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