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/raw_ostream.h"
21 #include <algorithm>
22 #include <cassert>
23 #include <cinttypes>
24 #include <cstdint>
25 #include <cstdio>
26 #include <utility>
27 
28 using namespace llvm;
29 using namespace dwarf;
30 
31 using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind;
32 
33 namespace {
34 
35 struct ContentDescriptor {
36   dwarf::LineNumberEntryFormat Type;
37   dwarf::Form Form;
38 };
39 
40 using ContentDescriptors = SmallVector<ContentDescriptor, 4>;
41 
42 } // end anonmyous namespace
43 
44 void DWARFDebugLine::ContentTypeTracker::trackContentType(
45     dwarf::LineNumberEntryFormat ContentType) {
46   switch (ContentType) {
47   case dwarf::DW_LNCT_timestamp:
48     HasModTime = true;
49     break;
50   case dwarf::DW_LNCT_size:
51     HasLength = true;
52     break;
53   case dwarf::DW_LNCT_MD5:
54     HasMD5 = true;
55     break;
56   case dwarf::DW_LNCT_LLVM_source:
57     HasSource = true;
58     break;
59   default:
60     // We only care about values we consider optional, and new values may be
61     // added in the vendor extension range, so we do not match exhaustively.
62     break;
63   }
64 }
65 
66 DWARFDebugLine::Prologue::Prologue() { clear(); }
67 
68 void DWARFDebugLine::Prologue::clear() {
69   TotalLength = PrologueLength = 0;
70   SegSelectorSize = 0;
71   MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
72   OpcodeBase = 0;
73   FormParams = DWARFFormParams({0, 0, DWARF32});
74   ContentTypes = ContentTypeTracker();
75   StandardOpcodeLengths.clear();
76   IncludeDirectories.clear();
77   FileNames.clear();
78 }
79 
80 void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
81                                     DIDumpOptions DumpOptions) const {
82   OS << "Line table prologue:\n"
83      << format("    total_length: 0x%8.8" PRIx64 "\n", TotalLength)
84      << format("         version: %u\n", getVersion());
85   if (getVersion() >= 5)
86     OS << format("    address_size: %u\n", getAddressSize())
87        << format(" seg_select_size: %u\n", SegSelectorSize);
88   OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength)
89      << format(" min_inst_length: %u\n", MinInstLength)
90      << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
91      << format(" default_is_stmt: %u\n", DefaultIsStmt)
92      << format("       line_base: %i\n", LineBase)
93      << format("      line_range: %u\n", LineRange)
94      << format("     opcode_base: %u\n", OpcodeBase);
95 
96   for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
97     OS << format("standard_opcode_lengths[%s] = %u\n",
98                  LNStandardString(I + 1).data(), StandardOpcodeLengths[I]);
99 
100   if (!IncludeDirectories.empty()) {
101     // DWARF v5 starts directory indexes at 0.
102     uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
103     for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
104       OS << format("include_directories[%3u] = ", I + DirBase);
105       IncludeDirectories[I].dump(OS, DumpOptions);
106       OS << '\n';
107     }
108   }
109 
110   if (!FileNames.empty()) {
111     // DWARF v5 starts file indexes at 0.
112     uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
113     for (uint32_t I = 0; I != FileNames.size(); ++I) {
114       const FileNameEntry &FileEntry = FileNames[I];
115       OS <<   format("file_names[%3u]:\n", I + FileBase);
116       OS <<          "           name: ";
117       FileEntry.Name.dump(OS, DumpOptions);
118       OS << '\n'
119          <<   format("      dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
120       if (ContentTypes.HasMD5)
121         OS <<        "   md5_checksum: " << FileEntry.Checksum.digest() << '\n';
122       if (ContentTypes.HasModTime)
123         OS << format("       mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
124       if (ContentTypes.HasLength)
125         OS << format("         length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
126       if (ContentTypes.HasSource) {
127         OS <<        "         source: ";
128         FileEntry.Source.dump(OS, DumpOptions);
129         OS << '\n';
130       }
131     }
132   }
133 }
134 
135 // Parse v2-v4 directory and file tables.
