1 //===- DWARFDebugLine.cpp -------------------------------------------------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 
10 #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h"
11 #include "llvm/ADT/Optional.h"
12 #include "llvm/ADT/SmallString.h"
13 #include "llvm/ADT/SmallVector.h"
14 #include "llvm/ADT/StringRef.h"
15 #include "llvm/BinaryFormat/Dwarf.h"
16 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
17 #include "llvm/DebugInfo/DWARF/DWARFRelocMap.h"
18 #include "llvm/Support/Format.h"
19 #include "llvm/Support/Path.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 anonmyous namespace
44 
45 void DWARFDebugLine::ContentTypeTracker::trackContentType(
46     dwarf::LineNumberEntryFormat ContentType) {
47   switch (ContentType) {
48   case dwarf::DW_LNCT_timestamp:
49     HasModTime = true;
50     break;
51   case dwarf::DW_LNCT_size:
52     HasLength = true;
53     break;
54   case dwarf::DW_LNCT_MD5:
55     HasMD5 = true;
56     break;
57   case dwarf::DW_LNCT_LLVM_source:
58     HasSource = true;
59     break;
60   default:
61     // We only care about values we consider optional, and new values may be
62     // added in the vendor extension range, so we do not match exhaustively.
63     break;
64   }
65 }
66 
67 DWARFDebugLine::Prologue::Prologue() { clear(); }
68 
69 void DWARFDebugLine::Prologue::clear() {
70   TotalLength = PrologueLength = 0;
71   SegSelectorSize = 0;
72   MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
73   OpcodeBase = 0;
74   FormParams = dwarf::FormParams({0, 0, DWARF32});
75   ContentTypes = ContentTypeTracker();
76   StandardOpcodeLengths.clear();
77   IncludeDirectories.clear();
78   FileNames.clear();
79 }
80 
81 void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
82                                     DIDumpOptions DumpOptions) const {
83   OS << "Line table prologue:\n"
84      << format("    total_length: 0x%8.8" PRIx64 "\n", TotalLength)
85      << format("         version: %u\n", getVersion());
86   if (getVersion() >= 5)
87     OS << format("    address_size: %u\n", getAddressSize())
88        << format(" seg_select_size: %u\n", SegSelectorSize);
89   OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength)
90      << format(" min_inst_length: %u\n", MinInstLength)
91      << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
92      << format(" default_is_stmt: %u\n", DefaultIsStmt)
93      << format("       line_base: %i\n", LineBase)
94      << format("      line_range: %u\n", LineRange)
95      << format("     opcode_base: %u\n", OpcodeBase);
96 
97   for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
98     OS << format("standard_opcode_lengths[%s] = %u\n",
99                  LNStandardString(I + 1).data(), StandardOpcodeLengths[I]);
100 
101   if (!IncludeDirectories.empty()) {
102     // DWARF v5 starts directory indexes at 0.
103     uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
104     for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
105       OS << format("include_directories[%3u] = ", I + DirBase);
106       IncludeDirectories[I].dump(OS, DumpOptions);
107       OS << '\n';
108     }
109   }
110 
111   if (!FileNames.empty()) {
112     // DWARF v5 starts file indexes at 0.
113     uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
114     for (uint32_t I = 0; I != FileNames.size(); ++I) {
115       const FileNameEntry &FileEntry = FileNames[I];
116       OS <<   format("file_names[%3u]:\n", I + FileBase);
117       OS <<          "           name: ";
118       FileEntry.Name.dump(OS, DumpOptions);
119       OS << '\n'
120          <<   format("      dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
121       if (ContentTypes.HasMD5)
122         OS <<        "   md5_checksum: " << FileEntry.Checksum.digest() << '\n';
123       if (ContentTypes.HasModTime)
124         OS << format("       mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
125       if (ContentTypes.HasLength)
126         OS << format("         length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
127       if (ContentTypes.HasSource) {
128         OS <<        "         source: ";
129         FileEntry.Source.dump(OS, DumpOptions);
130         OS << '\n';
131       }
132     }
133   }
134 }
135 
136 // Parse v2-v4 directory and file tables.
