1 //===- DWARFDebugLine.cpp -------------------------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 
9 #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h"
10 #include "llvm/ADT/SmallString.h"
11 #include "llvm/ADT/SmallVector.h"
12 #include "llvm/ADT/StringRef.h"
13 #include "llvm/BinaryFormat/Dwarf.h"
14 #include "llvm/DebugInfo/DWARF/DWARFDataExtractor.h"
15 #include "llvm/DebugInfo/DWARF/DWARFDie.h"
16 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h"
17 #include "llvm/Support/Errc.h"
18 #include "llvm/Support/Format.h"
19 #include "llvm/Support/FormatVariadic.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 anonymous namespace
43 
versionIsSupported(uint16_t Version)44 static bool versionIsSupported(uint16_t Version) {
45   return Version >= 2 && Version <= 5;
46 }
47 
trackContentType(dwarf::LineNumberEntryFormat ContentType)48 void DWARFDebugLine::ContentTypeTracker::trackContentType(
49     dwarf::LineNumberEntryFormat ContentType) {
50   switch (ContentType) {
51   case dwarf::DW_LNCT_timestamp:
52     HasModTime = true;
53     break;
54   case dwarf::DW_LNCT_size:
55     HasLength = true;
56     break;
57   case dwarf::DW_LNCT_MD5:
58     HasMD5 = true;
59     break;
60   case dwarf::DW_LNCT_LLVM_source:
61     HasSource = true;
62     break;
63   default:
64     // We only care about values we consider optional, and new values may be
65     // added in the vendor extension range, so we do not match exhaustively.
66     break;
67   }
68 }
69 
Prologue()70 DWARFDebugLine::Prologue::Prologue() { clear(); }
71 
hasFileAtIndex(uint64_t FileIndex) const72 bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const {
73   uint16_t DwarfVersion = getVersion();
74   assert(DwarfVersion != 0 &&
75          "line table prologue has no dwarf version information");
76   if (DwarfVersion >= 5)
77     return FileIndex < FileNames.size();
78   return FileIndex != 0 && FileIndex <= FileNames.size();
79 }
80 
81 std::optional<uint64_t>
getLastValidFileIndex() const82 DWARFDebugLine::Prologue::getLastValidFileIndex() const {
83   if (FileNames.empty())
84     return std::nullopt;
85   uint16_t DwarfVersion = getVersion();
86   assert(DwarfVersion != 0 &&
87          "line table prologue has no dwarf version information");
88   // In DWARF v5 the file names are 0-indexed.
89   if (DwarfVersion >= 5)
90     return FileNames.size() - 1;
91   return FileNames.size();
92 }
93 
94 const llvm::DWARFDebugLine::FileNameEntry &
getFileNameEntry(uint64_t Index) const95 DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const {
96   uint16_t DwarfVersion = getVersion();
97   assert(DwarfVersion != 0 &&
98          "line table prologue has no dwarf version information");
99   // In DWARF v5 the file names are 0-indexed.
100   if (DwarfVersion >= 5)
101     return FileNames[Index];
102   return FileNames[Index - 1];
103 }
104 
clear()105 void DWARFDebugLine::Prologue::clear() {
106   TotalLength = PrologueLength = 0;
107   SegSelectorSize = 0;
108   MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
109   OpcodeBase = 0;
110   FormParams = dwarf::FormParams({0, 0, DWARF32});
111   ContentTypes = ContentTypeTracker();
112   StandardOpcodeLengths.clear();
113   IncludeDirectories.clear();
114   FileNames.clear();
115 }
116 
dump(raw_ostream & OS,DIDumpOptions DumpOptions) const117 void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
118                                     DIDumpOptions DumpOptions) const {
119   if (!totalLengthIsValid())
120     return;
121   int OffsetDumpWidth = 2 * dwarf::getDwarfOffsetByteSize(FormParams.Format);
122   OS << "Line table prologue:\n"
123      << format("    total_length: 0x%0*" PRIx64 "\n", OffsetDumpWidth,
124                TotalLength)
125      << "          format: " << dwarf::FormatString(FormParams.Format) << "\n"
126      << format("         version: %u\n", getVersion());
127   if (!versionIsSupported(getVersion()))
128     return;
129   if (getVersion() >= 5)
130     OS << format("    address_size: %u\n", getAddressSize())
131        << format(" seg_select_size: %u\n", SegSelectorSize);
132   OS << format(" prologue_length: 0x%0*" PRIx64 "\n", OffsetDumpWidth,
133                PrologueLength)
134      << format(" min_inst_length: %u\n", MinInstLength)
135      << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
136      << format(" default_is_stmt: %u\n", DefaultIsStmt)
137      << format("       line_base: %i\n", LineBase)
138      << format("      line_range: %u\n", LineRange)
139      << format("     opcode_base: %u\n", OpcodeBase);
140 
141   for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
142     OS << formatv("standard_opcode_lengths[{0}] = {1}\n",
143                   static_cast<dwarf::LineNumberOps>(I + 1),
144                   StandardOpcodeLengths[I]);
145 
146   if (!IncludeDirectories.empty()) {
147     // DWARF v5 starts directory indexes at 0.