136 static void
137 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
138                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
139                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
140                      std::vector<DWARFFormValue> &IncludeDirectories,
141                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
142   while (*OffsetPtr < EndPrologueOffset) {
143     StringRef S = DebugLineData.getCStrRef(OffsetPtr);
144     if (S.empty())
145       break;
146     DWARFFormValue Dir(dwarf::DW_FORM_string);
147     Dir.setPValue(S.data());
148     IncludeDirectories.push_back(Dir);
149   }
150 
151   while (*OffsetPtr < EndPrologueOffset) {
152     StringRef Name = DebugLineData.getCStrRef(OffsetPtr);
153     if (Name.empty())
154       break;
155     DWARFDebugLine::FileNameEntry FileEntry;
156     FileEntry.Name.setForm(dwarf::DW_FORM_string);
157     FileEntry.Name.setPValue(Name.data());
158     FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
159     FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
160     FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
161     FileNames.push_back(FileEntry);
162   }
163 
164   ContentTypes.HasModTime = true;
165   ContentTypes.HasLength = true;
166 }
167 
168 // Parse v5 directory/file entry content descriptions.
169 // Returns the descriptors, or an empty vector if we did not find a path or
170 // ran off the end of the prologue.
171 static ContentDescriptors
172 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint32_t
173     *OffsetPtr, uint64_t EndPrologueOffset, DWARFDebugLine::ContentTypeTracker
174     *ContentTypes) {
175   ContentDescriptors Descriptors;
176   int FormatCount = DebugLineData.getU8(OffsetPtr);
177   bool HasPath = false;
178   for (int I = 0; I != FormatCount; ++I) {
179     if (*OffsetPtr >= EndPrologueOffset)
180       return ContentDescriptors();
181     ContentDescriptor Descriptor;
182     Descriptor.Type =
183       dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr));
184     Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr));
185     if (Descriptor.Type == dwarf::DW_LNCT_path)
186       HasPath = true;
187     if (ContentTypes)
188       ContentTypes->trackContentType(Descriptor.Type);
189     Descriptors.push_back(Descriptor);
190   }
191   return HasPath ? Descriptors : ContentDescriptors();
192 }
193 
194 static bool
195 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
196                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
197                      const DWARFFormParams &FormParams, const DWARFContext
198                      &Ctx, const DWARFUnit *U,
199                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
200                      std::vector<DWARFFormValue> &IncludeDirectories,
201                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
202   // Get the directory entry description.
203   ContentDescriptors DirDescriptors =
204       parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset, nullptr);
205   if (DirDescriptors.empty())
206     return false;
207 
208   // Get the directory entries, according to the format described above.
209   int DirEntryCount = DebugLineData.getU8(OffsetPtr);
210   for (int I = 0; I != DirEntryCount; ++I) {
211     if (*OffsetPtr >= EndPrologueOffset)
212       return false;
213     for (auto Descriptor : DirDescriptors) {
214       DWARFFormValue Value(Descriptor.Form);
215       switch (Descriptor.Type) {
216       case DW_LNCT_path:
217         if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
218           return false;
219         IncludeDirectories.push_back(Value);
220         break;
221       default:
222         if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
223           return false;
224       }
225     }
226   }
227 
228   // Get the file entry description.
229   ContentDescriptors FileDescriptors =
230       parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset,
231           &ContentTypes);
232   if (FileDescriptors.empty())
233     return false;
234 
235   // Get the file entries, according to the format described above.
236   int FileEntryCount = DebugLineData.getU8(OffsetPtr);
237   for (int I = 0; I != FileEntryCount; ++I) {
238     if (*OffsetPtr >= EndPrologueOffset)
239       return false;
240     DWARFDebugLine::FileNameEntry FileEntry;
241     for (auto Descriptor : FileDescriptors) {
242       DWARFFormValue Value(Descriptor.Form);
243       if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
244         return false;
245       switch (Descriptor.Type) {
246       case DW_LNCT_path:
247         FileEntry.Name = Value;
248         break;
249       case DW_LNCT_LLVM_source:
250         FileEntry.Source = Value;
251         break;
252       case DW_LNCT_directory_index:
253         FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
254         break;
255       case DW_LNCT_timestamp:
256         FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
257         break;
258       case DW_LNCT_size:
259         FileEntry.Length = Value.getAsUnsignedConstant().getValue();
260         break;
261       case DW_LNCT_MD5:
262         assert(Value.getAsBlock().getValue().size() == 16);
263         std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16,
264                                   FileEntry.Checksum.Bytes.begin());