137 static void
138 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
139                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
140                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
141                      std::vector<DWARFFormValue> &IncludeDirectories,
142                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
143   while (*OffsetPtr < EndPrologueOffset) {
144     StringRef S = DebugLineData.getCStrRef(OffsetPtr);
145     if (S.empty())
146       break;
147     DWARFFormValue Dir(dwarf::DW_FORM_string);
148     Dir.setPValue(S.data());
149     IncludeDirectories.push_back(Dir);
150   }
151 
152   while (*OffsetPtr < EndPrologueOffset) {
153     StringRef Name = DebugLineData.getCStrRef(OffsetPtr);
154     if (Name.empty())
155       break;
156     DWARFDebugLine::FileNameEntry FileEntry;
157     FileEntry.Name.setForm(dwarf::DW_FORM_string);
158     FileEntry.Name.setPValue(Name.data());
159     FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
160     FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
161     FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
162     FileNames.push_back(FileEntry);
163   }
164 
165   ContentTypes.HasModTime = true;
166   ContentTypes.HasLength = true;
167 }
168 
169 // Parse v5 directory/file entry content descriptions.
170 // Returns the descriptors, or an empty vector if we did not find a path or
171 // ran off the end of the prologue.
172 static ContentDescriptors
173 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint32_t
174     *OffsetPtr, uint64_t EndPrologueOffset, DWARFDebugLine::ContentTypeTracker
175     *ContentTypes) {
176   ContentDescriptors Descriptors;
177   int FormatCount = DebugLineData.getU8(OffsetPtr);
178   bool HasPath = false;
179   for (int I = 0; I != FormatCount; ++I) {
180     if (*OffsetPtr >= EndPrologueOffset)
181       return ContentDescriptors();
182     ContentDescriptor Descriptor;
183     Descriptor.Type =
184       dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr));
185     Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr));
186     if (Descriptor.Type == dwarf::DW_LNCT_path)
187       HasPath = true;
188     if (ContentTypes)
189       ContentTypes->trackContentType(Descriptor.Type);
190     Descriptors.push_back(Descriptor);
191   }
192   return HasPath ? Descriptors : ContentDescriptors();
193 }
194 
195 static bool
196 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
197                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
198                      const dwarf::FormParams &FormParams,
199                      const DWARFContext &Ctx, const DWARFUnit *U,
200                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
201                      std::vector<DWARFFormValue> &IncludeDirectories,
202                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
203   // Get the directory entry description.
204   ContentDescriptors DirDescriptors =
205       parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset, nullptr);
206   if (DirDescriptors.empty())
207     return false;
208 
209   // Get the directory entries, according to the format described above.
210   int DirEntryCount = DebugLineData.getU8(OffsetPtr);
211   for (int I = 0; I != DirEntryCount; ++I) {
212     if (*OffsetPtr >= EndPrologueOffset)
213       return false;
214     for (auto Descriptor : DirDescriptors) {
215       DWARFFormValue Value(Descriptor.Form);
216       switch (Descriptor.Type) {
217       case DW_LNCT_path:
218         if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
219           return false;
220         IncludeDirectories.push_back(Value);
221         break;
222       default:
223         if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
224           return false;
225       }
226     }
227   }
228 
229   // Get the file entry description.
230   ContentDescriptors FileDescriptors =
231       parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset,
232           &ContentTypes);
233   if (FileDescriptors.empty())
234     return false;
235 
236   // Get the file entries, according to the format described above.
237   int FileEntryCount = DebugLineData.getU8(OffsetPtr);
238   for (int I = 0; I != FileEntryCount; ++I) {
239     if (*OffsetPtr >= EndPrologueOffset)
240       return false;
241     DWARFDebugLine::FileNameEntry FileEntry;
242     for (auto Descriptor : FileDescriptors) {
243       DWARFFormValue Value(Descriptor.Form);
244       if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
245         return false;
246       switch (Descriptor.Type) {
247       case DW_LNCT_path:
248         FileEntry.Name = Value;
249         break;
250       case DW_LNCT_LLVM_source:
251         FileEntry.Source = Value;
252         break;
253       case DW_LNCT_directory_index:
254         FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
255         break;
256       case DW_LNCT_timestamp:
257         FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
258         break;
259       case DW_LNCT_size:
260         FileEntry.Length = Value.getAsUnsignedConstant().getValue();
261         break;
262       case DW_LNCT_MD5:
263         assert(Value.getAsBlock().getValue().size() == 16);
264         std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16,
265                                   FileEntry.Checksum.Bytes.begin());
