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, uint64_t EndPrologueOffset,
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 || *OffsetPtr > EndPrologueOffset) {
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 && *OffsetPtr <= EndPrologueOffset && 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 || *OffsetPtr > EndPrologueOffset) {
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     const 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   FormParams.Version = DebugLineData.getU16(OffsetPtr);
364   if (!versionIsSupported(getVersion()))
365     // Treat this error as unrecoverable - we cannot be sure what any of
366     // the data represents including the length field, so cannot skip it or make
367     // any reasonable assumptions.
368     return createStringError(
369         errc::not_supported,
370         "parsing line table prologue at offset 0x%8.8" PRIx64
371         ": unsupported version %" PRIu16,
372         PrologueOffset, getVersion());
373 
374   if (getVersion() >= 5) {
375     FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
376     assert((DebugLineData.getAddressSize() == 0 ||
377             DebugLineData.getAddressSize() == getAddressSize()) &&
378            "Line table header and data extractor disagree");
379     SegSelectorSize = DebugLineData.getU8(OffsetPtr);
380   }
381 
382   PrologueLength =
383       DebugLineData.getRelocatedValue(sizeofPrologueLength(), OffsetPtr);
384   const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
385   MinInstLength = DebugLineData.getU8(OffsetPtr);
386   if (getVersion() >= 4)
387     MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
388   DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
389   LineBase = DebugLineData.getU8(OffsetPtr);
390   LineRange = DebugLineData.getU8(OffsetPtr);
391   OpcodeBase = DebugLineData.getU8(OffsetPtr);
392 
393   if (OpcodeBase == 0) {
394     // If the opcode base is 0, we cannot read the standard opcode lengths (of
395     // which there are supposed to be one fewer than the opcode base). Assume
396     // there are no standard opcodes and continue parsing.
397     RecoverableErrorHandler(createStringError(
398         errc::invalid_argument,
399         "parsing line table prologue at offset 0x%8.8" PRIx64
400         " found opcode base of 0. Assuming no standard opcodes",
401         PrologueOffset));
402   } else {
403     StandardOpcodeLengths.reserve(OpcodeBase - 1);
404     for (uint32_t I = 1; I < OpcodeBase; ++I) {
405       uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
406       StandardOpcodeLengths.push_back(OpLen);
407     }
408   }
409 
410   Error E =
411       getVersion() >= 5
412           ? parseV5DirFileTables(DebugLineData, OffsetPtr, FormParams, Ctx, U,
413                                  ContentTypes, IncludeDirectories, FileNames)
414           : parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
415                                  ContentTypes, IncludeDirectories, FileNames);
416   if (E) {
417     RecoverableErrorHandler(joinErrors(
418         createStringError(
419             errc::invalid_argument,
420             "parsing line table prologue at 0x%8.8" PRIx64
421             " found an invalid directory or file table description at"
422             " 0x%8.8" PRIx64,
423             PrologueOffset, *OffsetPtr),
424         std::move(E)));
425     return Error::success();
426   }
427 
428   if (*OffsetPtr != EndPrologueOffset) {
429     RecoverableErrorHandler(createStringError(
430         errc::invalid_argument,
431         "parsing line table prologue at 0x%8.8" PRIx64
432         " should have ended at 0x%8.8" PRIx64 " but it ended at 0x%8.8" PRIx64,
433         PrologueOffset, EndPrologueOffset, *OffsetPtr));
434     *OffsetPtr = EndPrologueOffset;
435   }
436   return Error::success();
437 }
438 
439 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
440 
441 void DWARFDebugLine::Row::postAppend() {
442   Discriminator = 0;
443   BasicBlock = false;
444   PrologueEnd = false;
445   EpilogueBegin = false;
446 }
447 
448 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
449   Address.Address = 0;
450   Address.SectionIndex = object::SectionedAddress::UndefSection;
451   Line = 1;
452   Column = 0;
453   File = 1;
454   Isa = 0;
455   Discriminator = 0;
456   IsStmt = DefaultIsStmt;
457   BasicBlock = false;
458   EndSequence = false;
459   PrologueEnd = false;
460   EpilogueBegin = false;
461 }
462 
463 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS, unsigned Indent) {
464   OS.indent(Indent)
465       << "Address            Line   Column File   ISA Discriminator Flags\n";
466   OS.indent(Indent)
467       << "------------------ ------ ------ ------ --- ------------- "
468          "-------------\n";
469 }
470 
471 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
472   OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column)
473      << format(" %6u %3u %13u ", File, Isa, Discriminator)
474      << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
475      << (PrologueEnd ? " prologue_end" : "")
476      << (EpilogueBegin ? " epilogue_begin" : "")
477      << (EndSequence ? " end_sequence" : "") << '\n';
478 }
479 
480 DWARFDebugLine::Sequence::Sequence() { reset(); }
481 
482 void DWARFDebugLine::Sequence::reset() {
483   LowPC = 0;
484   HighPC = 0;
485   SectionIndex = object::SectionedAddress::UndefSection;
486   FirstRowIndex = 0;
487   LastRowIndex = 0;
488   Empty = true;
489 }
490 
491 DWARFDebugLine::LineTable::LineTable() { clear(); }
492 
493 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
494                                      DIDumpOptions DumpOptions) const {
495   Prologue.dump(OS, DumpOptions);
496 
497   if (!Rows.empty()) {
498     OS << '\n';
499     Row::dumpTableHeader(OS, 0);
500     for (const Row &R : Rows) {
501       R.dump(OS);
502     }
503   }
504 
505   // Terminate the table with a final blank line to clearly delineate it from
506   // later dumps.