265         break;
266       default:
267         break;
268       }
269     }
270     FileNames.push_back(FileEntry);
271   }
272   return true;
273 }
274 
275 bool DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData,
276                                      uint32_t *OffsetPtr,
277                                      const DWARFContext &Ctx,
278                                      const DWARFUnit *U) {
279   const uint64_t PrologueOffset = *OffsetPtr;
280 
281   clear();
282   TotalLength = DebugLineData.getU32(OffsetPtr);
283   if (TotalLength == UINT32_MAX) {
284     FormParams.Format = dwarf::DWARF64;
285     TotalLength = DebugLineData.getU64(OffsetPtr);
286   } else if (TotalLength >= 0xffffff00) {
287     return false;
288   }
289   FormParams.Version = DebugLineData.getU16(OffsetPtr);
290   if (getVersion() < 2)
291     return false;
292 
293   if (getVersion() >= 5) {
294     FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
295     assert((DebugLineData.getAddressSize() == 0 ||
296             DebugLineData.getAddressSize() == getAddressSize()) &&
297            "Line table header and data extractor disagree");
298     SegSelectorSize = DebugLineData.getU8(OffsetPtr);
299   }
300 
301   PrologueLength = DebugLineData.getUnsigned(OffsetPtr, sizeofPrologueLength());
302   const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
303   MinInstLength = DebugLineData.getU8(OffsetPtr);
304   if (getVersion() >= 4)
305     MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
306   DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
307   LineBase = DebugLineData.getU8(OffsetPtr);
308   LineRange = DebugLineData.getU8(OffsetPtr);
309   OpcodeBase = DebugLineData.getU8(OffsetPtr);
310 
311   StandardOpcodeLengths.reserve(OpcodeBase - 1);
312   for (uint32_t I = 1; I < OpcodeBase; ++I) {
313     uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
314     StandardOpcodeLengths.push_back(OpLen);
315   }
316 
317   if (getVersion() >= 5) {
318     if (!parseV5DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
319                               FormParams, Ctx, U, ContentTypes,
320                               IncludeDirectories, FileNames)) {
321       fprintf(stderr,
322               "warning: parsing line table prologue at 0x%8.8" PRIx64
323               " found an invalid directory or file table description at"
324               " 0x%8.8" PRIx64 "\n", PrologueOffset, (uint64_t)*OffsetPtr);
325       return false;
326     }
327   } else
328     parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
329                          ContentTypes, IncludeDirectories, FileNames);
330 
331   if (*OffsetPtr != EndPrologueOffset) {
332     fprintf(stderr,
333             "warning: parsing line table prologue at 0x%8.8" PRIx64
334             " should have ended at 0x%8.8" PRIx64
335             " but it ended at 0x%8.8" PRIx64 "\n",
336             PrologueOffset, EndPrologueOffset, (uint64_t)*OffsetPtr);
337     return false;
338   }
339   return true;
340 }
341 
342 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
343 
344 void DWARFDebugLine::Row::postAppend() {
345   BasicBlock = false;
346   PrologueEnd = false;
347   EpilogueBegin = false;
348 }
349 
350 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
351   Address = 0;
352   Line = 1;
353   Column = 0;
354   File = 1;
355   Isa = 0;
356   Discriminator = 0;
357   IsStmt = DefaultIsStmt;
358   BasicBlock = false;
359   EndSequence = false;
360   PrologueEnd = false;
361   EpilogueBegin = false;
362 }
363 
364 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) {
365   OS << "Address            Line   Column File   ISA Discriminator Flags\n"
366      << "------------------ ------ ------ ------ --- ------------- "
367         "-------------\n";
368 }
369 
370 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
371   OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column)
372      << format(" %6u %3u %13u ", File, Isa, Discriminator)
373      << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
374      << (PrologueEnd ? " prologue_end" : "")
375      << (EpilogueBegin ? " epilogue_begin" : "")
376      << (EndSequence ? " end_sequence" : "") << '\n';
377 }
378 
379 DWARFDebugLine::Sequence::Sequence() { reset(); }
380 
381 void DWARFDebugLine::Sequence::reset() {
382   LowPC = 0;
383   HighPC = 0;
384   FirstRowIndex = 0;
385   LastRowIndex = 0;
386   Empty = true;
387 }
388 
389 DWARFDebugLine::LineTable::LineTable() { clear(); }
390 
391 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
392                                      DIDumpOptions DumpOptions) const {
393   Prologue.dump(OS, DumpOptions);
394   OS << '\n';
395 
396   if (!Rows.empty()) {
397     Row::dumpTableHeader(OS);
398     for (const Row &R : Rows) {
399       R.dump(OS);
400     }
401   }
402 }
403 
404 void DWARFDebugLine::LineTable::clear() {
405   Prologue.clear();
406   Rows.clear();
407   Sequences.clear();
408 }
409 
410 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT)
411     : LineTable(LT) {
412   resetRowAndSequence();
413 }
414 
415 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
416   Row.reset(LineTable->Prologue.DefaultIsStmt);
417   Sequence.reset();
418 }
419 
420 void DWARFDebugLine::ParsingState::appendRowToMatrix(uint32_t Offset) {
421   if (Sequence.Empty) {
422     // Record the beginning of instruction sequence.