266         break;
267       default:
268         break;
269       }
270     }
271     FileNames.push_back(FileEntry);
272   }
273   return true;
274 }
275 
276 bool DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData,
277                                      uint32_t *OffsetPtr,
278                                      const DWARFContext &Ctx,
279                                      const DWARFUnit *U) {
280   const uint64_t PrologueOffset = *OffsetPtr;
281 
282   clear();
283   TotalLength = DebugLineData.getU32(OffsetPtr);
284   if (TotalLength == UINT32_MAX) {
285     FormParams.Format = dwarf::DWARF64;
286     TotalLength = DebugLineData.getU64(OffsetPtr);
287   } else if (TotalLength >= 0xffffff00) {
288     return false;
289   }
290   FormParams.Version = DebugLineData.getU16(OffsetPtr);
291   if (getVersion() < 2)
292     return false;
293 
294   if (getVersion() >= 5) {
295     FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
296     assert((DebugLineData.getAddressSize() == 0 ||
297             DebugLineData.getAddressSize() == getAddressSize()) &&
298            "Line table header and data extractor disagree");
299     SegSelectorSize = DebugLineData.getU8(OffsetPtr);
300   }
301 
302   PrologueLength = DebugLineData.getUnsigned(OffsetPtr, sizeofPrologueLength());
303   const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
304   MinInstLength = DebugLineData.getU8(OffsetPtr);
305   if (getVersion() >= 4)
306     MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
307   DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
308   LineBase = DebugLineData.getU8(OffsetPtr);
309   LineRange = DebugLineData.getU8(OffsetPtr);
310   OpcodeBase = DebugLineData.getU8(OffsetPtr);
311 
312   StandardOpcodeLengths.reserve(OpcodeBase - 1);
313   for (uint32_t I = 1; I < OpcodeBase; ++I) {
314     uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
315     StandardOpcodeLengths.push_back(OpLen);
316   }
317 
318   if (getVersion() >= 5) {
319     if (!parseV5DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
320                               FormParams, Ctx, U, ContentTypes,
321                               IncludeDirectories, FileNames)) {
322       WithColor::warning() << format(
323           "parsing line table prologue at 0x%8.8" PRIx64
324           " found an invalid directory or file table description at"
325           " 0x%8.8" PRIx64 "\n",
326           PrologueOffset, (uint64_t)*OffsetPtr);
327       return false;
328     }
329   } else
330     parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
331                          ContentTypes, IncludeDirectories, FileNames);
332 
333   if (*OffsetPtr != EndPrologueOffset) {
334     WithColor::warning() << format(
335         "parsing line table prologue at 0x%8.8" PRIx64
336         " should have ended at 0x%8.8" PRIx64 " but it ended at 0x%8.8" PRIx64
337         "\n",
338         PrologueOffset, EndPrologueOffset, (uint64_t)*OffsetPtr);
339     return false;
340   }
341   return true;
342 }
343 
344 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
345 
346 void DWARFDebugLine::Row::postAppend() {
347   BasicBlock = false;
348   PrologueEnd = false;
349   EpilogueBegin = false;
350 }
351 
352 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
353   Address = 0;
354   Line = 1;
355   Column = 0;
356   File = 1;
357   Isa = 0;
358   Discriminator = 0;
359   IsStmt = DefaultIsStmt;
360   BasicBlock = false;
361   EndSequence = false;
362   PrologueEnd = false;
363   EpilogueBegin = false;
364 }
365 
366 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) {
367   OS << "Address            Line   Column File   ISA Discriminator Flags\n"
368      << "------------------ ------ ------ ------ --- ------------- "
369         "-------------\n";
370 }
371 
372 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
373   OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column)
374      << format(" %6u %3u %13u ", File, Isa, Discriminator)
375      << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
376      << (PrologueEnd ? " prologue_end" : "")
377      << (EpilogueBegin ? " epilogue_begin" : "")
378      << (EndSequence ? " end_sequence" : "") << '\n';
379 }
380 
381 DWARFDebugLine::Sequence::Sequence() { reset(); }
382 
383 void DWARFDebugLine::Sequence::reset() {
384   LowPC = 0;
385   HighPC = 0;
386   FirstRowIndex = 0;
387   LastRowIndex = 0;
388   Empty = true;
389 }
390 
391 DWARFDebugLine::LineTable::LineTable() { clear(); }
392 
393 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
394                                      DIDumpOptions DumpOptions) const {
395   Prologue.dump(OS, DumpOptions);
396   OS << '\n';
397 
398   if (!Rows.empty()) {
399     Row::dumpTableHeader(OS);
400     for (const Row &R : Rows) {
401       R.dump(OS);
402     }
403   }
404 }
405 
406 void DWARFDebugLine::LineTable::clear() {
407   Prologue.clear();
408   Rows.clear();
409   Sequences.clear();
410 }
411 
412 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT)
413     : LineTable(LT) {
414   resetRowAndSequence();
415 }
416 
417 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
418   Row.reset(LineTable->Prologue.DefaultIsStmt);
419   Sequence.reset();
420 }
421 
422 void DWARFDebugLine::ParsingState::appendRowToMatrix(uint32_t Offset) {
423   if (Sequence.Empty) {
424     // Record the beginning of instruction sequence.
