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/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/DWARFContext.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 DWARFDebugLine::Prologue::Prologue() { clear(); }
45 
46 void DWARFDebugLine::Prologue::clear() {
47   TotalLength = PrologueLength = 0;
48   SegSelectorSize = 0;
49   MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
50   OpcodeBase = 0;
51   FormParams = DWARFFormParams({0, 0, DWARF32});
52   StandardOpcodeLengths.clear();
53   IncludeDirectories.clear();
54   FileNames.clear();
55 }
56 
57 void DWARFDebugLine::Prologue::dump(raw_ostream &OS) const {
58   OS << "Line table prologue:\n"
59      << format("    total_length: 0x%8.8" PRIx64 "\n", TotalLength)
60      << format("         version: %u\n", getVersion());
61   if (getVersion() >= 5)
62     OS << format("    address_size: %u\n", getAddressSize())
63        << format(" seg_select_size: %u\n", SegSelectorSize);
64   OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength)
65      << format(" min_inst_length: %u\n", MinInstLength)
66      << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
67      << format(" default_is_stmt: %u\n", DefaultIsStmt)
68      << format("       line_base: %i\n", LineBase)
69      << format("      line_range: %u\n", LineRange)
70      << format("     opcode_base: %u\n", OpcodeBase);
71 
72   for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
73     OS << format("standard_opcode_lengths[%s] = %u\n",
74                  LNStandardString(I + 1).data(), StandardOpcodeLengths[I]);
75 
76   if (!IncludeDirectories.empty())
77     for (uint32_t I = 0; I != IncludeDirectories.size(); ++I)
78       OS << format("include_directories[%3u] = '", I + 1)
79          << IncludeDirectories[I] << "'\n";
80 
81   if (!FileNames.empty()) {
82     OS << "                Dir  Mod Time   File Len   File Name\n"
83        << "                ---- ---------- ---------- -----------"
84           "----------------\n";
85     for (uint32_t I = 0; I != FileNames.size(); ++I) {
86       const FileNameEntry &FileEntry = FileNames[I];
87       OS << format("file_names[%3u] %4" PRIu64 " ", I + 1, FileEntry.DirIdx)
88          << format("0x%8.8" PRIx64 " 0x%8.8" PRIx64 " ", FileEntry.ModTime,
89                    FileEntry.Length)
90          << FileEntry.Name << '\n';
91     }
92   }
93 }
94 
95 // Parse v2-v4 directory and file tables.
96 static void
97 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
98                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
99                      std::vector<StringRef> &IncludeDirectories,
100                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
101   while (*OffsetPtr < EndPrologueOffset) {
102     StringRef S = DebugLineData.getCStrRef(OffsetPtr);
103     if (S.empty())
104       break;
105     IncludeDirectories.push_back(S);
106   }
107 
108   while (*OffsetPtr < EndPrologueOffset) {
109     StringRef Name = DebugLineData.getCStrRef(OffsetPtr);
110     if (Name.empty())
111       break;
112     DWARFDebugLine::FileNameEntry FileEntry;
113     FileEntry.Name = Name;
114     FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
115     FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
116     FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
117     FileNames.push_back(FileEntry);
118   }
119 }
120 
121 // Parse v5 directory/file entry content descriptions.
122 // Returns the descriptors, or an empty vector if we did not find a path or
123 // ran off the end of the prologue.
124 static ContentDescriptors
125 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint32_t *OffsetPtr,
126                    uint64_t EndPrologueOffset) {
127   ContentDescriptors Descriptors;
128   int FormatCount = DebugLineData.getU8(OffsetPtr);
129   bool HasPath = false;
130   for (int I = 0; I != FormatCount; ++I) {
131     if (*OffsetPtr >= EndPrologueOffset)
132       return ContentDescriptors();
133     ContentDescriptor Descriptor;
134     Descriptor.Type =
135       dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr));
136     Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr));
137     if (Descriptor.Type == dwarf::DW_LNCT_path)
138       HasPath = true;
139     Descriptors.push_back(Descriptor);
140   }
141   return HasPath ? Descriptors : ContentDescriptors();
142 }
143 
144 static bool
145 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
146                      uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
147                      const DWARFFormParams &FormParams, const DWARFUnit *U,
148                      std::vector<StringRef> &IncludeDirectories,
149                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
150   // Get the directory entry description.