423     Sequence.Empty = false;
424     Sequence.LowPC = Row.Address;
425     Sequence.FirstRowIndex = RowNumber;
426   }
427   ++RowNumber;
428   LineTable->appendRow(Row);
429   if (Row.EndSequence) {
430     // Record the end of instruction sequence.
431     Sequence.HighPC = Row.Address;
432     Sequence.LastRowIndex = RowNumber;
433     if (Sequence.isValid())
434       LineTable->appendSequence(Sequence);
435     Sequence.reset();
436   }
437   Row.postAppend();
438 }
439 
440 const DWARFDebugLine::LineTable *
441 DWARFDebugLine::getLineTable(uint32_t Offset) const {
442   LineTableConstIter Pos = LineTableMap.find(Offset);
443   if (Pos != LineTableMap.end())
444     return &Pos->second;
445   return nullptr;
446 }
447 
448 const DWARFDebugLine::LineTable *
449 DWARFDebugLine::getOrParseLineTable(DWARFDataExtractor &DebugLineData,
450                                     uint32_t Offset, const DWARFContext &Ctx,
451                                     const DWARFUnit *U) {
452   std::pair<LineTableIter, bool> Pos =
453       LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
454   LineTable *LT = &Pos.first->second;
455   if (Pos.second) {
456     if (!LT->parse(DebugLineData, &Offset, Ctx, U))
457       return nullptr;
458   }
459   return LT;
460 }
461 
462 bool DWARFDebugLine::LineTable::parse(DWARFDataExtractor &DebugLineData,
463                                       uint32_t *OffsetPtr,
464                                       const DWARFContext &Ctx,
465                                       const DWARFUnit *U, raw_ostream *OS) {
466   const uint32_t DebugLineOffset = *OffsetPtr;
467 
468   clear();
469 
470   if (!Prologue.parse(DebugLineData, OffsetPtr, Ctx, U)) {
471     // Restore our offset and return false to indicate failure!
472     *OffsetPtr = DebugLineOffset;
473     return false;
474   }
475 
476   if (OS) {
477     // The presence of OS signals verbose dumping.
478     DIDumpOptions DumpOptions;
479     DumpOptions.Verbose = true;
480     Prologue.dump(*OS, DumpOptions);
481   }
482 
483   const uint32_t EndOffset =
484       DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength();
485 
486   // See if we should tell the data extractor the address size.
487   if (DebugLineData.getAddressSize() == 0)
488     DebugLineData.setAddressSize(Prologue.getAddressSize());
489   else
490     assert(Prologue.getAddressSize() == 0 ||
491            Prologue.getAddressSize() == DebugLineData.getAddressSize());
492 
493   ParsingState State(this);
494 
495   while (*OffsetPtr < EndOffset) {
496     if (OS)
497       *OS << format("0x%08.08" PRIx32 ": ", *OffsetPtr);
498 
499     uint8_t Opcode = DebugLineData.getU8(OffsetPtr);
500 
501     if (OS)
502       *OS << format("%02.02" PRIx8 " ", Opcode);
503 
504     if (Opcode == 0) {
505       // Extended Opcodes always start with a zero opcode followed by
506       // a uleb128 length so you can skip ones you don't know about
507       uint64_t Len = DebugLineData.getULEB128(OffsetPtr);
508       uint32_t ExtOffset = *OffsetPtr;
509 
510       // Tolerate zero-length; assume length is correct and soldier on.
511       if (Len == 0) {
512         if (OS)
513           *OS << "Badly formed extended line op (length 0)\n";
514         continue;
515       }
516 
517       uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr);
518       if (OS)
519         *OS << LNExtendedString(SubOpcode);
520       switch (SubOpcode) {
521       case DW_LNE_end_sequence:
522         // Set the end_sequence register of the state machine to true and
523         // append a row to the matrix using the current values of the
524         // state-machine registers. Then reset the registers to the initial
525         // values specified above. Every statement program sequence must end
526         // with a DW_LNE_end_sequence instruction which creates a row whose
527         // address is that of the byte after the last target machine instruction
528         // of the sequence.