425     Sequence.Empty = false;
426     Sequence.LowPC = Row.Address;
427     Sequence.FirstRowIndex = RowNumber;
428   }
429   ++RowNumber;
430   LineTable->appendRow(Row);
431   if (Row.EndSequence) {
432     // Record the end of instruction sequence.
433     Sequence.HighPC = Row.Address;
434     Sequence.LastRowIndex = RowNumber;
435     if (Sequence.isValid())
436       LineTable->appendSequence(Sequence);
437     Sequence.reset();
438   }
439   Row.postAppend();
440 }
441 
442 const DWARFDebugLine::LineTable *
443 DWARFDebugLine::getLineTable(uint32_t Offset) const {
444   LineTableConstIter Pos = LineTableMap.find(Offset);
445   if (Pos != LineTableMap.end())
446     return &Pos->second;
447   return nullptr;
448 }
449 
450 const DWARFDebugLine::LineTable *
451 DWARFDebugLine::getOrParseLineTable(DWARFDataExtractor &DebugLineData,
452                                     uint32_t Offset, const DWARFContext &Ctx,
453                                     const DWARFUnit *U) {
454   if (!DebugLineData.isValidOffset(Offset))
455     return nullptr;
456 
457   std::pair<LineTableIter, bool> Pos =
458       LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
459   LineTable *LT = &Pos.first->second;
460   if (Pos.second) {
461     if (!LT->parse(DebugLineData, &Offset, Ctx, U))
462       return nullptr;
463   }
464   return LT;
465 }
466 
467 bool DWARFDebugLine::LineTable::parse(DWARFDataExtractor &DebugLineData,
468                                       uint32_t *OffsetPtr,
469                                       const DWARFContext &Ctx,
470                                       const DWARFUnit *U, raw_ostream *OS) {
471   const uint32_t DebugLineOffset = *OffsetPtr;
472 
473   clear();
474 
475   if (!Prologue.parse(DebugLineData, OffsetPtr, Ctx, U)) {
476     // Restore our offset and return false to indicate failure!
477     *OffsetPtr = DebugLineOffset;
478     return false;
479   }
480 
481   if (OS) {
482     // The presence of OS signals verbose dumping.
483     DIDumpOptions DumpOptions;
484     DumpOptions.Verbose = true;
485     Prologue.dump(*OS, DumpOptions);
486   }
487 
488   const uint32_t EndOffset =
489       DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength();
490 
491   // See if we should tell the data extractor the address size.
492   if (DebugLineData.getAddressSize() == 0)
493     DebugLineData.setAddressSize(Prologue.getAddressSize());
494   else
495     assert(Prologue.getAddressSize() == 0 ||
496            Prologue.getAddressSize() == DebugLineData.getAddressSize());
497 
498   ParsingState State(this);
499 
500   while (*OffsetPtr < EndOffset) {
501     if (OS)
502       *OS << format("0x%08.08" PRIx32 ": ", *OffsetPtr);
503 
504     uint8_t Opcode = DebugLineData.getU8(OffsetPtr);
505 
506     if (OS)
507       *OS << format("%02.02" PRIx8 " ", Opcode);
508 
509     if (Opcode == 0) {
510       // Extended Opcodes always start with a zero opcode followed by
511       // a uleb128 length so you can skip ones you don't know about
512       uint64_t Len = DebugLineData.getULEB128(OffsetPtr);
513       uint32_t ExtOffset = *OffsetPtr;
514 
515       // Tolerate zero-length; assume length is correct and soldier on.
516       if (Len == 0) {
517         if (OS)
518           *OS << "Badly formed extended line op (length 0)\n";
519         continue;
520       }
521 
522       uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr);
523       if (OS)
524         *OS << LNExtendedString(SubOpcode);
525       switch (SubOpcode) {
526       case DW_LNE_end_sequence:
527         // Set the end_sequence register of the state machine to true and
528         // append a row to the matrix using the current values of the
529         // state-machine registers. Then reset the registers to the initial
530         // values specified above. Every statement program sequence must end
531         // with a DW_LNE_end_sequence instruction which creates a row whose
532         // address is that of the byte after the last target machine instruction
533         // of the sequence.