507   OS << '\n';
508 }
509 
510 void DWARFDebugLine::LineTable::clear() {
511   Prologue.clear();
512   Rows.clear();
513   Sequences.clear();
514 }
515 
516 DWARFDebugLine::ParsingState::ParsingState(
517     struct LineTable *LT, uint64_t TableOffset,
518     function_ref<void(Error)> ErrorHandler)
519     : LineTable(LT), LineTableOffset(TableOffset), ErrorHandler(ErrorHandler) {
520   resetRowAndSequence();
521 }
522 
523 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
524   Row.reset(LineTable->Prologue.DefaultIsStmt);
525   Sequence.reset();
526 }
527 
528 void DWARFDebugLine::ParsingState::appendRowToMatrix() {
529   unsigned RowNumber = LineTable->Rows.size();
530   if (Sequence.Empty) {
531     // Record the beginning of instruction sequence.
532     Sequence.Empty = false;
533     Sequence.LowPC = Row.Address.Address;
534     Sequence.FirstRowIndex = RowNumber;
535   }
536   LineTable->appendRow(Row);
537   if (Row.EndSequence) {
538     // Record the end of instruction sequence.
539     Sequence.HighPC = Row.Address.Address;
540     Sequence.LastRowIndex = RowNumber + 1;
541     Sequence.SectionIndex = Row.Address.SectionIndex;
542     if (Sequence.isValid())
543       LineTable->appendSequence(Sequence);
544     Sequence.reset();
545   }
546   Row.postAppend();
547 }
548 
549 const DWARFDebugLine::LineTable *
550 DWARFDebugLine::getLineTable(uint64_t Offset) const {
551   LineTableConstIter Pos = LineTableMap.find(Offset);
552   if (Pos != LineTableMap.end())
553     return &Pos->second;
554   return nullptr;
555 }
556 
557 Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable(
558     DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx,
559     const DWARFUnit *U, function_ref<void(Error)> RecoverableErrorHandler) {
560   if (!DebugLineData.isValidOffset(Offset))
561     return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx64
562                        " is not a valid debug line section offset",
563                        Offset);
564 
565   std::pair<LineTableIter, bool> Pos =
566       LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
567   LineTable *LT = &Pos.first->second;
568   if (Pos.second) {
569     if (Error Err =
570             LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorHandler))
571       return std::move(Err);
572     return LT;
573   }
574   return LT;
575 }
576 
577 static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) {
578   assert(Opcode != 0);
579   if (Opcode < OpcodeBase)
580     return LNStandardString(Opcode);
581   return "special";
582 }
583 
584 uint64_t DWARFDebugLine::ParsingState::advanceAddr(uint64_t OperationAdvance,
585                                                    uint8_t Opcode,
586                                                    uint64_t OpcodeOffset) {
587   StringRef OpcodeName = getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
588   // For versions less than 4, the MaxOpsPerInst member is set to 0, as the
589   // maximum_operations_per_instruction field wasn't introduced until DWARFv4.
590   // Don't warn about bad values in this situation.
591   if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 &&
592       LineTable->Prologue.MaxOpsPerInst != 1)
593     ErrorHandler(createStringError(
594         errc::not_supported,
595         "line table program at offset 0x%8.8" PRIx64
596         " contains a %s opcode at offset 0x%8.8" PRIx64
597         ", but the prologue maximum_operations_per_instruction value is %" PRId8
598         ", which is unsupported. Assuming a value of 1 instead",
599         LineTableOffset, OpcodeName.data(), OpcodeOffset,
600         LineTable->Prologue.MaxOpsPerInst));
601   if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0)
602     ErrorHandler(
603         createStringError(errc::invalid_argument,
604                           "line table program at offset 0x%8.8" PRIx64
605                           " contains a %s opcode at offset 0x%8.8" PRIx64
606                           ", but the prologue minimum_instruction_length value "
607                           "is 0, which prevents any address advancing",
608                           LineTableOffset, OpcodeName.data(), OpcodeOffset));
609   ReportAdvanceAddrProblem = false;
610   uint64_t AddrOffset = OperationAdvance * LineTable->Prologue.MinInstLength;
611   Row.Address.Address += AddrOffset;
612   return AddrOffset;
613 }
614 
615 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode
616 DWARFDebugLine::ParsingState::advanceAddrForOpcode(uint8_t Opcode,
617                                                    uint64_t OpcodeOffset) {
618   assert(Opcode == DW_LNS_const_add_pc ||
619          Opcode >= LineTable->Prologue.OpcodeBase);
620   if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) {
621     StringRef OpcodeName =
622         getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
623     ErrorHandler(
624         createStringError(errc::not_supported,
625                           "line table program at offset 0x%8.8" PRIx64
626                           " contains a %s opcode at offset 0x%8.8" PRIx64
627                           ", but the prologue line_range value is 0. The "
628                           "address and line will not be adjusted",
629                           LineTableOffset, OpcodeName.data(), OpcodeOffset));
630     ReportBadLineRange = false;
631   }
632 
633   uint8_t OpcodeValue = Opcode;
634   if (Opcode == DW_LNS_const_add_pc)
635     OpcodeValue = 255;
636   uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase;
637   uint64_t OperationAdvance =
638       LineTable->Prologue.LineRange != 0
639           ? AdjustedOpcode / LineTable->Prologue.LineRange
640           : 0;
641   uint64_t AddrOffset = advanceAddr(OperationAdvance, Opcode, OpcodeOffset);
