '\n';
168      if (ContentTypes.HasModTime)
169        OS << format("       mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
170      if (ContentTypes.HasLength)
171        OS << format("         length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
172      if (ContentTypes.HasSource) {
173        OS <<        "         source: ";
174        FileEntry.Source.dump(OS, DumpOptions);
175        OS << '\n';
176      }
177    }
178  }
179}
180
181// Parse v2-v4 directory and file tables.
182static Error
183parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
184                     uint64_t *OffsetPtr,
185                     DWARFDebugLine::ContentTypeTracker &ContentTypes,
186                     std::vector<DWARFFormValue> &IncludeDirectories,
187                     std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
188  while (true) {
189    Error Err = Error::success();
190    StringRef S = DebugLineData.getCStrRef(OffsetPtr, &Err);
191    if (Err) {
192      consumeError(std::move(Err));
193      return createStringError(errc::invalid_argument,
194                               "include directories table was not null "
195                               "terminated before the end of the prologue");
196    }
197    if (S.empty())
198      break;
199    DWARFFormValue Dir =
200        DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, S.data());
201    IncludeDirectories.push_back(Dir);
202  }
203
204  ContentTypes.HasModTime = true;
205  ContentTypes.HasLength = true;
206
207  while (true) {
208    Error Err = Error::success();
209    StringRef Name = DebugLineData.getCStrRef(OffsetPtr, &Err);
210    if (!Err && Name.empty())
211      break;
212
213    DWARFDebugLine::FileNameEntry FileEntry;
214    FileEntry.Name =
215        DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name.data());
216    FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr, &Err);
217    FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr, &Err);
218    FileEntry.Length = DebugLineData.getULEB128(OffsetPtr, &Err);
219
220    if (Err) {
221      consumeError(std::move(Err));
222      return createStringError(
223          errc::invalid_argument,
224          "file names table was not null terminated before "
225          "the end of the prologue");
226    }
227    FileNames.push_back(FileEntry);
228  }
229
230  return Error::success();
231}
232
233// Parse v5 directory/file entry content descriptions.
234// Returns the descriptors, or an error if we did not find a path or ran off
235// the end of the prologue.
236static llvm::Expected<ContentDescriptors>
237parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
238                   DWARFDebugLine::ContentTypeTracker *ContentTypes) {
239  Error Err = Error::success();
240  ContentDescriptors Descriptors;
241  int FormatCount = DebugLineData.getU8(OffsetPtr, &Err);
242  bool HasPath = false;
243  for (int I = 0; I != FormatCount && !Err; ++I) {
244    ContentDescriptor Descriptor;
245    Descriptor.Type =
246        dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr, &Err));
247    Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr, &Err));
248    if (Descriptor.Type == dwarf::DW_LNCT_path)
249      HasPath = true;
250    if (ContentTypes)
251      ContentTypes->trackContentType(Descriptor.Type);
252    Descriptors.push_back(Descriptor);
253  }
254
255  if (Err)
256    return createStringError(errc::invalid_argument,
257                             "failed to parse entry content descriptors: %s",
258                             toString(std::move(Err)).c_str());
259
260  if (!HasPath)
261    return createStringError(errc::invalid_argument,
262                             "failed to parse entry content descriptions"
263                             " because no path was found");
264  return Descriptors;
265}
266
267static Error
268parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
269                     uint64_t *OffsetPtr, const dwarf::FormParams &FormParams,
270                     const DWARFContext &Ctx, const DWARFUnit *U,
271                     DWARFDebugLine::ContentTypeTracker &ContentTypes,
272                     std::vector<DWARFFormValue> &IncludeDirectories,
273                     std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
274  // Get the directory entry description.
275  llvm::Expected<ContentDescriptors> DirDescriptors =
276      parseV5EntryFormat(DebugLineData, OffsetPtr, nullptr);
277  if (!DirDescriptors)
278    return DirDescriptors.takeError();
279
280  // Get the directory entries, according to the format described above.
281  uint64_t DirEntryCount = DebugLineData.getULEB128(OffsetPtr);
282  for (uint64_t I = 0; I != DirEntryCount; ++I) {
283    for (auto Descriptor : *DirDescriptors) {
284      DWARFFormValue Value(Descriptor.Form);
285      switch (Descriptor.Type) {
286      case DW_LNCT_path:
287        if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
288          return createStringError(errc::invalid_argument,
289                                   "failed to parse directory entry because "
290                                   "extracting the form value failed");
291        IncludeDirectories.push_back(Value);
292        break;
293      default:
294        if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
295          return createStringError(errc::invalid_argument,
296                                   "failed to parse directory entry because "
297                                   "skipping the form value failed");
298      }
299    }
300  }
301
302  // Get the file entry description.
303  llvm::Expected<ContentDescriptors> FileDescriptors =
304      parseV5EntryFormat(DebugLineData, OffsetPtr, &ContentTypes);
305  if (!FileDescriptors)
306    return FileDescriptors.takeError();
307
308  // Get the file entries, according to the format described above.
309  uint64_t FileEntryCount = DebugLineData.getULEB128(OffsetPtr);
310  for (uint64_t I = 0; I != FileEntryCount; ++I) {
311    DWARFDebugLine::FileNameEntry FileEntry;
312    for (auto Descriptor : *FileDescriptors) {
313      DWARFFormValue Value(Descriptor.Form);
314      if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
315        return createStringError(errc::invalid_argument,
316                                 "failed to parse file entry because "
317                                 "extracting the form value failed");
318      switch (Descriptor.Type) {
319      case DW_LNCT_path:
320        FileEntry.Name = Value;
321        break;
322      case DW_LNCT_LLVM_source:
323        FileEntry.Source = Value;
324        break;
325      case DW_LNCT_directory_index:
326        FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
327        break;
328      case DW_LNCT_timestamp:
329        FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
330        break;
331      case DW_LNCT_size:
332        FileEntry.Length = Value.getAsUnsignedConstant().getValue();
333        break;
334      case DW_LNCT_MD5:
335        if (!Value.getAsBlock() || Value.getAsBlock().getValue().size() != 16)
336          return createStringError(
337              errc::invalid_argument,
338              "failed to parse file entry because the MD5 hash is invalid");
339        std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16,
340                                  FileEntry.Checksum.Bytes.begin());
341        break;
342      default:
343        break;
344      }
345    }
346    FileNames.push_back(FileEntry);
347  }
348  return Error::success();
349}
350
351uint64_t DWARFDebugLine::Prologue::getLength() const {
352  uint64_t Length = PrologueLength + sizeofTotalLength() +
353                    sizeof(getVersion()) + sizeofPrologueLength();
354  if (getVersion() >= 5)