529         State.Row.EndSequence = true;
530         State.appendRowToMatrix(*OffsetPtr);
531         if (OS) {
532           *OS << "\n";
533           OS->indent(12);
534           State.Row.dump(*OS);
535         }
536         State.resetRowAndSequence();
537         break;
538 
539       case DW_LNE_set_address:
540         // Takes a single relocatable address as an operand. The size of the
541         // operand is the size appropriate to hold an address on the target
542         // machine. Set the address register to the value given by the
543         // relocatable address. All of the other statement program opcodes
544         // that affect the address register add a delta to it. This instruction
545         // stores a relocatable value into it instead.
546         //
547         // Make sure the extractor knows the address size.  If not, infer it
548         // from the size of the operand.
549         if (DebugLineData.getAddressSize() == 0)
550           DebugLineData.setAddressSize(Len - 1);
551         else
552           assert(DebugLineData.getAddressSize() == Len - 1);
553         State.Row.Address = DebugLineData.getRelocatedAddress(OffsetPtr);
554         if (OS)
555           *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address);
556         break;
557 
558       case DW_LNE_define_file:
559         // Takes 4 arguments. The first is a null terminated string containing
560         // a source file name. The second is an unsigned LEB128 number
561         // representing the directory index of the directory in which the file
562         // was found. The third is an unsigned LEB128 number representing the
563         // time of last modification of the file. The fourth is an unsigned
564         // LEB128 number representing the length in bytes of the file. The time
565         // and length fields may contain LEB128(0) if the information is not
566         // available.
567         //
568         // The directory index represents an entry in the include_directories
569         // section of the statement program prologue. The index is LEB128(0)
570         // if the file was found in the current directory of the compilation,
571         // LEB128(1) if it was found in the first directory in the
572         // include_directories section, and so on. The directory index is
573         // ignored for file names that represent full path names.
574         //
575         // The files are numbered, starting at 1, in the order in which they
576         // appear; the names in the prologue come before names defined by
577         // the DW_LNE_define_file instruction. These numbers are used in the
578         // the file register of the state machine.
579         {
580           FileNameEntry FileEntry;
581           const char *Name = DebugLineData.getCStr(OffsetPtr);
582           FileEntry.Name.setForm(dwarf::DW_FORM_string);
583           FileEntry.Name.setPValue(Name);
584           FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
585           FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
586           FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
587           Prologue.FileNames.push_back(FileEntry);
588           if (OS)
589             *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
590                 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
591                 << ", length=" << FileEntry.Length << ")";
592         }
593         break;
594 
595       case DW_LNE_set_discriminator:
596         State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr);
597         if (OS)
598           *OS << " (" << State.Row.Discriminator << ")";
599         break;
600 
601       default:
602         if (OS)
603           *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
604               << format(" length %" PRIx64, Len);
605         // Len doesn't include the zero opcode byte or the length itself, but
606         // it does include the sub_opcode, so we have to adjust for that.
607         (*OffsetPtr) += Len - 1;
608         break;
609       }
610       // Make sure the stated and parsed lengths are the same.
611       // Otherwise we have an unparseable line-number program.
612       if (*OffsetPtr - ExtOffset != Len) {
613         fprintf(stderr, "Unexpected line op length at offset 0x%8.8" PRIx32
614                 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx32 "\n",
615                 ExtOffset, Len, *OffsetPtr - ExtOffset);
616         // Skip the rest of the line-number program.
617         *OffsetPtr = EndOffset;
618         return false;
619       }
620     } else if (Opcode < Prologue.OpcodeBase) {
621       if (OS)
622         *OS << LNStandardString(Opcode);
623       switch (Opcode) {
624       // Standard Opcodes
625       case DW_LNS_copy:
626         // Takes no arguments. Append a row to the matrix using the
627         // current values of the state-machine registers. Then set
628         // the basic_block register to false.
629         State.appendRowToMatrix(*OffsetPtr);
630         if (OS) {
631           *OS << "\n";
632           OS->indent(12);
633           State.Row.dump(*OS);
634           *OS << "\n";
635         }
636         break;
637 
638       case DW_LNS_advance_pc:
639         // Takes a single unsigned LEB128 operand, multiplies it by the
640         // min_inst_length field of the prologue, and adds the
641         // result to the address register of the state machine.