642   return {AddrOffset, AdjustedOpcode};
643 }
644 
645 DWARFDebugLine::ParsingState::AddrAndLineDelta
646 DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode,
647                                                   uint64_t OpcodeOffset) {
648   // A special opcode value is chosen based on the amount that needs
649   // to be added to the line and address registers. The maximum line
650   // increment for a special opcode is the value of the line_base
651   // field in the header, plus the value of the line_range field,
652   // minus 1 (line base + line range - 1). If the desired line
653   // increment is greater than the maximum line increment, a standard
654   // opcode must be used instead of a special opcode. The "address
655   // advance" is calculated by dividing the desired address increment
656   // by the minimum_instruction_length field from the header. The
657   // special opcode is then calculated using the following formula:
658   //
659   //  opcode = (desired line increment - line_base) +
660   //           (line_range * address advance) + opcode_base
661   //
662   // If the resulting opcode is greater than 255, a standard opcode
663   // must be used instead.
664   //
665   // To decode a special opcode, subtract the opcode_base from the
666   // opcode itself to give the adjusted opcode. The amount to
667   // increment the address register is the result of the adjusted
668   // opcode divided by the line_range multiplied by the
669   // minimum_instruction_length field from the header. That is:
670   //
671   //  address increment = (adjusted opcode / line_range) *
672   //                      minimum_instruction_length
673   //
674   // The amount to increment the line register is the line_base plus
675   // the result of the adjusted opcode modulo the line_range. That is:
676   //
677   // line increment = line_base + (adjusted opcode % line_range)
678 
679   DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode AddrAdvanceResult =
680       advanceAddrForOpcode(Opcode, OpcodeOffset);
681   int32_t LineOffset = 0;
682   if (LineTable->Prologue.LineRange != 0)
683     LineOffset =
684         LineTable->Prologue.LineBase +
685         (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange);
686   Row.Line += LineOffset;
687   return {AddrAdvanceResult.AddrDelta, LineOffset};
688 }
689 
690 Error DWARFDebugLine::LineTable::parse(
691     DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
692     const DWARFContext &Ctx, const DWARFUnit *U,
693     function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS,
694     bool Verbose) {
695   assert((OS || !Verbose) && "cannot have verbose output without stream");
696   const uint64_t DebugLineOffset = *OffsetPtr;
697 
698   clear();
699 
700   Error PrologueErr =
701       Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U);
702 
703   if (OS) {
704     DIDumpOptions DumpOptions;
705     DumpOptions.Verbose = Verbose;
706     Prologue.dump(*OS, DumpOptions);
707   }
708 
709   if (PrologueErr) {
710     // Ensure there is a blank line after the prologue to clearly delineate it
711     // from later dumps.
712     if (OS)
713       *OS << "\n";
714     return PrologueErr;
715   }
716 
717   uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength();
718   if (!DebugLineData.isValidOffsetForDataOfSize(DebugLineOffset,
719                                                 ProgramLength)) {
720     assert(DebugLineData.size() > DebugLineOffset &&
721            "prologue parsing should handle invalid offset");
722     uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset;
723     RecoverableErrorHandler(
724         createStringError(errc::invalid_argument,
725                           "line table program with offset 0x%8.8" PRIx64
726                           " has length 0x%8.8" PRIx64 " but only 0x%8.8" PRIx64
727                           " bytes are available",
728                           DebugLineOffset, ProgramLength, BytesRemaining));
729     // Continue by capping the length at the number of remaining bytes.
730     ProgramLength = BytesRemaining;
731   }
732 
733   // Create a DataExtractor which can only see the data up to the end of the
734   // table, to prevent reading past the end.
735   const uint64_t EndOffset = DebugLineOffset + ProgramLength;
736   DWARFDataExtractor TableData(DebugLineData, EndOffset);
737 
738   // See if we should tell the data extractor the address size.
739   if (TableData.getAddressSize() == 0)
740     TableData.setAddressSize(Prologue.getAddressSize());
741   else
742     assert(Prologue.getAddressSize() == 0 ||
743            Prologue.getAddressSize() == TableData.getAddressSize());
744 
745   ParsingState State(this, DebugLineOffset, RecoverableErrorHandler);
746 
747   *OffsetPtr = DebugLineOffset + Prologue.getLength();
748   if (OS && *OffsetPtr < EndOffset) {
749     *OS << '\n';
750     Row::dumpTableHeader(*OS, /*Indent=*/Verbose ? 12 : 0);
751   }
752   while (*OffsetPtr < EndOffset) {
753     if (Verbose)
754       *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr);
755 
756     uint64_t OpcodeOffset = *OffsetPtr;
757     uint8_t Opcode = TableData.getU8(OffsetPtr);
758     size_t RowCount = Rows.size();
759 
760     if (Verbose)
761       *OS << format("%02.02" PRIx8 " ", Opcode);
762 
763     if (Opcode == 0) {
764       // Extended Opcodes always start with a zero opcode followed by
765       // a uleb128 length so you can skip ones you don't know about
766       DataExtractor::Cursor Cursor(*OffsetPtr);
767       uint64_t Len = TableData.getULEB128(Cursor);
768       uint64_t ExtOffset = Cursor.tell();
769 
770       // Tolerate zero-length; assume length is correct and soldier on.