355    Length += 2; // Address + Segment selector sizes.
356  return Length;
357}
358
359Error DWARFDebugLine::Prologue::parse(
360    DWARFDataExtractor DebugLineData, uint64_t *OffsetPtr,
361    function_ref<void(Error)> RecoverableErrorHandler, const DWARFContext &Ctx,
362    const DWARFUnit *U) {
363  const uint64_t PrologueOffset = *OffsetPtr;
364
365  clear();
366  DataExtractor::Cursor Cursor(*OffsetPtr);
367  std::tie(TotalLength, FormParams.Format) =
368      DebugLineData.getInitialLength(Cursor);
369
370  DebugLineData =
371      DWARFDataExtractor(DebugLineData, Cursor.tell() + TotalLength);
372  FormParams.Version = DebugLineData.getU16(Cursor);
373  if (Cursor && !versionIsSupported(getVersion())) {
374    // Treat this error as unrecoverable - we cannot be sure what any of
375    // the data represents including the length field, so cannot skip it or make
376    // any reasonable assumptions.
377    *OffsetPtr = Cursor.tell();
378    return createStringError(
379        errc::not_supported,
380        "parsing line table prologue at offset 0x%8.8" PRIx64
381        ": unsupported version %" PRIu16,
382        PrologueOffset, getVersion());
383  }
384
385  if (getVersion() >= 5) {
386    FormParams.AddrSize = DebugLineData.getU8(Cursor);
387    assert((!Cursor || DebugLineData.getAddressSize() == 0 ||
388            DebugLineData.getAddressSize() == getAddressSize()) &&
389           "Line table header and data extractor disagree");
390    SegSelectorSize = DebugLineData.getU8(Cursor);
391  }
392
393  PrologueLength =
394      DebugLineData.getRelocatedValue(Cursor, sizeofPrologueLength());
395  const uint64_t EndPrologueOffset = PrologueLength + Cursor.tell();
396  DebugLineData = DWARFDataExtractor(DebugLineData, EndPrologueOffset);
397  MinInstLength = DebugLineData.getU8(Cursor);
398  if (getVersion() >= 4)
399    MaxOpsPerInst = DebugLineData.getU8(Cursor);
400  DefaultIsStmt = DebugLineData.getU8(Cursor);
401  LineBase = DebugLineData.getU8(Cursor);
402  LineRange = DebugLineData.getU8(Cursor);
403  OpcodeBase = DebugLineData.getU8(Cursor);
404
405  if (Cursor && OpcodeBase == 0) {
406    // If the opcode base is 0, we cannot read the standard opcode lengths (of
407    // which there are supposed to be one fewer than the opcode base). Assume
408    // there are no standard opcodes and continue parsing.
409    RecoverableErrorHandler(createStringError(
410        errc::invalid_argument,
411        "parsing line table prologue at offset 0x%8.8" PRIx64
412        " found opcode base of 0. Assuming no standard opcodes",
413        PrologueOffset));
414  } else if (Cursor) {
415    StandardOpcodeLengths.reserve(OpcodeBase - 1);
416    for (uint32_t I = 1; I < OpcodeBase; ++I) {
417      uint8_t OpLen = DebugLineData.getU8(Cursor);
418      StandardOpcodeLengths.push_back(OpLen);
419    }
420  }
421
422  *OffsetPtr = Cursor.tell();
423  // A corrupt file name or directory table does not prevent interpretation of
424  // the main line program, so check the cursor state now so that its errors can
425  // be handled separately.
426  if (!Cursor)
427    return createStringError(
428        errc::invalid_argument,
429        "parsing line table prologue at offset 0x%8.8" PRIx64 ": %s",
430        PrologueOffset, toString(Cursor.takeError()).c_str());
431
432  Error E =
433      getVersion() >= 5
434          ? parseV5DirFileTables(DebugLineData, OffsetPtr, FormParams, Ctx, U,
435                                 ContentTypes, IncludeDirectories, FileNames)
436          : parseV2DirFileTables(DebugLineData, OffsetPtr, ContentTypes,
437                                 IncludeDirectories, FileNames);
438  if (E) {
439    RecoverableErrorHandler(joinErrors(
440        createStringError(
441            errc::invalid_argument,
442            "parsing line table prologue at 0x%8.8" PRIx64
443            " found an invalid directory or file table description at"
444            " 0x%8.8" PRIx64,
445            PrologueOffset, *OffsetPtr),
446        std::move(E)));
447    return Error::success();
448  }
449
450  assert(*OffsetPtr <= EndPrologueOffset);
451  if (*OffsetPtr != EndPrologueOffset) {
452    RecoverableErrorHandler(createStringError(
453        errc::invalid_argument,
454        "unknown data in line table prologue at offset 0x%8.8" PRIx64
455        ": parsing ended (at offset 0x%8.8" PRIx64
456        ") before reaching the prologue end at offset 0x%8.8" PRIx64,
457        PrologueOffset, *OffsetPtr, EndPrologueOffset));
458  }
459  return Error::success();
460}
461
462DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
463
464void DWARFDebugLine::Row::postAppend() {
465  Discriminator = 0;
466  BasicBlock = false;
467  PrologueEnd = false;
468  EpilogueBegin = false;
469}
470
471void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
472  Address.Address = 0;
473  Address.SectionIndex = object::SectionedAddress::UndefSection;
474  Line = 1;
475  Column = 0;
476  File = 1;
477  Isa = 0;
478  Discriminator = 0;
479  IsStmt = DefaultIsStmt;
480  BasicBlock = false;
481  EndSequence = false;
482  PrologueEnd = false;
483  EpilogueBegin = false;
484}
485
486void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS, unsigned Indent) {
487  OS.indent(Indent)
488      << "Address            Line   Column File   ISA Discriminator Flags\n";
489  OS.indent(Indent)
490      << "------------------ ------ ------ ------ --- ------------- "
491         "-------------\n";
492}
493
494void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
495  OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column)
496     << format(" %6u %3u %13u ", File, Isa, Discriminator)
497     << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
498     << (PrologueEnd ? " prologue_end" : "")
499     << (EpilogueBegin ? " epilogue_begin" : "")
500     << (EndSequence ? " end_sequence" : "") << '\n';
501}
502
503DWARFDebugLine::Sequence::Sequence() { reset(); }
504
505void DWARFDebugLine::Sequence::reset() {
506  LowPC = 0;
507  HighPC = 0;
508  SectionIndex = object::SectionedAddress::UndefSection;
509  FirstRowIndex = 0;
510  LastRowIndex = 0;
511  Empty = true;
512}
513
514DWARFDebugLine::LineTable::LineTable() { clear(); }
515
516void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
517                                     DIDumpOptions DumpOptions) const {
518  Prologue.dump(OS, DumpOptions);
519
520  if (!Rows.empty()) {
521    OS << '\n';
522    Row::dumpTableHeader(OS, 0);
523    for (const Row &R : Rows) {
524      R.dump(OS);
525    }
526  }
527
528  // Terminate the table with a final blank line to clearly delineate it from
529  // later dumps.