151   ContentDescriptors DirDescriptors =
152     parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset);
153   if (DirDescriptors.empty())
154     return false;
155 
156   // Get the directory entries, according to the format described above.
157   int DirEntryCount = DebugLineData.getU8(OffsetPtr);
158   for (int I = 0; I != DirEntryCount; ++I) {
159     if (*OffsetPtr >= EndPrologueOffset)
160       return false;
161     for (auto Descriptor : DirDescriptors) {
162       DWARFFormValue Value(Descriptor.Form);
163       switch (Descriptor.Type) {
164       case DW_LNCT_path:
165         if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, U))
166           return false;
167         IncludeDirectories.push_back(Value.getAsCString().getValue());
168         break;
169       default:
170         if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
171           return false;
172       }
173     }
174   }
175 
176   // Get the file entry description.
177   ContentDescriptors FileDescriptors =
178     parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset);
179   if (FileDescriptors.empty())
180     return false;
181 
182   // Get the file entries, according to the format described above.
183   int FileEntryCount = DebugLineData.getU8(OffsetPtr);
184   for (int I = 0; I != FileEntryCount; ++I) {
185     if (*OffsetPtr >= EndPrologueOffset)
186       return false;
187     DWARFDebugLine::FileNameEntry FileEntry;
188     for (auto Descriptor : FileDescriptors) {
189       DWARFFormValue Value(Descriptor.Form);
190       if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, U))
191         return false;
192       switch (Descriptor.Type) {
193       case DW_LNCT_path:
194         FileEntry.Name = Value.getAsCString().getValue();
195         break;
196       case DW_LNCT_directory_index:
197         FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
198         break;
199       case DW_LNCT_timestamp:
200         FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
201         break;
202       case DW_LNCT_size:
203         FileEntry.Length = Value.getAsUnsignedConstant().getValue();
204         break;
205       // FIXME: Add MD5
206       default:
207         break;
208       }
209     }
210     FileNames.push_back(FileEntry);
211   }
212   return true;
213 }
214 
215 bool DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData,
216                                      uint32_t *OffsetPtr, const DWARFUnit *U) {
217   const uint64_t PrologueOffset = *OffsetPtr;
218 
219   clear();
220   TotalLength = DebugLineData.getU32(OffsetPtr);
221   if (TotalLength == UINT32_MAX) {
222     FormParams.Format = dwarf::DWARF64;
223     TotalLength = DebugLineData.getU64(OffsetPtr);
224   } else if (TotalLength >= 0xffffff00) {
225     return false;
226   }
227   FormParams.Version = DebugLineData.getU16(OffsetPtr);
228   if (getVersion() < 2)
229     return false;
230 
231   if (getVersion() >= 5) {
232     FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
233     assert(getAddressSize() == DebugLineData.getAddressSize() &&
234            "Line table header and data extractor disagree");
235     SegSelectorSize = DebugLineData.getU8(OffsetPtr);
236   }
237 
238   PrologueLength = DebugLineData.getUnsigned(OffsetPtr, sizeofPrologueLength());
239   const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
240   MinInstLength = DebugLineData.getU8(OffsetPtr);
241   if (getVersion() >= 4)
242     MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
243   DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
244   LineBase = DebugLineData.getU8(OffsetPtr);
245   LineRange = DebugLineData.getU8(OffsetPtr);
246   OpcodeBase = DebugLineData.getU8(OffsetPtr);
247 
248   StandardOpcodeLengths.reserve(OpcodeBase - 1);