771       if (Len == 0) {
772         if (Verbose)
773           *OS << "Badly formed extended line op (length 0)\n";
774         if (!Cursor)
775           RecoverableErrorHandler(Cursor.takeError());
776         *OffsetPtr = Cursor.tell();
777         continue;
778       }
779 
780       uint8_t SubOpcode = TableData.getU8(Cursor);
781       if (Verbose)
782         *OS << LNExtendedString(SubOpcode);
783       switch (SubOpcode) {
784       case DW_LNE_end_sequence:
785         // Set the end_sequence register of the state machine to true and
786         // append a row to the matrix using the current values of the
787         // state-machine registers. Then reset the registers to the initial
788         // values specified above. Every statement program sequence must end
789         // with a DW_LNE_end_sequence instruction which creates a row whose
790         // address is that of the byte after the last target machine instruction
791         // of the sequence.
792         State.Row.EndSequence = true;
793         if (Verbose) {
794           *OS << "\n";
795           OS->indent(12);
796         }
797         if (OS)
798           State.Row.dump(*OS);
799         State.appendRowToMatrix();
800         State.resetRowAndSequence();
801         break;
802 
803       case DW_LNE_set_address:
804         // Takes a single relocatable address as an operand. The size of the
805         // operand is the size appropriate to hold an address on the target
806         // machine. Set the address register to the value given by the
807         // relocatable address. All of the other statement program opcodes
808         // that affect the address register add a delta to it. This instruction
809         // stores a relocatable value into it instead.
810         //
811         // Make sure the extractor knows the address size.  If not, infer it
812         // from the size of the operand.
813         {
814           uint8_t ExtractorAddressSize = TableData.getAddressSize();
815           uint64_t OpcodeAddressSize = Len - 1;
816           if (ExtractorAddressSize != OpcodeAddressSize &&
817               ExtractorAddressSize != 0)
818             RecoverableErrorHandler(createStringError(
819                 errc::invalid_argument,
820                 "mismatching address size at offset 0x%8.8" PRIx64
821                 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64,
822                 ExtOffset, ExtractorAddressSize, Len - 1));
823 
824           // Assume that the line table is correct and temporarily override the
825           // address size. If the size is unsupported, give up trying to read
826           // the address and continue to the next opcode.
827           if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 &&
828               OpcodeAddressSize != 4 && OpcodeAddressSize != 8) {
829             RecoverableErrorHandler(createStringError(
830                 errc::invalid_argument,
831                 "address size 0x%2.2" PRIx64
832                 " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64
833                 " is unsupported",
834                 OpcodeAddressSize, ExtOffset));
835             TableData.skip(Cursor, OpcodeAddressSize);
836           } else {
837             TableData.setAddressSize(OpcodeAddressSize);
838             State.Row.Address.Address = TableData.getRelocatedAddress(
839                 Cursor, &State.Row.Address.SectionIndex);
840 
841             // Restore the address size if the extractor already had it.
842             if (ExtractorAddressSize != 0)
843               TableData.setAddressSize(ExtractorAddressSize);
844           }
845 
846           if (Verbose)
847             *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address.Address);
848         }
849         break;
850 
851       case DW_LNE_define_file:
852         // Takes 4 arguments. The first is a null terminated string containing
853         // a source file name. The second is an unsigned LEB128 number
854         // representing the directory index of the directory in which the file
855         // was found. The third is an unsigned LEB128 number representing the
856         // time of last modification of the file. The fourth is an unsigned
857         // LEB128 number representing the length in bytes of the file. The time
858         // and length fields may contain LEB128(0) if the information is not
859         // available.
860         //
861         // The directory index represents an entry in the include_directories
862         // section of the statement program prologue. The index is LEB128(0)
863         // if the file was found in the current directory of the compilation,
864         // LEB128(1) if it was found in the first directory in the
865         // include_directories section, and so on. The directory index is
866         // ignored for file names that represent full path names.
867         //
868         // The files are numbered, starting at 1, in the order in which they
869         // appear; the names in the prologue come before names defined by
870         // the DW_LNE_define_file instruction. These numbers are used in the
871         // the file register of the state machine.