148     uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
149     for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
150       OS << format("include_directories[%3u] = ", I + DirBase);
151       IncludeDirectories[I].dump(OS, DumpOptions);
152       OS << '\n';
153     }
154   }
155 
156   if (!FileNames.empty()) {
157     // DWARF v5 starts file indexes at 0.
158     uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
159     for (uint32_t I = 0; I != FileNames.size(); ++I) {
160       const FileNameEntry &FileEntry = FileNames[I];
161       OS <<   format("file_names[%3u]:\n", I + FileBase);
162       OS <<          "           name: ";
163       FileEntry.Name.dump(OS, DumpOptions);
164       OS << '\n'
165          <<   format("      dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
166       if (ContentTypes.HasMD5)
167         OS <<        "   md5_checksum: " << FileEntry.Checksum.digest() << '\n';
168       if (ContentTypes.HasModTime)
169         OS << format("       mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
170       if (ContentTypes.HasLength)
171         OS << format("         length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
172       if (ContentTypes.HasSource) {
173         OS <<        "         source: ";
174         FileEntry.Source.dump(OS, DumpOptions);
175         OS << '\n';
176       }
177     }
178   }
179 }
180 
181 // Parse v2-v4 directory and file tables.
182 static Error
parseV2DirFileTables(const DWARFDataExtractor & DebugLineData,uint64_t * OffsetPtr,DWARFDebugLine::ContentTypeTracker & ContentTypes,std::vector<DWARFFormValue> & IncludeDirectories,std::vector<DWARFDebugLine::FileNameEntry> & FileNames)183 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
184                      uint64_t *OffsetPtr,
185                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
186                      std::vector<DWARFFormValue> &IncludeDirectories,
187                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
188   while (true) {
189     Error Err = Error::success();
190     StringRef S = DebugLineData.getCStrRef(OffsetPtr, &Err);
191     if (Err) {
192       consumeError(std::move(Err));
193       return createStringError(errc::invalid_argument,
194                                "include directories table was not null "
195                                "terminated before the end of the prologue");
196     }
197     if (S.empty())
198       break;
199     DWARFFormValue Dir =
200         DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, S.data());
201     IncludeDirectories.push_back(Dir);
202   }
203 
204   ContentTypes.HasModTime = true;
205   ContentTypes.HasLength = true;
206 
207   while (true) {
208     Error Err = Error::success();
209     StringRef Name = DebugLineData.getCStrRef(OffsetPtr, &Err);
210     if (!Err && Name.empty())
211       break;
212 
213     DWARFDebugLine::FileNameEntry FileEntry;
214     FileEntry.Name =
215         DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name.data());
216     FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr, &Err);
217     FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr, &Err);
218     FileEntry.Length = DebugLineData.getULEB128(OffsetPtr, &Err);
219 
220     if (Err) {
221       consumeError(std::move(Err));
222       return createStringError(
223           errc::invalid_argument,
224           "file names table was not null terminated before "
225           "the end of the prologue");
226     }
227     FileNames.push_back(FileEntry);
228   }
229 
230   return Error::success();
231 }
232 
233 // Parse v5 directory/file entry content descriptions.
234 // Returns the descriptors, or an error if we did not find a path or ran off
235 // the end of the prologue.
236 static llvm::Expected<ContentDescriptors>
parseV5EntryFormat(const DWARFDataExtractor & DebugLineData,uint64_t * OffsetPtr,DWARFDebugLine::ContentTypeTracker * ContentTypes)237 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
238                    DWARFDebugLine::ContentTypeTracker *ContentTypes) {
239   Error Err = Error::success();
240   ContentDescriptors Descriptors;
241   int FormatCount = DebugLineData.getU8(OffsetPtr, &Err);
242   bool HasPath = false;
243   for (int I = 0; I != FormatCount && !Err; ++I) {
244     ContentDescriptor Descriptor;
245     Descriptor.Type =
246         dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr, &Err));
247     Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr, &Err));
248     if (Descriptor.Type == dwarf::DW_LNCT_path)
249       HasPath = true;
250     if (ContentTypes)
251       ContentTypes->trackContentType(Descriptor.Type);
252     Descriptors.push_back(Descriptor);
253   }
254 
255   if (Err)
256     return createStringError(errc::invalid_argument,
257                              "failed to parse entry content descriptors: %s",
258                              toString(std::move(Err)).c_str());
259 
260   if (!HasPath)
261     return createStringError(errc::invalid_argument,
262                              "failed to parse entry content descriptions"
263                              " because no path was found");
264   return Descriptors;
265 }
266 
267 static Error
parseV5DirFileTables(const DWARFDataExtractor & DebugLineData,uint64_t * OffsetPtr,const dwarf::FormParams & FormParams,const DWARFContext & Ctx,const DWARFUnit * U,DWARFDebugLine::ContentTypeTracker & ContentTypes,std::vector<DWARFFormValue> & IncludeDirectories,std::vector<DWARFDebugLine::FileNameEntry> & FileNames)268 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
269                      uint64_t *OffsetPtr, const dwarf::FormParams &FormParams,
270                      const DWARFContext &Ctx, const DWARFUnit *U,
271                      DWARFDebugLine::ContentTypeTracker &ContentTypes,
272                      std::vector<DWARFFormValue> &IncludeDirectories,
273                      std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
274   // Get the directory entry description.