249   for (uint32_t I = 1; I < OpcodeBase; ++I) {
250     uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
251     StandardOpcodeLengths.push_back(OpLen);
252   }
253 
254   if (getVersion() >= 5) {
255     if (!parseV5DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
256                               getFormParams(), U, IncludeDirectories,
257                               FileNames)) {
258       fprintf(stderr,
259               "warning: parsing line table prologue at 0x%8.8" PRIx64
260               " found an invalid directory or file table description at"
261               " 0x%8.8" PRIx64 "\n", PrologueOffset, (uint64_t)*OffsetPtr);
262       return false;
263     }
264   } else
265     parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
266                          IncludeDirectories, FileNames);
267 
268   if (*OffsetPtr != EndPrologueOffset) {
269     fprintf(stderr,
270             "warning: parsing line table prologue at 0x%8.8" PRIx64
271             " should have ended at 0x%8.8" PRIx64
272             " but it ended at 0x%8.8" PRIx64 "\n",
273             PrologueOffset, EndPrologueOffset, (uint64_t)*OffsetPtr);
274     return false;
275   }
276   return true;
277 }
278 
279 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
280 
281 void DWARFDebugLine::Row::postAppend() {
282   BasicBlock = false;
283   PrologueEnd = false;
284   EpilogueBegin = false;
285 }
286 
287 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
288   Address = 0;
289   Line = 1;
290   Column = 0;
291   File = 1;
292   Isa = 0;
293   Discriminator = 0;
294   IsStmt = DefaultIsStmt;
295   BasicBlock = false;
296   EndSequence = false;
297   PrologueEnd = false;
298   EpilogueBegin = false;
299 }
300 
301 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) {
302   OS << "Address            Line   Column File   ISA Discriminator Flags\n"
303      << "------------------ ------ ------ ------ --- ------------- "
304         "-------------\n";
305 }
306 
307 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
308   OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column)
309      << format(" %6u %3u %13u ", File, Isa, Discriminator)
310      << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
311      << (PrologueEnd ? " prologue_end" : "")
312      << (EpilogueBegin ? " epilogue_begin" : "")
313      << (EndSequence ? " end_sequence" : "") << '\n';
314 }
315 
316 DWARFDebugLine::Sequence::Sequence() { reset(); }
317 
318 void DWARFDebugLine::Sequence::reset() {
319   LowPC = 0;
320   HighPC = 0;
321   FirstRowIndex = 0;
322   LastRowIndex = 0;
323   Empty = true;
324 }
325 
326 DWARFDebugLine::LineTable::LineTable() { clear(); }
327 
328 void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const {
329   Prologue.dump(OS);
330   OS << '\n';
331 
332   if (!Rows.empty()) {
333     Row::dumpTableHeader(OS);
334     for (const Row &R : Rows) {
335       R.dump(OS);
336     }
337   }
338 }
339 
340 void DWARFDebugLine::LineTable::clear() {
341   Prologue.clear();
342   Rows.clear();
343   Sequences.clear();
344 }
345 
346 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT)
347     : LineTable(LT) {
348   resetRowAndSequence();
349 }
350 
351 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
352   Row.reset(LineTable->Prologue.DefaultIsStmt);
353   Sequence.reset();
354 }
355 
356 void DWARFDebugLine::ParsingState::appendRowToMatrix(uint32_t Offset) {
357   if (Sequence.Empty) {
358     // Record the beginning of instruction sequence.
359     Sequence.Empty = false;
360     Sequence.LowPC = Row.Address;
361     Sequence.FirstRowIndex = RowNumber;
362   }
363   ++RowNumber;
364   LineTable->appendRow(Row);
365   if (Row.EndSequence) {
366     // Record the end of instruction sequence.