642         {
643           uint64_t AddrOffset =
644               DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength;
645           State.Row.Address += AddrOffset;
646           if (OS)
647             *OS << " (" << AddrOffset << ")";
648         }
649         break;
650 
651       case DW_LNS_advance_line:
652         // Takes a single signed LEB128 operand and adds that value to
653         // the line register of the state machine.
654         State.Row.Line += DebugLineData.getSLEB128(OffsetPtr);
655         if (OS)
656           *OS << " (" << State.Row.Line << ")";
657         break;
658 
659       case DW_LNS_set_file:
660         // Takes a single unsigned LEB128 operand and stores it in the file
661         // register of the state machine.
662         State.Row.File = DebugLineData.getULEB128(OffsetPtr);
663         if (OS)
664           *OS << " (" << State.Row.File << ")";
665         break;
666 
667       case DW_LNS_set_column:
668         // Takes a single unsigned LEB128 operand and stores it in the
669         // column register of the state machine.
670         State.Row.Column = DebugLineData.getULEB128(OffsetPtr);
671         if (OS)
672           *OS << " (" << State.Row.Column << ")";
673         break;
674 
675       case DW_LNS_negate_stmt:
676         // Takes no arguments. Set the is_stmt register of the state
677         // machine to the logical negation of its current value.
678         State.Row.IsStmt = !State.Row.IsStmt;
679         break;
680 
681       case DW_LNS_set_basic_block:
682         // Takes no arguments. Set the basic_block register of the
683         // state machine to true
684         State.Row.BasicBlock = true;
685         break;
686 
687       case DW_LNS_const_add_pc:
688         // Takes no arguments. Add to the address register of the state
689         // machine the address increment value corresponding to special
690         // opcode 255. The motivation for DW_LNS_const_add_pc is this:
691         // when the statement program needs to advance the address by a
692         // small amount, it can use a single special opcode, which occupies
693         // a single byte. When it needs to advance the address by up to
694         // twice the range of the last special opcode, it can use
695         // DW_LNS_const_add_pc followed by a special opcode, for a total
696         // of two bytes. Only if it needs to advance the address by more
697         // than twice that range will it need to use both DW_LNS_advance_pc
698         // and a special opcode, requiring three or more bytes.
699         {
700           uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase;
701           uint64_t AddrOffset =
702               (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
703           State.Row.Address += AddrOffset;
704           if (OS)
705             *OS
706                 << format(" (0x%16.16" PRIx64 ")", AddrOffset);
707         }
708         break;
709 
710       case DW_LNS_fixed_advance_pc:
711         // Takes a single uhalf operand. Add to the address register of
712         // the state machine the value of the (unencoded) operand. This
713         // is the only extended opcode that takes an argument that is not
714         // a variable length number. The motivation for DW_LNS_fixed_advance_pc
715         // is this: existing assemblers cannot emit DW_LNS_advance_pc or
716         // special opcodes because they cannot encode LEB128 numbers or
717         // judge when the computation of a special opcode overflows and
718         // requires the use of DW_LNS_advance_pc. Such assemblers, however,
719         // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
720         {
721           uint16_t PCOffset = DebugLineData.getU16(OffsetPtr);
722           State.Row.Address += PCOffset;
723           if (OS)
724             *OS
725                 << format(" (0x%16.16" PRIx64 ")", PCOffset);
726         }
727         break;
728 
729       case DW_LNS_set_prologue_end:
730         // Takes no arguments. Set the prologue_end register of the
731         // state machine to true
732         State.Row.PrologueEnd = true;
733         break;
734 
735       case DW_LNS_set_epilogue_begin:
736         // Takes no arguments. Set the basic_block register of the
737         // state machine to true
738         State.Row.EpilogueBegin = true;
739         break;
740 
741       case DW_LNS_set_isa:
742         // Takes a single unsigned LEB128 operand and stores it in the
743         // column register of the state machine.
744         State.Row.Isa = DebugLineData.getULEB128(OffsetPtr);
745         if (OS)
746           *OS << " (" << State.Row.Isa << ")";
747         break;
748 
749       default:
750         // Handle any unknown standard opcodes here. We know the lengths
751         // of such opcodes because they are specified in the prologue
752         // as a multiple of LEB128 operands for each opcode.