275   llvm::Expected<ContentDescriptors> DirDescriptors =
276       parseV5EntryFormat(DebugLineData, OffsetPtr, nullptr);
277   if (!DirDescriptors)
278     return DirDescriptors.takeError();
279 
280   // Get the directory entries, according to the format described above.
281   uint64_t DirEntryCount = DebugLineData.getULEB128(OffsetPtr);
282   for (uint64_t I = 0; I != DirEntryCount; ++I) {
283     for (auto Descriptor : *DirDescriptors) {
284       DWARFFormValue Value(Descriptor.Form);
285       switch (Descriptor.Type) {
286       case DW_LNCT_path:
287         if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
288           return createStringError(errc::invalid_argument,
289                                    "failed to parse directory entry because "
290                                    "extracting the form value failed");
291         IncludeDirectories.push_back(Value);
292         break;
293       default:
294         if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
295           return createStringError(errc::invalid_argument,
296                                    "failed to parse directory entry because "
297                                    "skipping the form value failed");
298       }
299     }
300   }
301 
302   // Get the file entry description.
303   llvm::Expected<ContentDescriptors> FileDescriptors =
304       parseV5EntryFormat(DebugLineData, OffsetPtr, &ContentTypes);
305   if (!FileDescriptors)
306     return FileDescriptors.takeError();
307 
308   // Get the file entries, according to the format described above.
309   uint64_t FileEntryCount = DebugLineData.getULEB128(OffsetPtr);
310   for (uint64_t I = 0; I != FileEntryCount; ++I) {
311     DWARFDebugLine::FileNameEntry FileEntry;
312     for (auto Descriptor : *FileDescriptors) {
313       DWARFFormValue Value(Descriptor.Form);
314       if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
315         return createStringError(errc::invalid_argument,
316                                  "failed to parse file entry because "
317                                  "extracting the form value failed");
318       switch (Descriptor.Type) {
319       case DW_LNCT_path:
320         FileEntry.Name = Value;
321         break;
322       case DW_LNCT_LLVM_source:
323         FileEntry.Source = Value;
324         break;
325       case DW_LNCT_directory_index:
326         FileEntry.DirIdx = *Value.getAsUnsignedConstant();
327         break;
328       case DW_LNCT_timestamp:
329         FileEntry.ModTime = *Value.getAsUnsignedConstant();
330         break;
331       case DW_LNCT_size:
332         FileEntry.Length = *Value.getAsUnsignedConstant();
333         break;
334       case DW_LNCT_MD5:
335         if (!Value.getAsBlock() || Value.getAsBlock()->size() != 16)
336           return createStringError(
337               errc::invalid_argument,
338               "failed to parse file entry because the MD5 hash is invalid");
339         std::uninitialized_copy_n(Value.getAsBlock()->begin(), 16,
340                                   FileEntry.Checksum.begin());
341         break;
342       default:
343         break;
344       }
345     }
346     FileNames.push_back(FileEntry);
347   }
348   return Error::success();
349 }
350 
getLength() const351 uint64_t DWARFDebugLine::Prologue::getLength() const {
352   uint64_t Length = PrologueLength + sizeofTotalLength() +
353                     sizeof(getVersion()) + sizeofPrologueLength();
354   if (getVersion() >= 5)
355     Length += 2; // Address + Segment selector sizes.
356   return Length;
357 }
358 
parse(DWARFDataExtractor DebugLineData,uint64_t * OffsetPtr,function_ref<void (Error)> RecoverableErrorHandler,const DWARFContext & Ctx,const DWARFUnit * U)359 Error DWARFDebugLine::Prologue::parse(
360     DWARFDataExtractor DebugLineData, uint64_t *OffsetPtr,
361     function_ref<void(Error)> RecoverableErrorHandler, const DWARFContext &Ctx,
362     const DWARFUnit *U) {
363   const uint64_t PrologueOffset = *OffsetPtr;
364 
365   clear();
366   DataExtractor::Cursor Cursor(*OffsetPtr);
367   std::tie(TotalLength, FormParams.Format) =
368       DebugLineData.getInitialLength(Cursor);
369 
370   DebugLineData =
371       DWARFDataExtractor(DebugLineData, Cursor.tell() + TotalLength);
372   FormParams.Version = DebugLineData.getU16(Cursor);
373   if (Cursor && !versionIsSupported(getVersion())) {
374     // Treat this error as unrecoverable - we cannot be sure what any of
375     // the data represents including the length field, so cannot skip it or make
376     // any reasonable assumptions.