367     Sequence.HighPC = Row.Address;
368     Sequence.LastRowIndex = RowNumber;
369     if (Sequence.isValid())
370       LineTable->appendSequence(Sequence);
371     Sequence.reset();
372   }
373   Row.postAppend();
374 }
375 
376 const DWARFDebugLine::LineTable *
377 DWARFDebugLine::getLineTable(uint32_t Offset) const {
378   LineTableConstIter Pos = LineTableMap.find(Offset);
379   if (Pos != LineTableMap.end())
380     return &Pos->second;
381   return nullptr;
382 }
383 
384 const DWARFDebugLine::LineTable *
385 DWARFDebugLine::getOrParseLineTable(const DWARFDataExtractor &DebugLineData,
386                                     uint32_t Offset, const DWARFUnit *U) {
387   std::pair<LineTableIter, bool> Pos =
388       LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
389   LineTable *LT = &Pos.first->second;
390   if (Pos.second) {
391     if (!LT->parse(DebugLineData, &Offset, U))
392       return nullptr;
393   }
394   return LT;
395 }
396 
397 bool DWARFDebugLine::LineTable::parse(const DWARFDataExtractor &DebugLineData,
398                                       uint32_t *OffsetPtr, const DWARFUnit *U,
399                                       raw_ostream *OS) {
400   const uint32_t DebugLineOffset = *OffsetPtr;
401 
402   clear();
403 
404   if (!Prologue.parse(DebugLineData, OffsetPtr, U)) {
405     // Restore our offset and return false to indicate failure!
406     *OffsetPtr = DebugLineOffset;
407     return false;
408   }
409 
410   if (OS)
411     Prologue.dump(*OS);
412 
413   const uint32_t EndOffset =
414       DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength();
415 
416   ParsingState State(this);
417 
418   while (*OffsetPtr < EndOffset) {
419     if (OS)
420       *OS << format("0x%08.08" PRIx32 ": ", *OffsetPtr);
421 
422     uint8_t Opcode = DebugLineData.getU8(OffsetPtr);
423 
424     if (OS)
425       *OS << format("%02.02" PRIx8 " ", Opcode);
426 
427     if (Opcode == 0) {
428       // Extended Opcodes always start with a zero opcode followed by
429       // a uleb128 length so you can skip ones you don't know about
430       uint32_t ExtOffset = *OffsetPtr;
431       uint64_t Len = DebugLineData.getULEB128(OffsetPtr);
432       uint32_t ArgSize = Len - (*OffsetPtr - ExtOffset);
433 
434       uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr);
435       if (OS)
436         *OS << LNExtendedString(SubOpcode);
437       switch (SubOpcode) {
438       case DW_LNE_end_sequence:
439         // Set the end_sequence register of the state machine to true and
440         // append a row to the matrix using the current values of the
441         // state-machine registers. Then reset the registers to the initial
442         // values specified above. Every statement program sequence must end
443         // with a DW_LNE_end_sequence instruction which creates a row whose
444         // address is that of the byte after the last target machine instruction
445         // of the sequence.
446         State.Row.EndSequence = true;
447         State.appendRowToMatrix(*OffsetPtr);
448         if (OS) {
449           *OS << "\n";
450           OS->indent(12);
451           State.Row.dump(*OS);
452         }
453         State.resetRowAndSequence();
454         break;
455 
456       case DW_LNE_set_address:
457         // Takes a single relocatable address as an operand. The size of the
458         // operand is the size appropriate to hold an address on the target
459         // machine. Set the address register to the value given by the
460         // relocatable address. All of the other statement program opcodes
461         // that affect the address register add a delta to it. This instruction
462         // stores a relocatable value into it instead.
463         State.Row.Address = DebugLineData.getRelocatedAddress(OffsetPtr);
464         if (OS)
465           *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address);
466         break;
467 
468       case DW_LNE_define_file:
469         // Takes 4 arguments. The first is a null terminated string containing
470         // a source file name. The second is an unsigned LEB128 number
471         // representing the directory index of the directory in which the file
472         // was found. The third is an unsigned LEB128 number representing the
473         // time of last modification of the file. The fourth is an unsigned
474         // LEB128 number representing the length in bytes of the file. The time
475         // and length fields may contain LEB128(0) if the information is not
476         // available.