753         {
754           assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
755           uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
756           for (uint8_t I = 0; I < OpcodeLength; ++I) {
757             uint64_t Value = DebugLineData.getULEB128(OffsetPtr);
758             if (OS)
759               *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n",
760                             Value);
761           }
762         }
763         break;
764       }
765     } else {
766       // Special Opcodes
767 
768       // A special opcode value is chosen based on the amount that needs
769       // to be added to the line and address registers. The maximum line
770       // increment for a special opcode is the value of the line_base
771       // field in the header, plus the value of the line_range field,
772       // minus 1 (line base + line range - 1). If the desired line
773       // increment is greater than the maximum line increment, a standard
774       // opcode must be used instead of a special opcode. The "address
775       // advance" is calculated by dividing the desired address increment
776       // by the minimum_instruction_length field from the header. The
777       // special opcode is then calculated using the following formula:
778       //
779       //  opcode = (desired line increment - line_base) +
780       //           (line_range * address advance) + opcode_base
781       //
782       // If the resulting opcode is greater than 255, a standard opcode
783       // must be used instead.
784       //
785       // To decode a special opcode, subtract the opcode_base from the
786       // opcode itself to give the adjusted opcode. The amount to
787       // increment the address register is the result of the adjusted
788       // opcode divided by the line_range multiplied by the
789       // minimum_instruction_length field from the header. That is:
790       //
791       //  address increment = (adjusted opcode / line_range) *
792       //                      minimum_instruction_length
793       //
794       // The amount to increment the line register is the line_base plus
795       // the result of the adjusted opcode modulo the line_range. That is:
796       //
797       // line increment = line_base + (adjusted opcode % line_range)
798 
799       uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase;
800       uint64_t AddrOffset =
801           (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
802       int32_t LineOffset =
803           Prologue.LineBase + (AdjustOpcode % Prologue.LineRange);
804       State.Row.Line += LineOffset;
805       State.Row.Address += AddrOffset;
806 
807       if (OS) {
808         *OS << "address += " << ((uint32_t)AdjustOpcode)
809             << ",  line += " << LineOffset << "\n";
810         OS->indent(12);
811         State.Row.dump(*OS);
812       }
813 
814       State.appendRowToMatrix(*OffsetPtr);
815       // Reset discriminator to 0.
816       State.Row.Discriminator = 0;
817     }
818     if(OS)
819       *OS << "\n";
820   }
821 
822   if (!State.Sequence.Empty) {
823     fprintf(stderr, "warning: last sequence in debug line table is not"
824                     "terminated!\n");
825   }
826 
827   // Sort all sequences so that address lookup will work faster.
828   if (!Sequences.empty()) {
829     std::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC);
830     // Note: actually, instruction address ranges of sequences should not
831     // overlap (in shared objects and executables). If they do, the address
832     // lookup would still work, though, but result would be ambiguous.
833     // We don't report warning in this case. For example,
834     // sometimes .so compiled from multiple object files contains a few
835     // rudimentary sequences for address ranges [0x0, 0xsomething).
836   }
837 
838   return EndOffset;
839 }
840 
841 uint32_t
842 DWARFDebugLine::LineTable::findRowInSeq(const DWARFDebugLine::Sequence &Seq,
843                                         uint64_t Address) const {
844   if (!Seq.containsPC(Address))
845     return UnknownRowIndex;
846   // Search for instruction address in the rows describing the sequence.
847   // Rows are stored in a vector, so we may use arithmetical operations with
848   // iterators.
849   DWARFDebugLine::Row Row;
850   Row.Address = Address;
851   RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
852   RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
853   LineTable::RowIter RowPos = std::lower_bound(
854       FirstRow, LastRow, Row, DWARFDebugLine::Row::orderByAddress);
855   if (RowPos == LastRow) {
856     return Seq.LastRowIndex - 1;
857   }
858   uint32_t Index = Seq.FirstRowIndex + (RowPos - FirstRow);
859   if (RowPos->Address > Address) {
860     if (RowPos == FirstRow)
861       return UnknownRowIndex;
862     else
863       Index--;
864   }
865   return Index;
866 }
867 
868 uint32_t DWARFDebugLine::LineTable::lookupAddress(uint64_t Address) const {
869   if (Sequences.empty())
870     return UnknownRowIndex;
871   // First, find an instruction sequence containing the given address.