534         State.Row.EndSequence = true;
535         State.appendRowToMatrix(*OffsetPtr);
536         if (OS) {
537           *OS << "\n";
538           OS->indent(12);
539           State.Row.dump(*OS);
540         }
541         State.resetRowAndSequence();
542         break;
543 
544       case DW_LNE_set_address:
545         // Takes a single relocatable address as an operand. The size of the
546         // operand is the size appropriate to hold an address on the target
547         // machine. Set the address register to the value given by the
548         // relocatable address. All of the other statement program opcodes
549         // that affect the address register add a delta to it. This instruction
550         // stores a relocatable value into it instead.
551         //
552         // Make sure the extractor knows the address size.  If not, infer it
553         // from the size of the operand.
554         if (DebugLineData.getAddressSize() == 0)
555           DebugLineData.setAddressSize(Len - 1);
556         else if (DebugLineData.getAddressSize() != Len - 1) {
557           WithColor::warning()
558               << format("mismatching address size at offset 0x%8.8" PRIx32
559                         " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64 "\n",
560                         ExtOffset, DebugLineData.getAddressSize(), Len - 1);
561           // Skip the rest of the line-number program.
562           *OffsetPtr = EndOffset;
563           return false;
564         }
565         State.Row.Address = DebugLineData.getRelocatedAddress(OffsetPtr);
566         if (OS)
567           *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address);
568         break;
569 
570       case DW_LNE_define_file:
571         // Takes 4 arguments. The first is a null terminated string containing
572         // a source file name. The second is an unsigned LEB128 number
573         // representing the directory index of the directory in which the file
574         // was found. The third is an unsigned LEB128 number representing the
575         // time of last modification of the file. The fourth is an unsigned
576         // LEB128 number representing the length in bytes of the file. The time
577         // and length fields may contain LEB128(0) if the information is not
578         // available.
579         //
580         // The directory index represents an entry in the include_directories
581         // section of the statement program prologue. The index is LEB128(0)
582         // if the file was found in the current directory of the compilation,
583         // LEB128(1) if it was found in the first directory in the
584         // include_directories section, and so on. The directory index is
585         // ignored for file names that represent full path names.
586         //
587         // The files are numbered, starting at 1, in the order in which they
588         // appear; the names in the prologue come before names defined by
589         // the DW_LNE_define_file instruction. These numbers are used in the
590         // the file register of the state machine.
591         {
592           FileNameEntry FileEntry;
593           const char *Name = DebugLineData.getCStr(OffsetPtr);
594           FileEntry.Name.setForm(dwarf::DW_FORM_string);
595           FileEntry.Name.setPValue(Name);
596           FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
597           FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
598           FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
599           Prologue.FileNames.push_back(FileEntry);
600           if (OS)
601             *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
602                 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
603                 << ", length=" << FileEntry.Length << ")";
604         }
605         break;
606 
607       case DW_LNE_set_discriminator:
608         State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr);
609         if (OS)
610           *OS << " (" << State.Row.Discriminator << ")";
611         break;
612 
613       default:
614         if (OS)
615           *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
616               << format(" length %" PRIx64, Len);
617         // Len doesn't include the zero opcode byte or the length itself, but
618         // it does include the sub_opcode, so we have to adjust for that.
619         (*OffsetPtr) += Len - 1;
620         break;
621       }
622       // Make sure the stated and parsed lengths are the same.
623       // Otherwise we have an unparseable line-number program.
624       if (*OffsetPtr - ExtOffset != Len) {
625         WithColor::warning()
626             << format("unexpected line op length at offset 0x%8.8" PRIx32
627                       " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx32 "\n",
628                       ExtOffset, Len, *OffsetPtr - ExtOffset);
629         // Skip the rest of the line-number program.
630         *OffsetPtr = EndOffset;
631         return false;
632       }
633     } else if (Opcode < Prologue.OpcodeBase) {
634       if (OS)
635         *OS << LNStandardString(Opcode);
636       switch (Opcode) {
637       // Standard Opcodes
638       case DW_LNS_copy:
639         // Takes no arguments. Append a row to the matrix using the
640         // current values of the state-machine registers. Then set
641         // the basic_block register to false.
642         State.appendRowToMatrix(*OffsetPtr);
643         if (OS) {
644           *OS << "\n";
645           OS->indent(12);
646           State.Row.dump(*OS);
647           *OS << "\n";
648         }
649         break;
650 
651       case DW_LNS_advance_pc:
652         // Takes a single unsigned LEB128 operand, multiplies it by the
653         // min_inst_length field of the prologue, and adds the
654         // result to the address register of the state machine.
655         {
656           uint64_t AddrOffset =
657               DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength;
658           State.Row.Address += AddrOffset;
659           if (OS)
660             *OS << " (" << AddrOffset << ")";
661         }
662         break;
663 
664       case DW_LNS_advance_line:
665         // Takes a single signed LEB128 operand and adds that value to
666         // the line register of the state machine.