477         //
478         // The directory index represents an entry in the include_directories
479         // section of the statement program prologue. The index is LEB128(0)
480         // if the file was found in the current directory of the compilation,
481         // LEB128(1) if it was found in the first directory in the
482         // include_directories section, and so on. The directory index is
483         // ignored for file names that represent full path names.
484         //
485         // The files are numbered, starting at 1, in the order in which they
486         // appear; the names in the prologue come before names defined by
487         // the DW_LNE_define_file instruction. These numbers are used in the
488         // the file register of the state machine.
489         {
490           FileNameEntry FileEntry;
491           FileEntry.Name = DebugLineData.getCStr(OffsetPtr);
492           FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
493           FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
494           FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
495           Prologue.FileNames.push_back(FileEntry);
496           if (OS)
497             *OS << " (" << FileEntry.Name.str()
498                 << ", dir=" << FileEntry.DirIdx << ", mod_time="
499                 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
500                 << ", length=" << FileEntry.Length << ")";
501         }
502         break;
503 
504       case DW_LNE_set_discriminator:
505         State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr);
506         if (OS)
507           *OS << " (" << State.Row.Discriminator << ")";
508         break;
509 
510       default:
511         // Length doesn't include the zero opcode byte or the length itself, but
512         // it does include the sub_opcode, so we have to adjust for that below
513         (*OffsetPtr) += ArgSize;
514         break;
515       }
516     } else if (Opcode < Prologue.OpcodeBase) {
517       if (OS)
518         *OS << LNStandardString(Opcode);
519       switch (Opcode) {
520       // Standard Opcodes
521       case DW_LNS_copy:
522         // Takes no arguments. Append a row to the matrix using the
523         // current values of the state-machine registers. Then set
524         // the basic_block register to false.
525         State.appendRowToMatrix(*OffsetPtr);
526         if (OS) {
527           *OS << "\n";
528           OS->indent(12);
529           State.Row.dump(*OS);
530           *OS << "\n";
531         }
532         break;
533 
534       case DW_LNS_advance_pc:
535         // Takes a single unsigned LEB128 operand, multiplies it by the
536         // min_inst_length field of the prologue, and adds the
537         // result to the address register of the state machine.
538         {
539           uint64_t AddrOffset =
540               DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength;
541           State.Row.Address += AddrOffset;
542           if (OS)
543             *OS << " (" << AddrOffset << ")";
544         }
545         break;
546 
547       case DW_LNS_advance_line:
548         // Takes a single signed LEB128 operand and adds that value to
549         // the line register of the state machine.
550         State.Row.Line += DebugLineData.getSLEB128(OffsetPtr);
551         if (OS)
552           *OS << " (" << State.Row.Line << ")";
553         break;
554 
555       case DW_LNS_set_file:
556         // Takes a single unsigned LEB128 operand and stores it in the file
557         // register of the state machine.
558         State.Row.File = DebugLineData.getULEB128(OffsetPtr);
559         if (OS)
560           *OS << " (" << State.Row.File << ")";
561         break;
562 
563       case DW_LNS_set_column:
564         // Takes a single unsigned LEB128 operand and stores it in the
565         // column register of the state machine.
566         State.Row.Column = DebugLineData.getULEB128(OffsetPtr);
567         if (OS)
568           *OS << " (" << State.Row.Column << ")";
569         break;
570 
571       case DW_LNS_negate_stmt:
572         // Takes no arguments. Set the is_stmt register of the state
573         // machine to the logical negation of its current value.