377     *OffsetPtr = Cursor.tell();
378     return createStringError(
379         errc::not_supported,
380         "parsing line table prologue at offset 0x%8.8" PRIx64
381         ": unsupported version %" PRIu16,
382         PrologueOffset, getVersion());
383   }
384 
385   if (getVersion() >= 5) {
386     FormParams.AddrSize = DebugLineData.getU8(Cursor);
387     assert((!Cursor || DebugLineData.getAddressSize() == 0 ||
388             DebugLineData.getAddressSize() == getAddressSize()) &&
389            "Line table header and data extractor disagree");
390     SegSelectorSize = DebugLineData.getU8(Cursor);
391   }
392 
393   PrologueLength =
394       DebugLineData.getRelocatedValue(Cursor, sizeofPrologueLength());
395   const uint64_t EndPrologueOffset = PrologueLength + Cursor.tell();
396   DebugLineData = DWARFDataExtractor(DebugLineData, EndPrologueOffset);
397   MinInstLength = DebugLineData.getU8(Cursor);
398   if (getVersion() >= 4)
399     MaxOpsPerInst = DebugLineData.getU8(Cursor);
400   DefaultIsStmt = DebugLineData.getU8(Cursor);
401   LineBase = DebugLineData.getU8(Cursor);
402   LineRange = DebugLineData.getU8(Cursor);
403   OpcodeBase = DebugLineData.getU8(Cursor);
404 
405   if (Cursor && OpcodeBase == 0) {
406     // If the opcode base is 0, we cannot read the standard opcode lengths (of
407     // which there are supposed to be one fewer than the opcode base). Assume
408     // there are no standard opcodes and continue parsing.
409     RecoverableErrorHandler(createStringError(
410         errc::invalid_argument,
411         "parsing line table prologue at offset 0x%8.8" PRIx64
412         " found opcode base of 0. Assuming no standard opcodes",
413         PrologueOffset));
414   } else if (Cursor) {
415     StandardOpcodeLengths.reserve(OpcodeBase - 1);
416     for (uint32_t I = 1; I < OpcodeBase; ++I) {
417       uint8_t OpLen = DebugLineData.getU8(Cursor);
418       StandardOpcodeLengths.push_back(OpLen);
419     }
420   }
421 
422   *OffsetPtr = Cursor.tell();
423   // A corrupt file name or directory table does not prevent interpretation of
424   // the main line program, so check the cursor state now so that its errors can
425   // be handled separately.
426   if (!Cursor)
427     return createStringError(
428         errc::invalid_argument,
429         "parsing line table prologue at offset 0x%8.8" PRIx64 ": %s",
430         PrologueOffset, toString(Cursor.takeError()).c_str());
431 
432   Error E =
433       getVersion() >= 5
434           ? parseV5DirFileTables(DebugLineData, OffsetPtr, FormParams, Ctx, U,
435                                  ContentTypes, IncludeDirectories, FileNames)
436           : parseV2DirFileTables(DebugLineData, OffsetPtr, ContentTypes,
437                                  IncludeDirectories, FileNames);
438   if (E) {
439     RecoverableErrorHandler(joinErrors(
440         createStringError(
441             errc::invalid_argument,
442             "parsing line table prologue at 0x%8.8" PRIx64
443             " found an invalid directory or file table description at"
444             " 0x%8.8" PRIx64,
445             PrologueOffset, *OffsetPtr),
446         std::move(E)));
447     return Error::success();
448   }
449 
450   assert(*OffsetPtr <= EndPrologueOffset);
451   if (*OffsetPtr != EndPrologueOffset) {
452     RecoverableErrorHandler(createStringError(
453         errc::invalid_argument,
454         "unknown data in line table prologue at offset 0x%8.8" PRIx64
455         ": parsing ended (at offset 0x%8.8" PRIx64
456         ") before reaching the prologue end at offset 0x%8.8" PRIx64,
457         PrologueOffset, *OffsetPtr, EndPrologueOffset));
458   }
459   return Error::success();
460 }
461 
Row(bool DefaultIsStmt)462 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
463 
postAppend()464 void DWARFDebugLine::Row::postAppend() {
465   Discriminator = 0;
466   BasicBlock = false;
467   PrologueEnd = false;
468   EpilogueBegin = false;
469 }
470 
reset(bool DefaultIsStmt)471 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
472   Address.Address = 0;
473   Address.SectionIndex = object::SectionedAddress::UndefSection;
474   Line = 1;
475   Column = 0;
476   File = 1;
477   Isa = 0;
478   Discriminator = 0;
479   IsStmt = DefaultIsStmt;
480   BasicBlock = false;
481   EndSequence = false;
482   PrologueEnd = false;
483   EpilogueBegin = false;
484 }
485 
dumpTableHeader(raw_ostream & OS,unsigned Indent)486 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS, unsigned Indent) {
487   OS.indent(Indent)
488       << "Address            Line   Column File   ISA Discriminator Flags\n";
489   OS.indent(Indent)
490       << "------------------ ------ ------ ------ --- ------------- "
491          "-------------\n";
492 }
493 
dump(raw_ostream & OS) const494 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