530  OS << '\n';
531}
532
533void DWARFDebugLine::LineTable::clear() {
534  Prologue.clear();
535  Rows.clear();
536  Sequences.clear();
537}
538
539DWARFDebugLine::ParsingState::ParsingState(
540    struct LineTable *LT, uint64_t TableOffset,
541    function_ref<void(Error)> ErrorHandler)
542    : LineTable(LT), LineTableOffset(TableOffset), ErrorHandler(ErrorHandler) {
543  resetRowAndSequence();
544}
545
546void DWARFDebugLine::ParsingState::resetRowAndSequence() {
547  Row.reset(LineTable->Prologue.DefaultIsStmt);
548  Sequence.reset();
549}
550
551void DWARFDebugLine::ParsingState::appendRowToMatrix() {
552  unsigned RowNumber = LineTable->Rows.size();
553  if (Sequence.Empty) {
554    // Record the beginning of instruction sequence.
555    Sequence.Empty = false;
556    Sequence.LowPC = Row.Address.Address;
557    Sequence.FirstRowIndex = RowNumber;
558  }
559  LineTable->appendRow(Row);
560  if (Row.EndSequence) {
561    // Record the end of instruction sequence.
562    Sequence.HighPC = Row.Address.Address;
563    Sequence.LastRowIndex = RowNumber + 1;
564    Sequence.SectionIndex = Row.Address.SectionIndex;
565    if (Sequence.isValid())
566      LineTable->appendSequence(Sequence);
567    Sequence.reset();
568  }
569  Row.postAppend();
570}
571
572const DWARFDebugLine::LineTable *
573DWARFDebugLine::getLineTable(uint64_t Offset) const {
574  LineTableConstIter Pos = LineTableMap.find(Offset);
575  if (Pos != LineTableMap.end())
576    return &Pos->second;
577  return nullptr;
578}
579
580Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable(
581    DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx,
582    const DWARFUnit *U, function_ref<void(Error)> RecoverableErrorHandler) {
583  if (!DebugLineData.isValidOffset(Offset))
584    return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx64
585                       " is not a valid debug line section offset",
586                       Offset);
587
588  std::pair<LineTableIter, bool> Pos =
589      LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
590  LineTable *LT = &Pos.first->second;
591  if (Pos.second) {
592    if (Error Err =
593            LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorHandler))
594      return std::move(Err);
595    return LT;
596  }
597  return LT;
598}
599
600static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) {
601  assert(Opcode != 0);
602  if (Opcode < OpcodeBase)
603    return LNStandardString(Opcode);
604  return "special";
605}
606
607uint64_t DWARFDebugLine::ParsingState::advanceAddr(uint64_t OperationAdvance,
608                                                   uint8_t Opcode,
609                                                   uint64_t OpcodeOffset) {
610  StringRef OpcodeName = getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
611  // For versions less than 4, the MaxOpsPerInst member is set to 0, as the
612  // maximum_operations_per_instruction field wasn't introduced until DWARFv4.
613  // Don't warn about bad values in this situation.
614  if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 &&
615      LineTable->Prologue.MaxOpsPerInst != 1)
616    ErrorHandler(createStringError(
617        errc::not_supported,
618        "line table program at offset 0x%8.8" PRIx64
619        " contains a %s opcode at offset 0x%8.8" PRIx64
620        ", but the prologue maximum_operations_per_instruction value is %" PRId8
621        ", which is unsupported. Assuming a value of 1 instead",
622        LineTableOffset, OpcodeName.data(), OpcodeOffset,
623        LineTable->Prologue.MaxOpsPerInst));
624  if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0)
625    ErrorHandler(
626        createStringError(errc::invalid_argument,
627                          "line table program at offset 0x%8.8" PRIx64
628                          " contains a %s opcode at offset 0x%8.8" PRIx64
629                          ", but the prologue minimum_instruction_length value "
630                          "is 0, which prevents any address advancing",
631                          LineTableOffset, OpcodeName.data(), OpcodeOffset));
632  ReportAdvanceAddrProblem = false;
633  uint64_t AddrOffset = OperationAdvance * LineTable->Prologue.MinInstLength;
634  Row.Address.Address += AddrOffset;
635  return AddrOffset;
636}
637
638DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode
639DWARFDebugLine::ParsingState::advanceAddrForOpcode(uint8_t Opcode,
640                                                   uint64_t OpcodeOffset) {
641  assert(Opcode == DW_LNS_const_add_pc ||
642         Opcode >= LineTable->Prologue.OpcodeBase);
643  if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) {
644    StringRef OpcodeName =
645        getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
646    ErrorHandler(
647        createStringError(errc::not_supported,
648                          "line table program at offset 0x%8.8" PRIx64
649                          " contains a %s opcode at offset 0x%8.8" PRIx64
650                          ", but the prologue line_range value is 0. The "
651                          "address and line will not be adjusted",
652                          LineTableOffset, OpcodeName.data(), OpcodeOffset));
653    ReportBadLineRange = false;
654  }
655
656  uint8_t OpcodeValue = Opcode;
657  if (Opcode == DW_LNS_const_add_pc)
658    OpcodeValue = 255;
659  uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase;
660  uint64_t OperationAdvance =
661      LineTable->Prologue.LineRange != 0
662          ? AdjustedOpcode / LineTable->Prologue.LineRange
663          : 0;
664  uint64_t AddrOffset = advanceAddr(OperationAdvance, Opcode, OpcodeOffset);
665  return {AddrOffset, AdjustedOpcode};
666}
667
668DWARFDebugLine::ParsingState::AddrAndLineDelta
669DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode,
670                                                  uint64_t OpcodeOffset) {
671  // A special opcode value is chosen based on the amount that needs
672  // to be added to the line and address registers. The maximum line
673  // increment for a special opcode is the value of the line_base
674  // field in the header, plus the value of the line_range field,
675  // minus 1 (line base + line range - 1). If the desired line
676  // increment is greater than the maximum line increment, a standard
677  // opcode must be used instead of a special opcode. The "address
678  // advance" is calculated by dividing the desired address increment
679  // by the minimum_instruction_length field from the header. The
680  // special opcode is then calculated using the following formula:
681  //
682  //  opcode = (desired line increment - line_base) +
683  //           (line_range * address advance) + opcode_base
684  //
685  // If the resulting opcode is greater than 255, a standard opcode
686  // must be used instead.