667         State.Row.Line += DebugLineData.getSLEB128(OffsetPtr);
668         if (OS)
669           *OS << " (" << State.Row.Line << ")";
670         break;
671 
672       case DW_LNS_set_file:
673         // Takes a single unsigned LEB128 operand and stores it in the file
674         // register of the state machine.
675         State.Row.File = DebugLineData.getULEB128(OffsetPtr);
676         if (OS)
677           *OS << " (" << State.Row.File << ")";
678         break;
679 
680       case DW_LNS_set_column:
681         // Takes a single unsigned LEB128 operand and stores it in the
682         // column register of the state machine.
683         State.Row.Column = DebugLineData.getULEB128(OffsetPtr);
684         if (OS)
685           *OS << " (" << State.Row.Column << ")";
686         break;
687 
688       case DW_LNS_negate_stmt:
689         // Takes no arguments. Set the is_stmt register of the state
690         // machine to the logical negation of its current value.
691         State.Row.IsStmt = !State.Row.IsStmt;
692         break;
693 
694       case DW_LNS_set_basic_block:
695         // Takes no arguments. Set the basic_block register of the
696         // state machine to true
697         State.Row.BasicBlock = true;
698         break;
699 
700       case DW_LNS_const_add_pc:
701         // Takes no arguments. Add to the address register of the state
702         // machine the address increment value corresponding to special
703         // opcode 255. The motivation for DW_LNS_const_add_pc is this:
704         // when the statement program needs to advance the address by a
705         // small amount, it can use a single special opcode, which occupies
706         // a single byte. When it needs to advance the address by up to
707         // twice the range of the last special opcode, it can use
708         // DW_LNS_const_add_pc followed by a special opcode, for a total
709         // of two bytes. Only if it needs to advance the address by more
710         // than twice that range will it need to use both DW_LNS_advance_pc
711         // and a special opcode, requiring three or more bytes.
712         {
713           uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase;
714           uint64_t AddrOffset =
715               (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
716           State.Row.Address += AddrOffset;
717           if (OS)
718             *OS
719                 << format(" (0x%16.16" PRIx64 ")", AddrOffset);
720         }
721         break;
722 
723       case DW_LNS_fixed_advance_pc:
724         // Takes a single uhalf operand. Add to the address register of
725         // the state machine the value of the (unencoded) operand. This
726         // is the only extended opcode that takes an argument that is not
727         // a variable length number. The motivation for DW_LNS_fixed_advance_pc
728         // is this: existing assemblers cannot emit DW_LNS_advance_pc or
729         // special opcodes because they cannot encode LEB128 numbers or
730         // judge when the computation of a special opcode overflows and
731         // requires the use of DW_LNS_advance_pc. Such assemblers, however,
732         // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
733         {
734           uint16_t PCOffset = DebugLineData.getU16(OffsetPtr);
735           State.Row.Address += PCOffset;
736           if (OS)
737             *OS
738                 << format(" (0x%16.16" PRIx64 ")", PCOffset);
739         }
740         break;
741 
742       case DW_LNS_set_prologue_end:
743         // Takes no arguments. Set the prologue_end register of the
744         // state machine to true
745         State.Row.PrologueEnd = true;
746         break;
747 
748       case DW_LNS_set_epilogue_begin:
749         // Takes no arguments. Set the basic_block register of the
750         // state machine to true
751         State.Row.EpilogueBegin = true;
752         break;
753 
754       case DW_LNS_set_isa:
755         // Takes a single unsigned LEB128 operand and stores it in the
756         // column register of the state machine.
757         State.Row.Isa = DebugLineData.getULEB128(OffsetPtr);
758         if (OS)
759           *OS << " (" << State.Row.Isa << ")";
760         break;
761 
762       default:
763         // Handle any unknown standard opcodes here. We know the lengths
764         // of such opcodes because they are specified in the prologue
765         // as a multiple of LEB128 operands for each opcode.
766         {
767           assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
768           uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
769           for (uint8_t I = 0; I < OpcodeLength; ++I) {
770             uint64_t Value = DebugLineData.getULEB128(OffsetPtr);
771             if (OS)
772               *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n",
773                             Value);
774           }
775         }
776         break;
777       }
778     } else {
779       // Special Opcodes
780 
781       // A special opcode value is chosen based on the amount that needs
782       // to be added to the line and address registers. The maximum line
783       // increment for a special opcode is the value of the line_base
784       // field in the header, plus the value of the line_range field,
785       // minus 1 (line base + line range - 1). If the desired line
786       // increment is greater than the maximum line increment, a standard
787       // opcode must be used instead of a special opcode. The "address
788       // advance" is calculated by dividing the desired address increment
789       // by the minimum_instruction_length field from the header. The
790       // special opcode is then calculated using the following formula:
791       //
792       //  opcode = (desired line increment - line_base) +
793       //           (line_range * address advance) + opcode_base
794       //
795       // If the resulting opcode is greater than 255, a standard opcode
796       // must be used instead.