495   OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column)
496      << format(" %6u %3u %13u ", File, Isa, Discriminator)
497      << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
498      << (PrologueEnd ? " prologue_end" : "")
499      << (EpilogueBegin ? " epilogue_begin" : "")
500      << (EndSequence ? " end_sequence" : "") << '\n';
501 }
502 
Sequence()503 DWARFDebugLine::Sequence::Sequence() { reset(); }
504 
reset()505 void DWARFDebugLine::Sequence::reset() {
506   LowPC = 0;
507   HighPC = 0;
508   SectionIndex = object::SectionedAddress::UndefSection;
509   FirstRowIndex = 0;
510   LastRowIndex = 0;
511   Empty = true;
512 }
513 
LineTable()514 DWARFDebugLine::LineTable::LineTable() { clear(); }
515 
dump(raw_ostream & OS,DIDumpOptions DumpOptions) const516 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
517                                      DIDumpOptions DumpOptions) const {
518   Prologue.dump(OS, DumpOptions);
519 
520   if (!Rows.empty()) {
521     OS << '\n';
522     Row::dumpTableHeader(OS, 0);
523     for (const Row &R : Rows) {
524       R.dump(OS);
525     }
526   }
527 
528   // Terminate the table with a final blank line to clearly delineate it from
529   // later dumps.
530   OS << '\n';
531 }
532 
clear()533 void DWARFDebugLine::LineTable::clear() {
534   Prologue.clear();
535   Rows.clear();
536   Sequences.clear();
537 }
538 
ParsingState(struct LineTable * LT,uint64_t TableOffset,function_ref<void (Error)> ErrorHandler)539 DWARFDebugLine::ParsingState::ParsingState(
540     struct LineTable *LT, uint64_t TableOffset,
541     function_ref<void(Error)> ErrorHandler)
542     : LineTable(LT), LineTableOffset(TableOffset), ErrorHandler(ErrorHandler) {
543   resetRowAndSequence();
544 }
545 
resetRowAndSequence()546 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
547   Row.reset(LineTable->Prologue.DefaultIsStmt);
548   Sequence.reset();
549 }
550 
appendRowToMatrix()551 void DWARFDebugLine::ParsingState::appendRowToMatrix() {
552   unsigned RowNumber = LineTable->Rows.size();
553   if (Sequence.Empty) {
554     // Record the beginning of instruction sequence.
555     Sequence.Empty = false;
556     Sequence.LowPC = Row.Address.Address;
557     Sequence.FirstRowIndex = RowNumber;
558   }
559   LineTable->appendRow(Row);
560   if (Row.EndSequence) {
561     // Record the end of instruction sequence.
562     Sequence.HighPC = Row.Address.Address;
563     Sequence.LastRowIndex = RowNumber + 1;
564     Sequence.SectionIndex = Row.Address.SectionIndex;
565     if (Sequence.isValid())
566       LineTable->appendSequence(Sequence);
567     Sequence.reset();
568   }
569   Row.postAppend();
570 }
571 
572 const DWARFDebugLine::LineTable *
getLineTable(uint64_t Offset) const573 DWARFDebugLine::getLineTable(uint64_t Offset) const {
574   LineTableConstIter Pos = LineTableMap.find(Offset);
575   if (Pos != LineTableMap.end())
576     return &Pos->second;
577   return nullptr;
578 }
579 
getOrParseLineTable(DWARFDataExtractor & DebugLineData,uint64_t Offset,const DWARFContext & Ctx,const DWARFUnit * U,function_ref<void (Error)> RecoverableErrorHandler)580 Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable(
581     DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx,
582     const DWARFUnit *U, function_ref<void(Error)> RecoverableErrorHandler) {
583   if (!DebugLineData.isValidOffset(Offset))
584     return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx64
585                        " is not a valid debug line section offset",
586                        Offset);
587 
588   std::pair<LineTableIter, bool> Pos =
589       LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
590   LineTable *LT = &Pos.first->second;
591   if (Pos.second) {
592     if (Error Err =
593             LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorHandler))
594       return std::move(Err);
595     return LT;
596   }
597   return LT;
598 }
599 
clearLineTable(uint64_t Offset)600 void DWARFDebugLine::clearLineTable(uint64_t Offset) {
601   LineTableMap.erase(Offset);
602 }
603 
getOpcodeName(uint8_t Opcode,uint8_t OpcodeBase)604 static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) {
605   assert(Opcode != 0);
606   if (Opcode < OpcodeBase)
607     return LNStandardString(Opcode);
608   return "special";
609 }
610 
advanceAddr(uint64_t OperationAdvance,uint8_t Opcode,uint64_t OpcodeOffset)611 uint64_t DWARFDebugLine::ParsingState::advanceAddr(uint64_t OperationAdvance,
612                                                    uint8_t Opcode,
613                                                    uint64_t OpcodeOffset) {
614   StringRef OpcodeName = getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
615   // For versions less than 4, the MaxOpsPerInst member is set to 0, as the
616   // maximum_operations_per_instruction field wasn't introduced until DWARFv4.