687  //
688  // To decode a special opcode, subtract the opcode_base from the
689  // opcode itself to give the adjusted opcode. The amount to
690  // increment the address register is the result of the adjusted
691  // opcode divided by the line_range multiplied by the
692  // minimum_instruction_length field from the header. That is:
693  //
694  //  address increment = (adjusted opcode / line_range) *
695  //                      minimum_instruction_length
696  //
697  // The amount to increment the line register is the line_base plus
698  // the result of the adjusted opcode modulo the line_range. That is:
699  //
700  // line increment = line_base + (adjusted opcode % line_range)
701
702  DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode AddrAdvanceResult =
703      advanceAddrForOpcode(Opcode, OpcodeOffset);
704  int32_t LineOffset = 0;
705  if (LineTable->Prologue.LineRange != 0)
706    LineOffset =
707        LineTable->Prologue.LineBase +
708        (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange);
709  Row.Line += LineOffset;
710  return {AddrAdvanceResult.AddrDelta, LineOffset};
711}
712
713/// Parse a ULEB128 using the specified \p Cursor. \returns the parsed value on
714/// success, or None if \p Cursor is in a failing state.
715template <typename T>
716static Optional<T> parseULEB128(DWARFDataExtractor &Data,
717                                DataExtractor::Cursor &Cursor) {
718  T Value = Data.getULEB128(Cursor);
719  if (Cursor)
720    return Value;
721  return None;
722}
723
724Error DWARFDebugLine::LineTable::parse(
725    DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
726    const DWARFContext &Ctx, const DWARFUnit *U,
727    function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS,
728    bool Verbose) {
729  assert((OS || !Verbose) && "cannot have verbose output without stream");
730  const uint64_t DebugLineOffset = *OffsetPtr;
731
732  clear();
733
734  Error PrologueErr =
735      Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U);
736
737  if (OS) {
738    DIDumpOptions DumpOptions;
739    DumpOptions.Verbose = Verbose;
740    Prologue.dump(*OS, DumpOptions);
741  }
742
743  if (PrologueErr) {
744    // Ensure there is a blank line after the prologue to clearly delineate it
745    // from later dumps.
746    if (OS)
747      *OS << "\n";
748    return PrologueErr;
749  }
750
751  uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength();
752  if (!DebugLineData.isValidOffsetForDataOfSize(DebugLineOffset,
753                                                ProgramLength)) {
754    assert(DebugLineData.size() > DebugLineOffset &&
755           "prologue parsing should handle invalid offset");
756    uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset;
757    RecoverableErrorHandler(
758        createStringError(errc::invalid_argument,
759                          "line table program with offset 0x%8.8" PRIx64
760                          " has length 0x%8.8" PRIx64 " but only 0x%8.8" PRIx64
761                          " bytes are available",
762                          DebugLineOffset, ProgramLength, BytesRemaining));
763    // Continue by capping the length at the number of remaining bytes.
764    ProgramLength = BytesRemaining;
765  }
766
767  // Create a DataExtractor which can only see the data up to the end of the
768  // table, to prevent reading past the end.
769  const uint64_t EndOffset = DebugLineOffset + ProgramLength;
770  DWARFDataExtractor TableData(DebugLineData, EndOffset);
771
772  // See if we should tell the data extractor the address size.
773  if (TableData.getAddressSize() == 0)
774    TableData.setAddressSize(Prologue.getAddressSize());
775  else
776    assert(Prologue.getAddressSize() == 0 ||
777           Prologue.getAddressSize() == TableData.getAddressSize());
778
779  ParsingState State(this, DebugLineOffset, RecoverableErrorHandler);
780
781  *OffsetPtr = DebugLineOffset + Prologue.getLength();
782  if (OS && *OffsetPtr < EndOffset) {
783    *OS << '\n';
784    Row::dumpTableHeader(*OS, /*Indent=*/Verbose ? 12 : 0);
785  }
786  bool TombstonedAddress = false;
787  auto EmitRow = [&] {
788    if (!TombstonedAddress) {
789      if (Verbose) {
790        *OS << "\n";
791        OS->indent(12);
792      }
793      if (OS)
794        State.Row.dump(*OS);
795      State.appendRowToMatrix();
796    }
797  };
798  while (*OffsetPtr < EndOffset) {
799    DataExtractor::Cursor Cursor(*OffsetPtr);
800
801    if (Verbose)
802      *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr);
803
804    uint64_t OpcodeOffset = *OffsetPtr;
805    uint8_t Opcode = TableData.getU8(Cursor);
806    size_t RowCount = Rows.size();
807
808    if (Cursor && Verbose)
809      *OS << format("%02.02" PRIx8 " ", Opcode);
810
811    if (Opcode == 0) {
812      // Extended Opcodes always start with a zero opcode followed by
813      // a uleb128 length so you can skip ones you don't know about
814      uint64_t Len = TableData.getULEB128(Cursor);
815      uint64_t ExtOffset = Cursor.tell();
816
817      // Tolerate zero-length; assume length is correct and soldier on.
818      if (Len == 0) {
819        if (Cursor && Verbose)
820          *OS << "Badly formed extended line op (length 0)\n";
821        if (!Cursor) {
822          if (Verbose)
823            *OS << "\n";
824          RecoverableErrorHandler(Cursor.takeError());
825        }
826        *OffsetPtr = Cursor.tell();
827        continue;
828      }
829
830      uint8_t SubOpcode = TableData.getU8(Cursor);
831      // OperandOffset will be the same as ExtOffset, if it was not possible to
832      // read the SubOpcode.
833      uint64_t OperandOffset = Cursor.tell();
834      if (Verbose)
835        *OS << LNExtendedString(SubOpcode);
836      switch (SubOpcode) {
837      case DW_LNE_end_sequence:
838        // Set the end_sequence register of the state machine to true and
839        // append a row to the matrix using the current values of the
840        // state-machine registers. Then reset the registers to the initial
841        // values specified above. Every statement program sequence must end
842        // with a DW_LNE_end_sequence instruction which creates a row whose
843        // address is that of the byte after the last target machine instruction
844        // of the sequence.