872         {
873           FileNameEntry FileEntry;
874           const char *Name = TableData.getCStr(Cursor);
875           FileEntry.Name =
876               DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name);
877           FileEntry.DirIdx = TableData.getULEB128(Cursor);
878           FileEntry.ModTime = TableData.getULEB128(Cursor);
879           FileEntry.Length = TableData.getULEB128(Cursor);
880           Prologue.FileNames.push_back(FileEntry);
881           if (Verbose)
882             *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
883                 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
884                 << ", length=" << FileEntry.Length << ")";
885         }
886         break;
887 
888       case DW_LNE_set_discriminator:
889         State.Row.Discriminator = TableData.getULEB128(Cursor);
890         if (Verbose)
891           *OS << " (" << State.Row.Discriminator << ")";
892         break;
893 
894       default:
895         if (Verbose)
896           *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
897               << format(" length %" PRIx64, Len);
898         // Len doesn't include the zero opcode byte or the length itself, but
899         // it does include the sub_opcode, so we have to adjust for that.
900         TableData.skip(Cursor, Len - 1);
901         break;
902       }
903       // Make sure the length as recorded in the table and the standard length
904       // for the opcode match. If they don't, continue from the end as claimed
905       // by the table. Similarly, continue from the claimed end in the event of
906       // a parsing error.
907       uint64_t End = ExtOffset + Len;
908       if (!Cursor)
909         RecoverableErrorHandler(Cursor.takeError());
910       else if (Cursor.tell() != End)
911         RecoverableErrorHandler(createStringError(
912             errc::illegal_byte_sequence,
913             "unexpected line op length at offset 0x%8.8" PRIx64
914             " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64,
915             ExtOffset, Len, Cursor.tell() - ExtOffset));
916       *OffsetPtr = End;
917     } else if (Opcode < Prologue.OpcodeBase) {
918       if (Verbose)
919         *OS << LNStandardString(Opcode);
920       switch (Opcode) {
921       // Standard Opcodes
922       case DW_LNS_copy:
923         // Takes no arguments. Append a row to the matrix using the
924         // current values of the state-machine registers.
925         if (Verbose) {
926           *OS << "\n";
927           OS->indent(12);
928         }
929         if (OS)
930           State.Row.dump(*OS);
931         State.appendRowToMatrix();
932         break;
933 
934       case DW_LNS_advance_pc:
935         // Takes a single unsigned LEB128 operand, multiplies it by the
936         // min_inst_length field of the prologue, and adds the
937         // result to the address register of the state machine.
938         {
939           uint64_t AddrOffset = State.advanceAddr(
940               TableData.getULEB128(OffsetPtr), Opcode, OpcodeOffset);
941           if (Verbose)
942             *OS << " (" << AddrOffset << ")";
943         }
944         break;
945 
946       case DW_LNS_advance_line:
947         // Takes a single signed LEB128 operand and adds that value to
948         // the line register of the state machine.
949         State.Row.Line += TableData.getSLEB128(OffsetPtr);
950         if (Verbose)
951           *OS << " (" << State.Row.Line << ")";
952         break;
953 
954       case DW_LNS_set_file:
955         // Takes a single unsigned LEB128 operand and stores it in the file
956         // register of the state machine.
957         State.Row.File = TableData.getULEB128(OffsetPtr);
958         if (Verbose)
959           *OS << " (" << State.Row.File << ")";
960         break;
961 
962       case DW_LNS_set_column:
963         // Takes a single unsigned LEB128 operand and stores it in the
964         // column register of the state machine.
965         State.Row.Column = TableData.getULEB128(OffsetPtr);
966         if (Verbose)
967           *OS << " (" << State.Row.Column << ")";
968         break;
969 
970       case DW_LNS_negate_stmt:
971         // Takes no arguments. Set the is_stmt register of the state
972         // machine to the logical negation of its current value.
973         State.Row.IsStmt = !State.Row.IsStmt;
974         break;
975 
976       case DW_LNS_set_basic_block:
977         // Takes no arguments. Set the basic_block register of the
978         // state machine to true
979         State.Row.BasicBlock = true;
980         break;
981 
982       case DW_LNS_const_add_pc:
983         // Takes no arguments. Add to the address register of the state
984         // machine the address increment value corresponding to special
985         // opcode 255. The motivation for DW_LNS_const_add_pc is this:
986         // when the statement program needs to advance the address by a
987         // small amount, it can use a single special opcode, which occupies
988         // a single byte. When it needs to advance the address by up to
989         // twice the range of the last special opcode, it can use
990         // DW_LNS_const_add_pc followed by a special opcode, for a total
991         // of two bytes. Only if it needs to advance the address by more
992         // than twice that range will it need to use both DW_LNS_advance_pc
993         // and a special opcode, requiring three or more bytes.