574         State.Row.IsStmt = !State.Row.IsStmt;
575         break;
576 
577       case DW_LNS_set_basic_block:
578         // Takes no arguments. Set the basic_block register of the
579         // state machine to true
580         State.Row.BasicBlock = true;
581         break;
582 
583       case DW_LNS_const_add_pc:
584         // Takes no arguments. Add to the address register of the state
585         // machine the address increment value corresponding to special
586         // opcode 255. The motivation for DW_LNS_const_add_pc is this:
587         // when the statement program needs to advance the address by a
588         // small amount, it can use a single special opcode, which occupies
589         // a single byte. When it needs to advance the address by up to
590         // twice the range of the last special opcode, it can use
591         // DW_LNS_const_add_pc followed by a special opcode, for a total
592         // of two bytes. Only if it needs to advance the address by more
593         // than twice that range will it need to use both DW_LNS_advance_pc
594         // and a special opcode, requiring three or more bytes.
595         {
596           uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase;
597           uint64_t AddrOffset =
598               (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
599           State.Row.Address += AddrOffset;
600           if (OS)
601             *OS
602                 << format(" (0x%16.16" PRIx64 ")", AddrOffset);
603         }
604         break;
605 
606       case DW_LNS_fixed_advance_pc:
607         // Takes a single uhalf operand. Add to the address register of
608         // the state machine the value of the (unencoded) operand. This
609         // is the only extended opcode that takes an argument that is not
610         // a variable length number. The motivation for DW_LNS_fixed_advance_pc
611         // is this: existing assemblers cannot emit DW_LNS_advance_pc or
612         // special opcodes because they cannot encode LEB128 numbers or
613         // judge when the computation of a special opcode overflows and
614         // requires the use of DW_LNS_advance_pc. Such assemblers, however,
615         // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
616         {
617           uint16_t PCOffset = DebugLineData.getU16(OffsetPtr);
618           State.Row.Address += PCOffset;
619           if (OS)
620             *OS
621                 << format(" (0x%16.16" PRIx64 ")", PCOffset);
622         }
623         break;
624 
625       case DW_LNS_set_prologue_end:
626         // Takes no arguments. Set the prologue_end register of the
627         // state machine to true
628         State.Row.PrologueEnd = true;
629         break;
630 
631       case DW_LNS_set_epilogue_begin:
632         // Takes no arguments. Set the basic_block register of the
633         // state machine to true
634         State.Row.EpilogueBegin = true;
635         break;
636 
637       case DW_LNS_set_isa:
638         // Takes a single unsigned LEB128 operand and stores it in the
639         // column register of the state machine.
640         State.Row.Isa = DebugLineData.getULEB128(OffsetPtr);
641         if (OS)
642           *OS << " (" << State.Row.Isa << ")";
643         break;
644 
645       default:
646         // Handle any unknown standard opcodes here. We know the lengths
647         // of such opcodes because they are specified in the prologue
648         // as a multiple of LEB128 operands for each opcode.
649         {
650           assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
651           uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
652           for (uint8_t I = 0; I < OpcodeLength; ++I) {
653             uint64_t Value = DebugLineData.getULEB128(OffsetPtr);
654             if (OS)
655               *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n",
656                             Value);
657           }
658         }
659         break;
660       }
661     } else {
662       // Special Opcodes
663 
664       // A special opcode value is chosen based on the amount that needs
665       // to be added to the line and address registers. The maximum line
666       // increment for a special opcode is the value of the line_base
667       // field in the header, plus the value of the line_range field,
668       // minus 1 (line base + line range - 1). If the desired line
669       // increment is greater than the maximum line increment, a standard
670       // opcode must be used instead of a special opcode. The "address
671       // advance" is calculated by dividing the desired address increment
672       // by the minimum_instruction_length field from the header. The
673       // special opcode is then calculated using the following formula:
674       //
675       //  opcode = (desired line increment - line_base) +
676       //           (line_range * address advance) + opcode_base
677       //
678       // If the resulting opcode is greater than 255, a standard opcode
679       // must be used instead.