845        State.Row.EndSequence = true;
846        // No need to test the Cursor is valid here, since it must be to get
847        // into this code path - if it were invalid, the default case would be
848        // followed.
849        EmitRow();
850        State.resetRowAndSequence();
851        break;
852
853      case DW_LNE_set_address:
854        // Takes a single relocatable address as an operand. The size of the
855        // operand is the size appropriate to hold an address on the target
856        // machine. Set the address register to the value given by the
857        // relocatable address. All of the other statement program opcodes
858        // that affect the address register add a delta to it. This instruction
859        // stores a relocatable value into it instead.
860        //
861        // Make sure the extractor knows the address size.  If not, infer it
862        // from the size of the operand.
863        {
864          uint8_t ExtractorAddressSize = TableData.getAddressSize();
865          uint64_t OpcodeAddressSize = Len - 1;
866          if (ExtractorAddressSize != OpcodeAddressSize &&
867              ExtractorAddressSize != 0)
868            RecoverableErrorHandler(createStringError(
869                errc::invalid_argument,
870                "mismatching address size at offset 0x%8.8" PRIx64
871                " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64,
872                ExtOffset, ExtractorAddressSize, Len - 1));
873
874          // Assume that the line table is correct and temporarily override the
875          // address size. If the size is unsupported, give up trying to read
876          // the address and continue to the next opcode.
877          if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 &&
878              OpcodeAddressSize != 4 && OpcodeAddressSize != 8) {
879            RecoverableErrorHandler(createStringError(
880                errc::invalid_argument,
881                "address size 0x%2.2" PRIx64
882                " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64
883                " is unsupported",
884                OpcodeAddressSize, ExtOffset));
885            TableData.skip(Cursor, OpcodeAddressSize);
886          } else {
887            TableData.setAddressSize(OpcodeAddressSize);
888            State.Row.Address.Address = TableData.getRelocatedAddress(
889                Cursor, &State.Row.Address.SectionIndex);
890
891            uint64_t Tombstone =
892                dwarf::computeTombstoneAddress(OpcodeAddressSize);
893            TombstonedAddress = State.Row.Address.Address == Tombstone;
894
895            // Restore the address size if the extractor already had it.
896            if (ExtractorAddressSize != 0)
897              TableData.setAddressSize(ExtractorAddressSize);
898          }
899
900          if (Cursor && Verbose) {
901            *OS << " (";
902            DWARFFormValue::dumpAddress(*OS, OpcodeAddressSize, State.Row.Address.Address);
903            *OS << ')';
904          }
905        }
906        break;
907
908      case DW_LNE_define_file:
909        // Takes 4 arguments. The first is a null terminated string containing
910        // a source file name. The second is an unsigned LEB128 number
911        // representing the directory index of the directory in which the file
912        // was found. The third is an unsigned LEB128 number representing the
913        // time of last modification of the file. The fourth is an unsigned
914        // LEB128 number representing the length in bytes of the file. The time
915        // and length fields may contain LEB128(0) if the information is not
916        // available.
917        //
918        // The directory index represents an entry in the include_directories
919        // section of the statement program prologue. The index is LEB128(0)
920        // if the file was found in the current directory of the compilation,
921        // LEB128(1) if it was found in the first directory in the
922        // include_directories section, and so on. The directory index is
923        // ignored for file names that represent full path names.
924        //
925        // The files are numbered, starting at 1, in the order in which they
926        // appear; the names in the prologue come before names defined by
927        // the DW_LNE_define_file instruction. These numbers are used in the
928        // the file register of the state machine.
929        {
930          FileNameEntry FileEntry;
931          const char *Name = TableData.getCStr(Cursor);
932          FileEntry.Name =
933              DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name);
934          FileEntry.DirIdx = TableData.getULEB128(Cursor);
935          FileEntry.ModTime = TableData.getULEB128(Cursor);
936          FileEntry.Length = TableData.getULEB128(Cursor);
937          Prologue.FileNames.push_back(FileEntry);
938          if (Cursor && Verbose)
939            *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
940                << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
941                << ", length=" << FileEntry.Length << ")";
942        }
943        break;
944
945      case DW_LNE_set_discriminator:
946        State.Row.Discriminator = TableData.getULEB128(Cursor);
947        if (Cursor && Verbose)
948          *OS << " (" << State.Row.Discriminator << ")";
949        break;
950
951      default:
952        if (Cursor && Verbose)
953          *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
954              << format(" length %" PRIx64, Len);
955        // Len doesn't include the zero opcode byte or the length itself, but
956        // it does include the sub_opcode, so we have to adjust for that.
957        TableData.skip(Cursor, Len - 1);
958        break;
959      }
960      // Make sure the length as recorded in the table and the standard length
961      // for the opcode match. If they don't, continue from the end as claimed
962      // by the table. Similarly, continue from the claimed end in the event of
963      // a parsing error.
964      uint64_t End = ExtOffset + Len;
965      if (Cursor && Cursor.tell() != End)
966        RecoverableErrorHandler(createStringError(
967            errc::illegal_byte_sequence,
968            "unexpected line op length at offset 0x%8.8" PRIx64
969            " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64,
970            ExtOffset, Len, Cursor.tell() - ExtOffset));
971      if (!Cursor && Verbose) {
972        DWARFDataExtractor::Cursor ByteCursor(OperandOffset);
973        uint8_t Byte = TableData.getU8(ByteCursor);
974        if (ByteCursor) {
975          *OS << " (<parsing error>";
976          do {
977            *OS << format(" %2.2" PRIx8, Byte);
978            Byte = TableData.getU8(ByteCursor);
979          } while (ByteCursor);
980          *OS << ")";
981        }
982
983        // The only parse failure in this case should be if the end was reached.
984        // In that case, throw away the error, as the main Cursor's error will
985        // be sufficient.
986        consumeError(ByteCursor.takeError());
987      }
988      *OffsetPtr = End;
989    } else if (Opcode < Prologue.OpcodeBase) {
990      if (Verbose)
991        *OS << LNStandardString(Opcode);
992      switch (Opcode) {
993      // Standard Opcodes
994      case DW_LNS_copy:
995        // Takes no arguments. Append a row to the matrix using the
996        // current values of the state-machine registers.
997        EmitRow();
998        break;
999
1000      case DW_LNS_advance_pc:
1001        // Takes a single unsigned LEB128 operand, multiplies it by the
1002        // min_inst_length field of the prologue, and adds the
1003        // result to the address register of the state machine.