994         {
995           uint64_t AddrOffset =
996               State.advanceAddrForOpcode(Opcode, OpcodeOffset).AddrDelta;
997           if (Verbose)
998             *OS << format(" (0x%16.16" PRIx64 ")", AddrOffset);
999         }
1000         break;
1001 
1002       case DW_LNS_fixed_advance_pc:
1003         // Takes a single uhalf operand. Add to the address register of
1004         // the state machine the value of the (unencoded) operand. This
1005         // is the only extended opcode that takes an argument that is not
1006         // a variable length number. The motivation for DW_LNS_fixed_advance_pc
1007         // is this: existing assemblers cannot emit DW_LNS_advance_pc or
1008         // special opcodes because they cannot encode LEB128 numbers or
1009         // judge when the computation of a special opcode overflows and
1010         // requires the use of DW_LNS_advance_pc. Such assemblers, however,
1011         // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
1012         {
1013           uint16_t PCOffset = TableData.getRelocatedValue(2, OffsetPtr);
1014           State.Row.Address.Address += PCOffset;
1015           if (Verbose)
1016             *OS << format(" (0x%4.4" PRIx16 ")", PCOffset);
1017         }
1018         break;
1019 
1020       case DW_LNS_set_prologue_end:
1021         // Takes no arguments. Set the prologue_end register of the
1022         // state machine to true
1023         State.Row.PrologueEnd = true;
1024         break;
1025 
1026       case DW_LNS_set_epilogue_begin:
1027         // Takes no arguments. Set the basic_block register of the
1028         // state machine to true
1029         State.Row.EpilogueBegin = true;
1030         break;
1031 
1032       case DW_LNS_set_isa:
1033         // Takes a single unsigned LEB128 operand and stores it in the
1034         // column register of the state machine.
1035         State.Row.Isa = TableData.getULEB128(OffsetPtr);
1036         if (Verbose)
1037           *OS << " (" << (uint64_t)State.Row.Isa << ")";
1038         break;
1039 
1040       default:
1041         // Handle any unknown standard opcodes here. We know the lengths
1042         // of such opcodes because they are specified in the prologue
1043         // as a multiple of LEB128 operands for each opcode.
1044         {
1045           assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
1046           if (Verbose)
1047             *OS << "Unrecognized standard opcode";
1048           uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
1049           if (OpcodeLength != 0) {
1050             if (Verbose)
1051               *OS << " (operands: ";
1052             for (uint8_t I = 0; I < OpcodeLength; ++I) {
1053               uint64_t Value = TableData.getULEB128(OffsetPtr);
1054               if (Verbose) {
1055                 if (I > 0)
1056                   *OS << ", ";
1057                 *OS << format("0x%16.16" PRIx64, Value);
1058               }
1059             }
1060             if (Verbose)
1061               *OS << ')';
1062           }
1063         }
1064         break;
1065       }
1066     } else {
1067       // Special Opcodes.
1068       ParsingState::AddrAndLineDelta Delta =
1069           State.handleSpecialOpcode(Opcode, OpcodeOffset);
1070 
1071       if (Verbose) {
1072         *OS << "address += " << Delta.Address << ",  line += " << Delta.Line
1073             << "\n";
1074         OS->indent(12);
1075       }
1076       if (OS)
1077         State.Row.dump(*OS);
1078 
1079       State.appendRowToMatrix();
1080     }
1081 
1082     // When a row is added to the matrix, it is also dumped, which includes a
1083     // new line already, so don't add an extra one.
1084     if (Verbose && Rows.size() == RowCount)
1085       *OS << "\n";
1086   }
1087 
1088   if (!State.Sequence.Empty)
1089     RecoverableErrorHandler(createStringError(
1090         errc::illegal_byte_sequence,
1091         "last sequence in debug line table at offset 0x%8.8" PRIx64
1092         " is not terminated",
1093         DebugLineOffset));
1094 
1095   // Sort all sequences so that address lookup will work faster.
1096   if (!Sequences.empty()) {
1097     llvm::sort(Sequences, Sequence::orderByHighPC);
1098     // Note: actually, instruction address ranges of sequences should not
1099     // overlap (in shared objects and executables). If they do, the address
1100     // lookup would still work, though, but result would be ambiguous.
1101     // We don't report warning in this case. For example,
1102     // sometimes .so compiled from multiple object files contains a few
1103     // rudimentary sequences for address ranges [0x0, 0xsomething).
1104   }
1105 
1106   // Terminate the table with a final blank line to clearly delineate it from
1107   // later dumps.
1108   if (OS)
1109     *OS << "\n";
1110 
1111   return Error::success();
1112 }
1113 
1114 uint32_t DWARFDebugLine::LineTable::findRowInSeq(
1115     const DWARFDebugLine::Sequence &Seq,
1116     object::SectionedAddress Address) const {
1117   if (!Seq.containsPC(Address))
1118     return UnknownRowIndex;
1119   assert(Seq.SectionIndex == Address.SectionIndex);
1120   // In some cases, e.g. first instruction in a function, the compiler generates
1121   // two entries, both with the same address. We want the last one.
1122   //
1123   // In general we want a non-empty range: the last row whose address is less
1124   // than or equal to Address. This can be computed as upper_bound - 1.
1125   DWARFDebugLine::Row Row;
1126   Row.Address = Address;
1127   RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
1128   RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
1129   assert(FirstRow->Address.Address <= Row.Address.Address &&
1130          Row.Address.Address < LastRow[-1].Address.Address);
1131   RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row,
1132                                     DWARFDebugLine::Row::orderByAddress) -
1133                    1;
1134   assert(Seq.SectionIndex == RowPos->Address.SectionIndex);
1135   return RowPos - Rows.begin();
1136 }
1137 
1138 uint32_t DWARFDebugLine::LineTable::lookupAddress(
1139     object::SectionedAddress Address) const {
1140 
1141   // Search for relocatable addresses
1142   uint32_t Result = lookupAddressImpl(Address);
1143 
1144   if (Result != UnknownRowIndex ||
1145       Address.SectionIndex == object::SectionedAddress::UndefSection)
1146     return Result;
1147 
1148   // Search for absolute addresses
1149   Address.SectionIndex = object::SectionedAddress::UndefSection;
1150   return lookupAddressImpl(Address);
1151 }
1152 
1153 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl(
1154     object::SectionedAddress Address) const {
1155   // First, find an instruction sequence containing the given address.