680       //
681       // To decode a special opcode, subtract the opcode_base from the
682       // opcode itself to give the adjusted opcode. The amount to
683       // increment the address register is the result of the adjusted
684       // opcode divided by the line_range multiplied by the
685       // minimum_instruction_length field from the header. That is:
686       //
687       //  address increment = (adjusted opcode / line_range) *
688       //                      minimum_instruction_length
689       //
690       // The amount to increment the line register is the line_base plus
691       // the result of the adjusted opcode modulo the line_range. That is:
692       //
693       // line increment = line_base + (adjusted opcode % line_range)
694 
695       uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase;
696       uint64_t AddrOffset =
697           (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
698       int32_t LineOffset =
699           Prologue.LineBase + (AdjustOpcode % Prologue.LineRange);
700       State.Row.Line += LineOffset;
701       State.Row.Address += AddrOffset;
702 
703       if (OS) {
704         *OS << "address += " << ((uint32_t)AdjustOpcode)
705             << ",  line += " << LineOffset << "\n";
706         OS->indent(12);
707         State.Row.dump(*OS);
708       }
709 
710       State.appendRowToMatrix(*OffsetPtr);
711       // Reset discriminator to 0.
712       State.Row.Discriminator = 0;
713     }
714     if(OS)
715       *OS << "\n";
716   }
717 
718   if (!State.Sequence.Empty) {
719     fprintf(stderr, "warning: last sequence in debug line table is not"
720                     "terminated!\n");
721   }
722 
723   // Sort all sequences so that address lookup will work faster.
724   if (!Sequences.empty()) {
725     std::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC);
726     // Note: actually, instruction address ranges of sequences should not
727     // overlap (in shared objects and executables). If they do, the address
728     // lookup would still work, though, but result would be ambiguous.
729     // We don't report warning in this case. For example,
730     // sometimes .so compiled from multiple object files contains a few
731     // rudimentary sequences for address ranges [0x0, 0xsomething).
732   }
733 
734   return EndOffset;
735 }
736 
737 uint32_t
738 DWARFDebugLine::LineTable::findRowInSeq(const DWARFDebugLine::Sequence &Seq,
739                                         uint64_t Address) const {
740   if (!Seq.containsPC(Address))
741     return UnknownRowIndex;
742   // Search for instruction address in the rows describing the sequence.
743   // Rows are stored in a vector, so we may use arithmetical operations with
744   // iterators.
745   DWARFDebugLine::Row Row;
746   Row.Address = Address;
747   RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
748   RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
749   LineTable::RowIter RowPos = std::lower_bound(
750       FirstRow, LastRow, Row, DWARFDebugLine::Row::orderByAddress);
751   if (RowPos == LastRow) {
752     return Seq.LastRowIndex - 1;
753   }
754   uint32_t Index = Seq.FirstRowIndex + (RowPos - FirstRow);
755   if (RowPos->Address > Address) {
756     if (RowPos == FirstRow)
757       return UnknownRowIndex;
758     else
759       Index--;
760   }
761   return Index;
762 }
763 
764 uint32_t DWARFDebugLine::LineTable::lookupAddress(uint64_t Address) const {
765   if (Sequences.empty())
766     return UnknownRowIndex;
767   // First, find an instruction sequence containing the given address.
768   DWARFDebugLine::Sequence Sequence;
769   Sequence.LowPC = Address;
770   SequenceIter FirstSeq = Sequences.begin();
771   SequenceIter LastSeq = Sequences.end();
772   SequenceIter SeqPos = std::lower_bound(
773       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
774   DWARFDebugLine::Sequence FoundSeq;
775   if (SeqPos == LastSeq) {
776     FoundSeq = Sequences.back();
777   } else if (SeqPos->LowPC == Address) {
778     FoundSeq = *SeqPos;
779   } else {
780     if (SeqPos == FirstSeq)
781       return UnknownRowIndex;
782     FoundSeq = *(SeqPos - 1);
783   }
784   return findRowInSeq(FoundSeq, Address);
785 }
786 
787 bool DWARFDebugLine::LineTable::lookupAddressRange(
788     uint64_t Address, uint64_t Size, std::vector<uint32_t> &Result) const {
789   if (Sequences.empty())
790     return false;
791   uint64_t EndAddr = Address + Size;
792   // First, find an instruction sequence containing the given address.