1004        if (Optional<uint64_t> Operand =
1005                parseULEB128<uint64_t>(TableData, Cursor)) {
1006          uint64_t AddrOffset =
1007              State.advanceAddr(*Operand, Opcode, OpcodeOffset);
1008          if (Verbose)
1009            *OS << " (" << AddrOffset << ")";
1010        }
1011        break;
1012
1013      case DW_LNS_advance_line:
1014        // Takes a single signed LEB128 operand and adds that value to
1015        // the line register of the state machine.
1016        {
1017          int64_t LineDelta = TableData.getSLEB128(Cursor);
1018          if (Cursor) {
1019            State.Row.Line += LineDelta;
1020            if (Verbose)
1021              *OS << " (" << State.Row.Line << ")";
1022          }
1023        }
1024        break;
1025
1026      case DW_LNS_set_file:
1027        // Takes a single unsigned LEB128 operand and stores it in the file
1028        // register of the state machine.
1029        if (Optional<uint16_t> File =
1030                parseULEB128<uint16_t>(TableData, Cursor)) {
1031          State.Row.File = *File;
1032          if (Verbose)
1033            *OS << " (" << State.Row.File << ")";
1034        }
1035        break;
1036
1037      case DW_LNS_set_column:
1038        // Takes a single unsigned LEB128 operand and stores it in the
1039        // column register of the state machine.
1040        if (Optional<uint16_t> Column =
1041                parseULEB128<uint16_t>(TableData, Cursor)) {
1042          State.Row.Column = *Column;
1043          if (Verbose)
1044            *OS << " (" << State.Row.Column << ")";
1045        }
1046        break;
1047
1048      case DW_LNS_negate_stmt:
1049        // Takes no arguments. Set the is_stmt register of the state
1050        // machine to the logical negation of its current value.
1051        State.Row.IsStmt = !State.Row.IsStmt;
1052        break;
1053
1054      case DW_LNS_set_basic_block:
1055        // Takes no arguments. Set the basic_block register of the
1056        // state machine to true
1057        State.Row.BasicBlock = true;
1058        break;
1059
1060      case DW_LNS_const_add_pc:
1061        // Takes no arguments. Add to the address register of the state
1062        // machine the address increment value corresponding to special
1063        // opcode 255. The motivation for DW_LNS_const_add_pc is this:
1064        // when the statement program needs to advance the address by a
1065        // small amount, it can use a single special opcode, which occupies
1066        // a single byte. When it needs to advance the address by up to
1067        // twice the range of the last special opcode, it can use
1068        // DW_LNS_const_add_pc followed by a special opcode, for a total
1069        // of two bytes. Only if it needs to advance the address by more
1070        // than twice that range will it need to use both DW_LNS_advance_pc
1071        // and a special opcode, requiring three or more bytes.
1072        {
1073          uint64_t AddrOffset =
1074              State.advanceAddrForOpcode(Opcode, OpcodeOffset).AddrDelta;
1075          if (Verbose)
1076            *OS << format(" (0x%16.16" PRIx64 ")", AddrOffset);
1077        }
1078        break;
1079
1080      case DW_LNS_fixed_advance_pc:
1081        // Takes a single uhalf operand. Add to the address register of
1082        // the state machine the value of the (unencoded) operand. This
1083        // is the only extended opcode that takes an argument that is not
1084        // a variable length number. The motivation for DW_LNS_fixed_advance_pc
1085        // is this: existing assemblers cannot emit DW_LNS_advance_pc or
1086        // special opcodes because they cannot encode LEB128 numbers or
1087        // judge when the computation of a special opcode overflows and
1088        // requires the use of DW_LNS_advance_pc. Such assemblers, however,
1089        // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
1090        {
1091          uint16_t PCOffset =
1092              TableData.getRelocatedValue(Cursor, 2);
1093          if (Cursor) {
1094            State.Row.Address.Address += PCOffset;
1095            if (Verbose)
1096              *OS << format(" (0x%4.4" PRIx16 ")", PCOffset);
1097          }
1098        }
1099        break;
1100
1101      case DW_LNS_set_prologue_end:
1102        // Takes no arguments. Set the prologue_end register of the
1103        // state machine to true
1104        State.Row.PrologueEnd = true;
1105        break;
1106
1107      case DW_LNS_set_epilogue_begin:
1108        // Takes no arguments. Set the basic_block register of the
1109        // state machine to true
1110        State.Row.EpilogueBegin = true;
1111        break;
1112
1113      case DW_LNS_set_isa:
1114        // Takes a single unsigned LEB128 operand and stores it in the
1115        // ISA register of the state machine.
1116        if (Optional<uint8_t> Isa = parseULEB128<uint8_t>(TableData, Cursor)) {
1117          State.Row.Isa = *Isa;
1118          if (Verbose)
1119            *OS << " (" << (uint64_t)State.Row.Isa << ")";
1120        }
1121        break;
1122
1123      default:
1124        // Handle any unknown standard opcodes here. We know the lengths
1125        // of such opcodes because they are specified in the prologue
1126        // as a multiple of LEB128 operands for each opcode.
1127        {
1128          assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
1129          if (Verbose)
1130            *OS << "Unrecognized standard opcode";
1131          uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
1132          std::vector<uint64_t> Operands;
1133          for (uint8_t I = 0; I < OpcodeLength; ++I) {
1134            if (Optional<uint64_t> Value =
1135                    parseULEB128<uint64_t>(TableData, Cursor))
1136              Operands.push_back(*Value);
1137            else
1138              break;
1139          }
1140          if (Verbose && !Operands.empty()) {
1141            *OS << " (operands: ";
1142            bool First = true;
1143            for (uint64_t Value : Operands) {
1144              if (!First)
1145                *OS << ", ";
1146              First = false;
1147              *OS << format("0x%16.16" PRIx64, Value);
1148            }
1149            if (Verbose)
1150              *OS << ')';
1151          }
1152        }
1153        break;
1154      }
1155
1156      *OffsetPtr = Cursor.tell();
1157    } else {
1158      // Special Opcodes.
1159      ParsingState::AddrAndLineDelta Delta =
1160          State.handleSpecialOpcode(Opcode, OpcodeOffset);
1161
1162      if (Verbose)
1163        *OS << "address += " << Delta.Address << ",  line += " << Delta.Line;
1164      EmitRow();
1165      *OffsetPtr = Cursor.tell();
1166    }
1167
1168    // When a row is added to the matrix, it is also dumped, which includes a
1169    // new line already, so don't add an extra one.