617   // Don't warn about bad values in this situation.
618   if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 &&
619       LineTable->Prologue.MaxOpsPerInst != 1)
620     ErrorHandler(createStringError(
621         errc::not_supported,
622         "line table program at offset 0x%8.8" PRIx64
623         " contains a %s opcode at offset 0x%8.8" PRIx64
624         ", but the prologue maximum_operations_per_instruction value is %" PRId8
625         ", which is unsupported. Assuming a value of 1 instead",
626         LineTableOffset, OpcodeName.data(), OpcodeOffset,
627         LineTable->Prologue.MaxOpsPerInst));
628   if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0)
629     ErrorHandler(
630         createStringError(errc::invalid_argument,
631                           "line table program at offset 0x%8.8" PRIx64
632                           " contains a %s opcode at offset 0x%8.8" PRIx64
633                           ", but the prologue minimum_instruction_length value "
634                           "is 0, which prevents any address advancing",
635                           LineTableOffset, OpcodeName.data(), OpcodeOffset));
636   ReportAdvanceAddrProblem = false;
637   uint64_t AddrOffset = OperationAdvance * LineTable->Prologue.MinInstLength;
638   Row.Address.Address += AddrOffset;
639   return AddrOffset;
640 }
641 
642 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode
advanceAddrForOpcode(uint8_t Opcode,uint64_t OpcodeOffset)643 DWARFDebugLine::ParsingState::advanceAddrForOpcode(uint8_t Opcode,
644                                                    uint64_t OpcodeOffset) {
645   assert(Opcode == DW_LNS_const_add_pc ||
646          Opcode >= LineTable->Prologue.OpcodeBase);
647   if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) {
648     StringRef OpcodeName =
649         getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase);
650     ErrorHandler(
651         createStringError(errc::not_supported,
652                           "line table program at offset 0x%8.8" PRIx64
653                           " contains a %s opcode at offset 0x%8.8" PRIx64
654                           ", but the prologue line_range value is 0. The "
655                           "address and line will not be adjusted",
656                           LineTableOffset, OpcodeName.data(), OpcodeOffset));
657     ReportBadLineRange = false;
658   }
659 
660   uint8_t OpcodeValue = Opcode;
661   if (Opcode == DW_LNS_const_add_pc)
662     OpcodeValue = 255;
663   uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase;
664   uint64_t OperationAdvance =
665       LineTable->Prologue.LineRange != 0
666           ? AdjustedOpcode / LineTable->Prologue.LineRange
667           : 0;
668   uint64_t AddrOffset = advanceAddr(OperationAdvance, Opcode, OpcodeOffset);
669   return {AddrOffset, AdjustedOpcode};
670 }
671 
672 DWARFDebugLine::ParsingState::AddrAndLineDelta
handleSpecialOpcode(uint8_t Opcode,uint64_t OpcodeOffset)673 DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode,
674                                                   uint64_t OpcodeOffset) {
675   // A special opcode value is chosen based on the amount that needs
676   // to be added to the line and address registers. The maximum line
677   // increment for a special opcode is the value of the line_base
678   // field in the header, plus the value of the line_range field,
679   // minus 1 (line base + line range - 1). If the desired line
680   // increment is greater than the maximum line increment, a standard
681   // opcode must be used instead of a special opcode. The "address
682   // advance" is calculated by dividing the desired address increment
683   // by the minimum_instruction_length field from the header. The
684   // special opcode is then calculated using the following formula:
685   //
686   //  opcode = (desired line increment - line_base) +
687   //           (line_range * address advance) + opcode_base
688   //
689   // If the resulting opcode is greater than 255, a standard opcode
690   // must be used instead.
691   //
692   // To decode a special opcode, subtract the opcode_base from the
693   // opcode itself to give the adjusted opcode. The amount to
694   // increment the address register is the result of the adjusted
695   // opcode divided by the line_range multiplied by the
696   // minimum_instruction_length field from the header. That is:
697   //
698   //  address increment = (adjusted opcode / line_range) *
699   //                      minimum_instruction_length
700   //
701   // The amount to increment the line register is the line_base plus
702   // the result of the adjusted opcode modulo the line_range. That is:
703   //
704   // line increment = line_base + (adjusted opcode % line_range)
705 
706   DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode AddrAdvanceResult =
707       advanceAddrForOpcode(Opcode, OpcodeOffset);
708   int32_t LineOffset = 0;
709   if (LineTable->Prologue.LineRange != 0)
710     LineOffset =
711         LineTable->Prologue.LineBase +
712         (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange);
713   Row.Line += LineOffset;
714   return {AddrAdvanceResult.AddrDelta, LineOffset};
715 }
716 
717 /// Parse a ULEB128 using the specified \p Cursor. \returns the parsed value on
718 /// success, or std::nullopt if \p Cursor is in a failing state.