1156   DWARFDebugLine::Sequence Sequence;
1157   Sequence.SectionIndex = Address.SectionIndex;
1158   Sequence.HighPC = Address.Address;
1159   SequenceIter It = llvm::upper_bound(Sequences, Sequence,
1160                                       DWARFDebugLine::Sequence::orderByHighPC);
1161   if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex)
1162     return UnknownRowIndex;
1163   return findRowInSeq(*It, Address);
1164 }
1165 
1166 bool DWARFDebugLine::LineTable::lookupAddressRange(
1167     object::SectionedAddress Address, uint64_t Size,
1168     std::vector<uint32_t> &Result) const {
1169 
1170   // Search for relocatable addresses
1171   if (lookupAddressRangeImpl(Address, Size, Result))
1172     return true;
1173 
1174   if (Address.SectionIndex == object::SectionedAddress::UndefSection)
1175     return false;
1176 
1177   // Search for absolute addresses
1178   Address.SectionIndex = object::SectionedAddress::UndefSection;
1179   return lookupAddressRangeImpl(Address, Size, Result);
1180 }
1181 
1182 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
1183     object::SectionedAddress Address, uint64_t Size,
1184     std::vector<uint32_t> &Result) const {
1185   if (Sequences.empty())
1186     return false;
1187   uint64_t EndAddr = Address.Address + Size;
1188   // First, find an instruction sequence containing the given address.
1189   DWARFDebugLine::Sequence Sequence;
1190   Sequence.SectionIndex = Address.SectionIndex;
1191   Sequence.HighPC = Address.Address;
1192   SequenceIter LastSeq = Sequences.end();
1193   SequenceIter SeqPos = llvm::upper_bound(
1194       Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC);
1195   if (SeqPos == LastSeq || !SeqPos->containsPC(Address))
1196     return false;
1197 
1198   SequenceIter StartPos = SeqPos;
1199 
1200   // Add the rows from the first sequence to the vector, starting with the
1201   // index we just calculated
1202 
1203   while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
1204     const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
1205     // For the first sequence, we need to find which row in the sequence is the
1206     // first in our range.
1207     uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
1208     if (SeqPos == StartPos)
1209       FirstRowIndex = findRowInSeq(CurSeq, Address);
1210 
1211     // Figure out the last row in the range.
1212     uint32_t LastRowIndex =
1213         findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
1214     if (LastRowIndex == UnknownRowIndex)
1215       LastRowIndex = CurSeq.LastRowIndex - 1;
1216 
1217     assert(FirstRowIndex != UnknownRowIndex);
1218     assert(LastRowIndex != UnknownRowIndex);
1219 
1220     for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
1221       Result.push_back(I);
1222     }
1223 
1224     ++SeqPos;
1225   }
1226 
1227   return true;
1228 }
1229 
1230 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
1231                                                                 FileLineInfoKind Kind) const {
1232   if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex))
1233     return None;
1234   const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex);
1235   if (Optional<const char *> source = Entry.Source.getAsCString())
1236     return StringRef(*source);
1237   return None;
1238 }
1239 
1240 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
1241   // Debug info can contain paths from any OS, not necessarily
1242   // an OS we're currently running on. Moreover different compilation units can
1243   // be compiled on different operating systems and linked together later.
1244   return sys::path::is_absolute(Path, sys::path::Style::posix) ||
1245          sys::path::is_absolute(Path, sys::path::Style::windows);
1246 }
1247 
1248 bool DWARFDebugLine::Prologue::getFileNameByIndex(
1249     uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind,
1250     std::string &Result, sys::path::Style Style) const {
1251   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
1252     return false;
1253   const FileNameEntry &Entry = getFileNameEntry(FileIndex);
1254   Optional<const char *> Name = Entry.Name.getAsCString();
1255   if (!Name)
1256     return false;
1257   StringRef FileName = *Name;
1258   if (Kind == FileLineInfoKind::RawValue ||
1259       isPathAbsoluteOnWindowsOrPosix(FileName)) {
1260     Result = std::string(FileName);
1261     return true;
1262   }
1263   if (Kind == FileLineInfoKind::BaseNameOnly) {
1264     Result = std::string(llvm::sys::path::filename(FileName));
1265     return true;
1266   }
1267 
1268   SmallString<16> FilePath;
1269   StringRef IncludeDir;
1270   // Be defensive about the contents of Entry.
1271   if (getVersion() >= 5) {
1272     // DirIdx 0 is the compilation directory, so don't include it for
1273     // relative names.
1274     if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) &&
1275         Entry.DirIdx < IncludeDirectories.size())
1276       IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue();
1277   } else {
1278     if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size())
1279       IncludeDir =
1280           IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue();
1281   }
1282 
1283   // For absolute paths only, include the compilation directory of compile unit.