1170    if (Verbose && Rows.size() == RowCount)
1171      *OS << "\n";
1172
1173    // Most parse failures other than when parsing extended opcodes are due to
1174    // failures to read ULEBs. Bail out of parsing, since we don't know where to
1175    // continue reading from as there is no stated length for such byte
1176    // sequences. Print the final trailing new line if needed before doing so.
1177    if (!Cursor && Opcode != 0) {
1178      if (Verbose)
1179        *OS << "\n";
1180      return Cursor.takeError();
1181    }
1182
1183    if (!Cursor)
1184      RecoverableErrorHandler(Cursor.takeError());
1185  }
1186
1187  if (!State.Sequence.Empty)
1188    RecoverableErrorHandler(createStringError(
1189        errc::illegal_byte_sequence,
1190        "last sequence in debug line table at offset 0x%8.8" PRIx64
1191        " is not terminated",
1192        DebugLineOffset));
1193
1194  // Sort all sequences so that address lookup will work faster.
1195  if (!Sequences.empty()) {
1196    llvm::sort(Sequences, Sequence::orderByHighPC);
1197    // Note: actually, instruction address ranges of sequences should not
1198    // overlap (in shared objects and executables). If they do, the address
1199    // lookup would still work, though, but result would be ambiguous.
1200    // We don't report warning in this case. For example,
1201    // sometimes .so compiled from multiple object files contains a few
1202    // rudimentary sequences for address ranges [0x0, 0xsomething).
1203  }
1204
1205  // Terminate the table with a final blank line to clearly delineate it from
1206  // later dumps.
1207  if (OS)
1208    *OS << "\n";
1209
1210  return Error::success();
1211}
1212
1213uint32_t DWARFDebugLine::LineTable::findRowInSeq(
1214    const DWARFDebugLine::Sequence &Seq,
1215    object::SectionedAddress Address) const {
1216  if (!Seq.containsPC(Address))
1217    return UnknownRowIndex;
1218  assert(Seq.SectionIndex == Address.SectionIndex);
1219  // In some cases, e.g. first instruction in a function, the compiler generates
1220  // two entries, both with the same address. We want the last one.
1221  //
1222  // In general we want a non-empty range: the last row whose address is less
1223  // than or equal to Address. This can be computed as upper_bound - 1.
1224  DWARFDebugLine::Row Row;
1225  Row.Address = Address;
1226  RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
1227  RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
1228  assert(FirstRow->Address.Address <= Row.Address.Address &&
1229         Row.Address.Address < LastRow[-1].Address.Address);
1230  RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row,
1231                                    DWARFDebugLine::Row::orderByAddress) -
1232                   1;
1233  assert(Seq.SectionIndex == RowPos->Address.SectionIndex);
1234  return RowPos - Rows.begin();
1235}
1236
1237uint32_t DWARFDebugLine::LineTable::lookupAddress(
1238    object::SectionedAddress Address) const {
1239
1240  // Search for relocatable addresses
1241  uint32_t Result = lookupAddressImpl(Address);
1242
1243  if (Result != UnknownRowIndex ||
1244      Address.SectionIndex == object::SectionedAddress::UndefSection)
1245    return Result;
1246
1247  // Search for absolute addresses
1248  Address.SectionIndex = object::SectionedAddress::UndefSection;
1249  return lookupAddressImpl(Address);
1250}
1251
1252uint32_t DWARFDebugLine::LineTable::lookupAddressImpl(
1253    object::SectionedAddress Address) const {
1254  // First, find an instruction sequence containing the given address.
1255  DWARFDebugLine::Sequence Sequence;
1256  Sequence.SectionIndex = Address.SectionIndex;
1257  Sequence.HighPC = Address.Address;
1258  SequenceIter It = llvm::upper_bound(Sequences, Sequence,
1259                                      DWARFDebugLine::Sequence::orderByHighPC);
1260  if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex)
1261    return UnknownRowIndex;
1262  return findRowInSeq(*It, Address);
1263}
1264
1265bool DWARFDebugLine::LineTable::lookupAddressRange(
1266    object::SectionedAddress Address, uint64_t Size,
1267    std::vector<uint32_t> &Result) const {
1268
1269  // Search for relocatable addresses
1270  if (lookupAddressRangeImpl(Address, Size, Result))
1271    return true;
1272
1273  if (Address.SectionIndex == object::SectionedAddress::UndefSection)
1274    return false;
1275
1276  // Search for absolute addresses
1277  Address.SectionIndex = object::SectionedAddress::UndefSection;
1278  return lookupAddressRangeImpl(Address, Size, Result);
1279}
1280
1281bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
1282    object::SectionedAddress Address, uint64_t Size,
1283    std::vector<uint32_t> &Result) const {
1284  if (Sequences.empty())
1285    return false;
1286  uint64_t EndAddr = Address.Address + Size;
1287  // First, find an instruction sequence containing the given address.
1288  DWARFDebugLine::Sequence Sequence;
1289  Sequence.SectionIndex = Address.SectionIndex;
1290  Sequence.HighPC = Address.Address;
1291  SequenceIter LastSeq = Sequences.end();
1292  SequenceIter SeqPos = llvm::upper_bound(
1293      Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC);
1294  if (SeqPos == LastSeq || !SeqPos->containsPC(Address))
1295    return false;
1296
1297  SequenceIter StartPos = SeqPos;
1298
1299  // Add the rows from the first sequence to the vector, starting with the
1300  // index we just calculated
1301
1302  while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
1303    const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
1304    // For the first sequence, we need to find which row in the sequence is the
1305    // first in our range.
1306    uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
1307    if (SeqPos == StartPos)
1308      FirstRowIndex = findRowInSeq(CurSeq, Address);
1309
1310    // Figure out the last row in the range.
1311    uint32_t LastRowIndex =
1312        findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
1313    if (LastRowIndex == UnknownRowIndex)
1314      LastRowIndex = CurSeq.LastRowIndex - 1;
1315
1316    assert(FirstRowIndex != UnknownRowIndex);
1317    assert(LastRowIndex != UnknownRowIndex);
1318
1319    for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
1320      Result.push_back(I);
1321    }
1322
1323    ++SeqPos;
1324  }
1325
1326  return true;
1327}
1328
1329Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
1330                                                                FileLineInfoKind Kind) const {
1331  if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex))
1332    return None;
1333  const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex);
1334  if (Optional<const char *> source = Entry.Source.getAsCString())
1335    return StringRef(*source);
1336  return None;
1337}
1338
1339static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
1340  // Debug info can contain paths from any OS, not necessarily
1341  // an OS we're currently running on. Moreover different compilation units can
1342  // be compiled on different operating systems and linked together later.