719 template <typename T>
parseULEB128(DWARFDataExtractor & Data,DataExtractor::Cursor & Cursor)720 static std::optional<T> parseULEB128(DWARFDataExtractor &Data,
721                                      DataExtractor::Cursor &Cursor) {
722   T Value = Data.getULEB128(Cursor);
723   if (Cursor)
724     return Value;
725   return std::nullopt;
726 }
727 
parse(DWARFDataExtractor & DebugLineData,uint64_t * OffsetPtr,const DWARFContext & Ctx,const DWARFUnit * U,function_ref<void (Error)> RecoverableErrorHandler,raw_ostream * OS,bool Verbose)728 Error DWARFDebugLine::LineTable::parse(
729     DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
730     const DWARFContext &Ctx, const DWARFUnit *U,
731     function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS,
732     bool Verbose) {
733   assert((OS || !Verbose) && "cannot have verbose output without stream");
734   const uint64_t DebugLineOffset = *OffsetPtr;
735 
736   clear();
737 
738   Error PrologueErr =
739       Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U);
740 
741   if (OS) {
742     DIDumpOptions DumpOptions;
743     DumpOptions.Verbose = Verbose;
744     Prologue.dump(*OS, DumpOptions);
745   }
746 
747   if (PrologueErr) {
748     // Ensure there is a blank line after the prologue to clearly delineate it
749     // from later dumps.
750     if (OS)
751       *OS << "\n";
752     return PrologueErr;
753   }
754 
755   uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength();
756   if (!DebugLineData.isValidOffsetForDataOfSize(DebugLineOffset,
757                                                 ProgramLength)) {
758     assert(DebugLineData.size() > DebugLineOffset &&
759            "prologue parsing should handle invalid offset");
760     uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset;
761     RecoverableErrorHandler(
762         createStringError(errc::invalid_argument,
763                           "line table program with offset 0x%8.8" PRIx64
764                           " has length 0x%8.8" PRIx64 " but only 0x%8.8" PRIx64
765                           " bytes are available",
766                           DebugLineOffset, ProgramLength, BytesRemaining));
767     // Continue by capping the length at the number of remaining bytes.
768     ProgramLength = BytesRemaining;
769   }
770 
771   // Create a DataExtractor which can only see the data up to the end of the
772   // table, to prevent reading past the end.
773   const uint64_t EndOffset = DebugLineOffset + ProgramLength;
774   DWARFDataExtractor TableData(DebugLineData, EndOffset);
775 
776   // See if we should tell the data extractor the address size.
777   if (TableData.getAddressSize() == 0)
778     TableData.setAddressSize(Prologue.getAddressSize());
779   else
780     assert(Prologue.getAddressSize() == 0 ||
781            Prologue.getAddressSize() == TableData.getAddressSize());
782 
783   ParsingState State(this, DebugLineOffset, RecoverableErrorHandler);
784 
785   *OffsetPtr = DebugLineOffset + Prologue.getLength();
786   if (OS && *OffsetPtr < EndOffset) {
787     *OS << '\n';
788     Row::dumpTableHeader(*OS, /*Indent=*/Verbose ? 12 : 0);
789   }
790   bool TombstonedAddress = false;
791   auto EmitRow = [&] {
792     if (!TombstonedAddress) {
793       if (Verbose) {
794         *OS << "\n";
795         OS->indent(12);
796       }
797       if (OS)
798         State.Row.dump(*OS);
799       State.appendRowToMatrix();
800     }
801   };
802   while (*OffsetPtr < EndOffset) {
803     DataExtractor::Cursor Cursor(*OffsetPtr);
804 
805     if (Verbose)
806       *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr);
807 
808     uint64_t OpcodeOffset = *OffsetPtr;
809     uint8_t Opcode = TableData.getU8(Cursor);
810     size_t RowCount = Rows.size();
811 
812     if (Cursor && Verbose)
813       *OS << format("%02.02" PRIx8 " ", Opcode);
814 
815     if (Opcode == 0) {
816       // Extended Opcodes always start with a zero opcode followed by
817       // a uleb128 length so you can skip ones you don't know about
818       uint64_t Len = TableData.getULEB128(Cursor);
819       uint64_t ExtOffset = Cursor.tell();
820 
821       // Tolerate zero-length; assume length is correct and soldier on.