1284   // We know that FileName is not absolute, the only way to have an absolute
1285   // path at this point would be if IncludeDir is absolute.
1286   if (Kind == FileLineInfoKind::AbsoluteFilePath && !CompDir.empty() &&
1287       !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1288     sys::path::append(FilePath, Style, CompDir);
1289 
1290   assert((Kind == FileLineInfoKind::AbsoluteFilePath ||
1291           Kind == FileLineInfoKind::RelativeFilePath) &&
1292          "invalid FileLineInfo Kind");
1293 
1294   // sys::path::append skips empty strings.
1295   sys::path::append(FilePath, Style, IncludeDir, FileName);
1296   Result = std::string(FilePath.str());
1297   return true;
1298 }
1299 
1300 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1301     object::SectionedAddress Address, const char *CompDir,
1302     FileLineInfoKind Kind, DILineInfo &Result) const {
1303   // Get the index of row we're looking for in the line table.
1304   uint32_t RowIndex = lookupAddress(Address);
1305   if (RowIndex == -1U)
1306     return false;
1307   // Take file number and line/column from the row.
1308   const auto &Row = Rows[RowIndex];
1309   if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1310     return false;
1311   Result.Line = Row.Line;
1312   Result.Column = Row.Column;
1313   Result.Discriminator = Row.Discriminator;
1314   Result.Source = getSourceByIndex(Row.File, Kind);
1315   return true;
1316 }
1317 
1318 // We want to supply the Unit associated with a .debug_line[.dwo] table when
1319 // we dump it, if possible, but still dump the table even if there isn't a Unit.
1320 // Therefore, collect up handles on all the Units that point into the
1321 // line-table section.
1322 static DWARFDebugLine::SectionParser::LineToUnitMap
1323 buildLineToUnitMap(DWARFDebugLine::SectionParser::cu_range CUs,
1324                    DWARFDebugLine::SectionParser::tu_range TUs) {
1325   DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit;
1326   for (const auto &CU : CUs)
1327     if (auto CUDIE = CU->getUnitDIE())
1328       if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list)))
1329         LineToUnit.insert(std::make_pair(*StmtOffset, &*CU));
1330   for (const auto &TU : TUs)
1331     if (auto TUDIE = TU->getUnitDIE())
1332       if (auto StmtOffset = toSectionOffset(TUDIE.find(DW_AT_stmt_list)))
1333         LineToUnit.insert(std::make_pair(*StmtOffset, &*TU));
1334   return LineToUnit;
1335 }
1336 
1337 DWARFDebugLine::SectionParser::SectionParser(DWARFDataExtractor &Data,
1338                                              const DWARFContext &C,
1339                                              cu_range CUs, tu_range TUs)
1340     : DebugLineData(Data), Context(C) {
1341   LineToUnit = buildLineToUnitMap(CUs, TUs);
1342   if (!DebugLineData.isValidOffset(Offset))
1343     Done = true;
1344 }
1345 
1346 bool DWARFDebugLine::Prologue::totalLengthIsValid() const {
1347   return TotalLength != 0u;
1348 }
1349 
1350 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext(
1351     function_ref<void(Error)> RecoverableErrorHandler,
1352     function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS,
1353     bool Verbose) {
1354   assert(DebugLineData.isValidOffset(Offset) &&
1355          "parsing should have terminated");
1356   DWARFUnit *U = prepareToParse(Offset);
1357   uint64_t OldOffset = Offset;
1358   LineTable LT;
1359   if (Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1360                            RecoverableErrorHandler, OS, Verbose))
1361     UnrecoverableErrorHandler(std::move(Err));
1362   moveToNextTable(OldOffset, LT.Prologue);
1363   return LT;
1364 }
1365 
1366 void DWARFDebugLine::SectionParser::skip(
1367     function_ref<void(Error)> RecoverableErrorHandler,
1368     function_ref<void(Error)> UnrecoverableErrorHandler) {
1369   assert(DebugLineData.isValidOffset(Offset) &&
1370          "parsing should have terminated");
1371   DWARFUnit *U = prepareToParse(Offset);
1372   uint64_t OldOffset = Offset;
1373   LineTable LT;
1374   if (Error Err = LT.Prologue.parse(DebugLineData, &Offset,
1375                                     RecoverableErrorHandler, Context, U))
1376     UnrecoverableErrorHandler(std::move(Err));
1377   moveToNextTable(OldOffset, LT.Prologue);
1378 }
1379 
1380 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) {
1381   DWARFUnit *U = nullptr;
1382   auto It = LineToUnit.find(Offset);
1383   if (It != LineToUnit.end())
1384     U = It->second;
1385   DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0);
1386   return U;
1387 }
1388 
1389 void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset,
1390                                                     const Prologue &P) {
1391   // If the length field is not valid, we don't know where the next table is, so
1392   // cannot continue to parse. Mark the parser as done, and leave the Offset
1393   // value as it currently is. This will be the end of the bad length field.
1394   if (!P.totalLengthIsValid()) {
1395     Done = true;
1396     return;
1397   }
1398 
1399   Offset = OldOffset + P.TotalLength + P.sizeofTotalLength();
1400   if (!DebugLineData.isValidOffset(Offset)) {
1401     Done = true;
1402   }
1403 }
1404