1343  return sys::path::is_absolute(Path, sys::path::Style::posix) ||
1344         sys::path::is_absolute(Path, sys::path::Style::windows);
1345}
1346
1347bool DWARFDebugLine::Prologue::getFileNameByIndex(
1348    uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind,
1349    std::string &Result, sys::path::Style Style) const {
1350  if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
1351    return false;
1352  const FileNameEntry &Entry = getFileNameEntry(FileIndex);
1353  Optional<const char *> Name = Entry.Name.getAsCString();
1354  if (!Name)
1355    return false;
1356  StringRef FileName = *Name;
1357  if (Kind == FileLineInfoKind::RawValue ||
1358      isPathAbsoluteOnWindowsOrPosix(FileName)) {
1359    Result = std::string(FileName);
1360    return true;
1361  }
1362  if (Kind == FileLineInfoKind::BaseNameOnly) {
1363    Result = std::string(llvm::sys::path::filename(FileName));
1364    return true;
1365  }
1366
1367  SmallString<16> FilePath;
1368  StringRef IncludeDir;
1369  // Be defensive about the contents of Entry.
1370  if (getVersion() >= 5) {
1371    // DirIdx 0 is the compilation directory, so don't include it for
1372    // relative names.
1373    if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) &&
1374        Entry.DirIdx < IncludeDirectories.size())
1375      IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue();
1376  } else {
1377    if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size())
1378      IncludeDir =
1379          IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue();
1380  }
1381
1382  // For absolute paths only, include the compilation directory of compile unit.
1383  // We know that FileName is not absolute, the only way to have an absolute
1384  // path at this point would be if IncludeDir is absolute.
1385  if (Kind == FileLineInfoKind::AbsoluteFilePath && !CompDir.empty() &&
1386      !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1387    sys::path::append(FilePath, Style, CompDir);
1388
1389  assert((Kind == FileLineInfoKind::AbsoluteFilePath ||
1390          Kind == FileLineInfoKind::RelativeFilePath) &&
1391         "invalid FileLineInfo Kind");
1392
1393  // sys::path::append skips empty strings.
1394  sys::path::append(FilePath, Style, IncludeDir, FileName);
1395  Result = std::string(FilePath.str());
1396  return true;
1397}
1398
1399bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1400    object::SectionedAddress Address, const char *CompDir,
1401    FileLineInfoKind Kind, DILineInfo &Result) const {
1402  // Get the index of row we're looking for in the line table.
1403  uint32_t RowIndex = lookupAddress(Address);
1404  if (RowIndex == -1U)
1405    return false;
1406  // Take file number and line/column from the row.
1407  const auto &Row = Rows[RowIndex];
1408  if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1409    return false;
1410  Result.Line = Row.Line;
1411  Result.Column = Row.Column;
1412  Result.Discriminator = Row.Discriminator;
1413  Result.Source = getSourceByIndex(Row.File, Kind);
1414  return true;
1415}
1416
1417// We want to supply the Unit associated with a .debug_line[.dwo] table when
1418// we dump it, if possible, but still dump the table even if there isn't a Unit.
1419// Therefore, collect up handles on all the Units that point into the
1420// line-table section.
1421static DWARFDebugLine::SectionParser::LineToUnitMap
1422buildLineToUnitMap(DWARFUnitVector::iterator_range Units) {
1423  DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit;
1424  for (const auto &U : Units)
1425    if (auto CUDIE = U->getUnitDIE())
1426      if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list)))
1427        LineToUnit.insert(std::make_pair(*StmtOffset, &*U));
1428  return LineToUnit;
1429}
1430
1431DWARFDebugLine::SectionParser::SectionParser(
1432    DWARFDataExtractor &Data, const DWARFContext &C,
1433    DWARFUnitVector::iterator_range Units)
1434    : DebugLineData(Data), Context(C) {
1435  LineToUnit = buildLineToUnitMap(Units);
1436  if (!DebugLineData.isValidOffset(Offset))
1437    Done = true;
1438}
1439
1440bool DWARFDebugLine::Prologue::totalLengthIsValid() const {
1441  return TotalLength != 0u;
1442}
1443
1444DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext(
1445    function_ref<void(Error)> RecoverableErrorHandler,
1446    function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS,
1447    bool Verbose) {
1448  assert(DebugLineData.isValidOffset(Offset) &&
1449         "parsing should have terminated");
1450  DWARFUnit *U = prepareToParse(Offset);
1451  uint64_t OldOffset = Offset;
1452  LineTable LT;
1453  if (Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1454                           RecoverableErrorHandler, OS, Verbose))
1455    UnrecoverableErrorHandler(std::move(Err));
1456  moveToNextTable(OldOffset, LT.Prologue);
1457  return LT;
1458}
1459
1460void DWARFDebugLine::SectionParser::skip(
1461    function_ref<void(Error)> RecoverableErrorHandler,
1462    function_ref<void(Error)> UnrecoverableErrorHandler) {
1463  assert(DebugLineData.isValidOffset(Offset) &&
1464         "parsing should have terminated");
1465  DWARFUnit *U = prepareToParse(Offset);
1466  uint64_t OldOffset = Offset;
1467  LineTable LT;
1468  if (Error Err = LT.Prologue.parse(DebugLineData, &Offset,
1469                                    RecoverableErrorHandler, Context, U))
1470    UnrecoverableErrorHandler(std::move(Err));
1471  moveToNextTable(OldOffset, LT.Prologue);
1472}
1473
1474DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) {
1475  DWARFUnit *U = nullptr;
1476  auto It = LineToUnit.find(Offset);
1477  if (It != LineToUnit.end())
1478    U = It->second;
1479  DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0);
1480  return U;
1481}
1482
1483void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset,
1484                                                    const Prologue &P) {
1485  // If the length field is not valid, we don't know where the next table is, so
1486  // cannot continue to parse. Mark the parser as done, and leave the Offset
1487  // value as it currently is. This will be the end of the bad length field.
1488  if (!P.totalLengthIsValid()) {
1489    Done = true;
1490    return;
1491  }
1492
1493  Offset = OldOffset + P.TotalLength + P.sizeofTotalLength();
1494  if (!DebugLineData.isValidOffset(Offset)) {
1495    Done = true;
1496  }
1497}
1498