793   DWARFDebugLine::Sequence Sequence;
794   Sequence.LowPC = Address;
795   SequenceIter FirstSeq = Sequences.begin();
796   SequenceIter LastSeq = Sequences.end();
797   SequenceIter SeqPos = std::lower_bound(
798       FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC);
799   if (SeqPos == LastSeq || SeqPos->LowPC != Address) {
800     if (SeqPos == FirstSeq)
801       return false;
802     SeqPos--;
803   }
804   if (!SeqPos->containsPC(Address))
805     return false;
806 
807   SequenceIter StartPos = SeqPos;
808 
809   // Add the rows from the first sequence to the vector, starting with the
810   // index we just calculated
811 
812   while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
813     const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
814     // For the first sequence, we need to find which row in the sequence is the
815     // first in our range.
816     uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
817     if (SeqPos == StartPos)
818       FirstRowIndex = findRowInSeq(CurSeq, Address);
819 
820     // Figure out the last row in the range.
821     uint32_t LastRowIndex = findRowInSeq(CurSeq, EndAddr - 1);
822     if (LastRowIndex == UnknownRowIndex)
823       LastRowIndex = CurSeq.LastRowIndex - 1;
824 
825     assert(FirstRowIndex != UnknownRowIndex);
826     assert(LastRowIndex != UnknownRowIndex);
827 
828     for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
829       Result.push_back(I);
830     }
831 
832     ++SeqPos;
833   }
834 
835   return true;
836 }
837 
838 bool DWARFDebugLine::LineTable::hasFileAtIndex(uint64_t FileIndex) const {
839   return FileIndex != 0 && FileIndex <= Prologue.FileNames.size();
840 }
841 
842 bool DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex,
843                                                    const char *CompDir,
844                                                    FileLineInfoKind Kind,
845                                                    std::string &Result) const {
846   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
847     return false;
848   const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
849   StringRef FileName = Entry.Name;
850   if (Kind != FileLineInfoKind::AbsoluteFilePath ||
851       sys::path::is_absolute(FileName)) {
852     Result = FileName;
853     return true;
854   }
855 
856   SmallString<16> FilePath;
857   uint64_t IncludeDirIndex = Entry.DirIdx;
858   StringRef IncludeDir;
859   // Be defensive about the contents of Entry.
860   if (IncludeDirIndex > 0 &&
861       IncludeDirIndex <= Prologue.IncludeDirectories.size())
862     IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1];
863 
864   // We may still need to append compilation directory of compile unit.
865   // We know that FileName is not absolute, the only way to have an
866   // absolute path at this point would be if IncludeDir is absolute.
867   if (CompDir && Kind == FileLineInfoKind::AbsoluteFilePath &&
868       sys::path::is_relative(IncludeDir))
869     sys::path::append(FilePath, CompDir);
870 
871   // sys::path::append skips empty strings.
872   sys::path::append(FilePath, IncludeDir, FileName);
873   Result = FilePath.str();
874   return true;
875 }
876 
877 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
878     uint64_t Address, const char *CompDir, FileLineInfoKind Kind,
879     DILineInfo &Result) const {
880   // Get the index of row we're looking for in the line table.
881   uint32_t RowIndex = lookupAddress(Address);
882   if (RowIndex == -1U)
883     return false;
884   // Take file number and line/column from the row.
885   const auto &Row = Rows[RowIndex];
886   if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
887     return false;
888   Result.Line = Row.Line;
889   Result.Column = Row.Column;
890   Result.Discriminator = Row.Discriminator;
891   return true;
892 }
893