797       //
798       // To decode a special opcode, subtract the opcode_base from the
799       // opcode itself to give the adjusted opcode. The amount to
800       // increment the address register is the result of the adjusted
801       // opcode divided by the line_range multiplied by the
802       // minimum_instruction_length field from the header. That is:
803       //
804       //  address increment = (adjusted opcode / line_range) *
805       //                      minimum_instruction_length
806       //
807       // The amount to increment the line register is the line_base plus
808       // the result of the adjusted opcode modulo the line_range. That is:
809       //
810       // line increment = line_base + (adjusted opcode % line_range)
811 
812       uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase;
813       uint64_t AddrOffset =
814           (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
815       int32_t LineOffset =
816           Prologue.LineBase + (AdjustOpcode % Prologue.LineRange);
817       State.Row.Line += LineOffset;
818       State.Row.Address += AddrOffset;
819 
820       if (OS) {
821         *OS << "address += " << ((uint32_t)AdjustOpcode)
822             << ",  line += " << LineOffset << "\n";
823         OS->indent(12);
824         State.Row.dump(*OS);
825       }
826 
827       State.appendRowToMatrix(*OffsetPtr);
828       // Reset discriminator to 0.
829       State.Row.Discriminator = 0;
830     }
831     if(OS)
832       *OS << "\n";
833   }
834 
835   if (!State.Sequence.Empty)
836     WithColor::warning() << "last sequence in debug line table is not"
837                             "terminated!\n";
838 
839   // Sort all sequences so that address lookup will work faster.
840   if (!Sequences.empty()) {
841     llvm::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC);
842     // Note: actually, instruction address ranges of sequences should not
843     // overlap (in shared objects and executables). If they do, the address
844     // lookup would still work, though, but result would be ambiguous.
845     // We don't report warning in this case. For example,
846     // sometimes .so compiled from multiple object files contains a few
847     // rudimentary sequences for address ranges [0x0, 0xsomething).
848   }
849 
850   return EndOffset;
851 }
852 
853 uint32_t
854 DWARFDebugLine::LineTable::findRowInSeq(const DWARFDebugLine::Sequence &Seq,
855                                         uint64_t Address) const {
856   if (!Seq.containsPC(Address))
857     return UnknownRowIndex;
858   // Search for instruction address in the rows describing the sequence.
859   // Rows are stored in a vector, so we may use arithmetical operations with
860   // iterators.
861   DWARFDebugLine::Row Row;
862   Row.Address = Address;
863   RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
864   RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
865   LineTable::RowIter RowPos = std::lower_bound(
866       FirstRow, LastRow, Row, DWARFDebugLine::Row::orderByAddress);
867   if (RowPos == LastRow) {
868     return Seq.LastRowIndex - 1;
869   }
870   uint32_t Index = Seq.FirstRowIndex + (RowPos - FirstRow);
871   if (RowPos->Address > Address) {
872     if (RowPos == FirstRow)
873       return UnknownRowIndex;
874     else
875       Index--;
876   }
877   return Index;
878 }
879 
880 uint32_t DWARFDebugLine::LineTable::lookupAddress(uint64_t Address) const {
881   if (Sequences.empty())
882     return UnknownRowIndex;
883   // First, find an instruction sequence containing the given address.
884   DWARFDebugLine::Sequence Sequence;
885   Sequence.LowPC = Address;
886   SequenceIter FirstSeq = Sequences.begin();
887   SequenceIter LastSeq = Sequences.end();
888   SequenceIter SeqPos = std::lower_bound(
889       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
890   DWARFDebugLine::Sequence FoundSeq;
891   if (SeqPos == LastSeq) {
892     FoundSeq = Sequences.back();
893   } else if (SeqPos->LowPC == Address) {
894     FoundSeq = *SeqPos;
895   } else {
896     if (SeqPos == FirstSeq)
897       return UnknownRowIndex;
898     FoundSeq = *(SeqPos - 1);
899   }
900   return findRowInSeq(FoundSeq, Address);
901 }
902 
903 bool DWARFDebugLine::LineTable::lookupAddressRange(
904     uint64_t Address, uint64_t Size, std::vector<uint32_t> &Result) const {
905   if (Sequences.empty())
906     return false;
907   uint64_t EndAddr = Address + Size;
908   // First, find an instruction sequence containing the given address.