872   DWARFDebugLine::Sequence Sequence;
873   Sequence.LowPC = Address;
874   SequenceIter FirstSeq = Sequences.begin();
875   SequenceIter LastSeq = Sequences.end();
876   SequenceIter SeqPos = std::lower_bound(
877       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
878   DWARFDebugLine::Sequence FoundSeq;
879   if (SeqPos == LastSeq) {
880     FoundSeq = Sequences.back();
881   } else if (SeqPos->LowPC == Address) {
882     FoundSeq = *SeqPos;
883   } else {
884     if (SeqPos == FirstSeq)
885       return UnknownRowIndex;
886     FoundSeq = *(SeqPos - 1);
887   }
888   return findRowInSeq(FoundSeq, Address);
889 }
890 
891 bool DWARFDebugLine::LineTable::lookupAddressRange(
892     uint64_t Address, uint64_t Size, std::vector<uint32_t> &Result) const {
893   if (Sequences.empty())
894     return false;
895   uint64_t EndAddr = Address + Size;
896   // First, find an instruction sequence containing the given address.
897   DWARFDebugLine::Sequence Sequence;
898   Sequence.LowPC = Address;
899   SequenceIter FirstSeq = Sequences.begin();
900   SequenceIter LastSeq = Sequences.end();
901   SequenceIter SeqPos = std::lower_bound(
902       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
903   if (SeqPos == LastSeq || SeqPos->LowPC != Address) {
904     if (SeqPos == FirstSeq)
905       return false;
906     SeqPos--;
907   }
908   if (!SeqPos->containsPC(Address))
909     return false;
910 
911   SequenceIter StartPos = SeqPos;
912 
913   // Add the rows from the first sequence to the vector, starting with the
914   // index we just calculated
915 
916   while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
917     const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
918     // For the first sequence, we need to find which row in the sequence is the
919     // first in our range.
920     uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
921     if (SeqPos == StartPos)
922       FirstRowIndex = findRowInSeq(CurSeq, Address);
923 
924     // Figure out the last row in the range.
925     uint32_t LastRowIndex = findRowInSeq(CurSeq, EndAddr - 1);
926     if (LastRowIndex == UnknownRowIndex)
927       LastRowIndex = CurSeq.LastRowIndex - 1;
928 
929     assert(FirstRowIndex != UnknownRowIndex);
930     assert(LastRowIndex != UnknownRowIndex);
931 
932     for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
933       Result.push_back(I);
934     }
935 
936     ++SeqPos;
937   }
938 
939   return true;
940 }
941 
942 bool DWARFDebugLine::LineTable::hasFileAtIndex(uint64_t FileIndex) const {
943   return FileIndex != 0 && FileIndex <= Prologue.FileNames.size();
944 }
945 
946 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
947                                                                 FileLineInfoKind Kind) const {
948   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
949     return None;
950   const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
951   if (Optional<const char *> source = Entry.Source.getAsCString())
952     return StringRef(*source);
953   return None;
954 }
955 
956 bool DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex,
957                                                    const char *CompDir,
958                                                    FileLineInfoKind Kind,
959                                                    std::string &Result) const {
960   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
961     return false;
962   const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
963   StringRef FileName = Entry.Name.getAsCString().getValue();
964   if (Kind != FileLineInfoKind::AbsoluteFilePath ||
965       sys::path::is_absolute(FileName)) {
966     Result = FileName;
967     return true;
968   }
969 
970   SmallString<16> FilePath;
971   uint64_t IncludeDirIndex = Entry.DirIdx;
972   StringRef IncludeDir;
973   // Be defensive about the contents of Entry.
974   if (IncludeDirIndex > 0 &&
975       IncludeDirIndex <= Prologue.IncludeDirectories.size())
976     IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1]
977                      .getAsCString()
978                      .getValue();
979 
980   // We may still need to append compilation directory of compile unit.
981   // We know that FileName is not absolute, the only way to have an
982   // absolute path at this point would be if IncludeDir is absolute.
983   if (CompDir && Kind == FileLineInfoKind::AbsoluteFilePath &&
984       sys::path::is_relative(IncludeDir))
985     sys::path::append(FilePath, CompDir);
986 
987   // sys::path::append skips empty strings.
988   sys::path::append(FilePath, IncludeDir, FileName);
989   Result = FilePath.str();
990   return true;
991 }
992 
993 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
994     uint64_t Address, const char *CompDir, FileLineInfoKind Kind,
995     DILineInfo &Result) const {
996   // Get the index of row we're looking for in the line table.
997   uint32_t RowIndex = lookupAddress(Address);
998   if (RowIndex == -1U)
999     return false;
1000   // Take file number and line/column from the row.
1001   const auto &Row = Rows[RowIndex];
1002   if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1003     return false;
1004   Result.Line = Row.Line;
1005   Result.Column = Row.Column;
1006   Result.Discriminator = Row.Discriminator;
1007   Result.Source = getSourceByIndex(Row.File, Kind);
1008   return true;
1009 }
1010