909   DWARFDebugLine::Sequence Sequence;
910   Sequence.LowPC = Address;
911   SequenceIter FirstSeq = Sequences.begin();
912   SequenceIter LastSeq = Sequences.end();
913   SequenceIter SeqPos = std::lower_bound(
914       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
915   if (SeqPos == LastSeq || SeqPos->LowPC != Address) {
916     if (SeqPos == FirstSeq)
917       return false;
918     SeqPos--;
919   }
920   if (!SeqPos->containsPC(Address))
921     return false;
922 
923   SequenceIter StartPos = SeqPos;
924 
925   // Add the rows from the first sequence to the vector, starting with the
926   // index we just calculated
927 
928   while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
929     const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
930     // For the first sequence, we need to find which row in the sequence is the
931     // first in our range.
932     uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
933     if (SeqPos == StartPos)
934       FirstRowIndex = findRowInSeq(CurSeq, Address);
935 
936     // Figure out the last row in the range.
937     uint32_t LastRowIndex = findRowInSeq(CurSeq, EndAddr - 1);
938     if (LastRowIndex == UnknownRowIndex)
939       LastRowIndex = CurSeq.LastRowIndex - 1;
940 
941     assert(FirstRowIndex != UnknownRowIndex);
942     assert(LastRowIndex != UnknownRowIndex);
943 
944     for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
945       Result.push_back(I);
946     }
947 
948     ++SeqPos;
949   }
950 
951   return true;
952 }
953 
954 bool DWARFDebugLine::LineTable::hasFileAtIndex(uint64_t FileIndex) const {
955   return FileIndex != 0 && FileIndex <= Prologue.FileNames.size();
956 }
957 
958 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
959                                                                 FileLineInfoKind Kind) const {
960   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
961     return None;
962   const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
963   if (Optional<const char *> source = Entry.Source.getAsCString())
964     return StringRef(*source);
965   return None;
966 }
967 
968 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
969   // Debug info can contain paths from any OS, not necessarily
970   // an OS we're currently running on. Moreover different compilation units can
971   // be compiled on different operating systems and linked together later.
972   return sys::path::is_absolute(Path, sys::path::Style::posix) ||
973          sys::path::is_absolute(Path, sys::path::Style::windows);
974 }
975 
976 bool DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex,
977                                                    const char *CompDir,
978                                                    FileLineInfoKind Kind,
979                                                    std::string &Result) const {
980   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
981     return false;
982   const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
983   StringRef FileName = Entry.Name.getAsCString().getValue();
984   if (Kind != FileLineInfoKind::AbsoluteFilePath ||
985       isPathAbsoluteOnWindowsOrPosix(FileName)) {
986     Result = FileName;
987     return true;
988   }
989 
990   SmallString<16> FilePath;
991   uint64_t IncludeDirIndex = Entry.DirIdx;
992   StringRef IncludeDir;
993   // Be defensive about the contents of Entry.
994   if (IncludeDirIndex > 0 &&
995       IncludeDirIndex <= Prologue.IncludeDirectories.size())
996     IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1]
997                      .getAsCString()
998                      .getValue();
999 
1000   // We may still need to append compilation directory of compile unit.
1001   // We know that FileName is not absolute, the only way to have an
1002   // absolute path at this point would be if IncludeDir is absolute.
1003   if (CompDir && Kind == FileLineInfoKind::AbsoluteFilePath &&
1004       !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1005     sys::path::append(FilePath, CompDir);
1006 
1007   // sys::path::append skips empty strings.
1008   sys::path::append(FilePath, IncludeDir, FileName);
1009   Result = FilePath.str();
1010   return true;
1011 }
1012 
1013 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1014     uint64_t Address, const char *CompDir, FileLineInfoKind Kind,
1015     DILineInfo &Result) const {
1016   // Get the index of row we're looking for in the line table.
1017   uint32_t RowIndex = lookupAddress(Address);
1018   if (RowIndex == -1U)
1019     return false;
1020   // Take file number and line/column from the row.
1021   const auto &Row = Rows[RowIndex];
1022   if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1023     return false;
1024   Result.Line = Row.Line;
1025   Result.Column = Row.Column;
1026   Result.Discriminator = Row.Discriminator;
1027   Result.Source = getSourceByIndex(Row.File, Kind);
1028   return true;
1029 }
1030