822       if (Len == 0) {
823         if (Cursor && Verbose)
824           *OS << "Badly formed extended line op (length 0)\n";
825         if (!Cursor) {
826           if (Verbose)
827             *OS << "\n";
828           RecoverableErrorHandler(Cursor.takeError());
829         }
830         *OffsetPtr = Cursor.tell();
831         continue;
832       }
833 
834       uint8_t SubOpcode = TableData.getU8(Cursor);
835       // OperandOffset will be the same as ExtOffset, if it was not possible to
836       // read the SubOpcode.
837       uint64_t OperandOffset = Cursor.tell();
838       if (Verbose)
839         *OS << LNExtendedString(SubOpcode);
840       switch (SubOpcode) {
841       case DW_LNE_end_sequence:
842         // Set the end_sequence register of the state machine to true and
843         // append a row to the matrix using the current values of the
844         // state-machine registers. Then reset the registers to the initial
845         // values specified above. Every statement program sequence must end
846         // with a DW_LNE_end_sequence instruction which creates a row whose
847         // address is that of the byte after the last target machine instruction
848         // of the sequence.
849         State.Row.EndSequence = true;
850         // No need to test the Cursor is valid here, since it must be to get
851         // into this code path - if it were invalid, the default case would be
852         // followed.
853         EmitRow();
854         State.resetRowAndSequence();
855         break;
856 
857       case DW_LNE_set_address:
858         // Takes a single relocatable address as an operand. The size of the
859         // operand is the size appropriate to hold an address on the target
860         // machine. Set the address register to the value given by the
861         // relocatable address. All of the other statement program opcodes
862         // that affect the address register add a delta to it. This instruction
863         // stores a relocatable value into it instead.
864         //
865         // Make sure the extractor knows the address size.  If not, infer it
866         // from the size of the operand.
867         {
868           uint8_t ExtractorAddressSize = TableData.getAddressSize();
869           uint64_t OpcodeAddressSize = Len - 1;
870           if (ExtractorAddressSize != OpcodeAddressSize &&
871               ExtractorAddressSize != 0)
872             RecoverableErrorHandler(createStringError(
873                 errc::invalid_argument,
874                 "mismatching address size at offset 0x%8.8" PRIx64
875                 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64,
876                 ExtOffset, ExtractorAddressSize, Len - 1));
877 
878           // Assume that the line table is correct and temporarily override the
879           // address size. If the size is unsupported, give up trying to read
880           // the address and continue to the next opcode.
881           if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 &&
882               OpcodeAddressSize != 4 && OpcodeAddressSize != 8) {
883             RecoverableErrorHandler(createStringError(
884                 errc::invalid_argument,
885                 "address size 0x%2.2" PRIx64
886                 " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64
887                 " is unsupported",
888                 OpcodeAddressSize, ExtOffset));
889             TableData.skip(Cursor, OpcodeAddressSize);
890           } else {
891             TableData.setAddressSize(OpcodeAddressSize);
892             State.Row.Address.Address = TableData.getRelocatedAddress(
893                 Cursor, &State.Row.Address.SectionIndex);
894 
895             uint64_t Tombstone =
896                 dwarf::computeTombstoneAddress(OpcodeAddressSize);
897             TombstonedAddress = State.Row.Address.Address == Tombstone;
898 
899             // Restore the address size if the extractor already had it.
900             if (ExtractorAddressSize != 0)
901               TableData.setAddressSize(ExtractorAddressSize);
902           }
903 
904           if (Cursor && Verbose) {
905             *OS << " (";
906             DWARFFormValue::dumpAddress(*OS, OpcodeAddressSize, State.Row.Address.Address);
907             *OS << ')';
908           }
909         }
910         break;
911 
912       case DW_LNE_define_file:
913         // Takes 4 arguments. The first is a null terminated string containing
914         // a source file name. The second is an unsigned LEB128 number
915         // representing the directory index of the directory in which the file
916         // was found. The third is an unsigned LEB128 number representing the
917         // time of last modification of the file. The fourth is an unsigned
918         // LEB128 number representing the length in bytes of the file. The time
919         // and length fields may contain LEB128(0) if the information is not
920         // available.
921         //
922         // The directory index represents an entry in the include_directories
923         // section of the statement program prologue. The index is LEB128(0)
924         // if the file was found in the current directory of the compilation,
925         // LEB128(1) if it was found in the first directory in the
926         // include_directories section, and so on. The directory index is
927         // ignored for file names that represent full path names.
928         //
929         // The files are numbered, starting at 1, in the order in which they
930         // appear; the names in the prologue come before names defined by
931         // the DW_LNE_define_file instruction. These numbers are used in the
932         // the file register of the state machine.
933         {
934           FileNameEntry FileEntry;
935           const char *Name = TableData.getCStr(Cursor);
936           FileEntry.Name =
937               DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name);
938           FileEntry.DirIdx = TableData.getULEB128(Cursor);
939           FileEntry.ModTime = TableData.getULEB128(Cursor);
940           FileEntry.Length = TableData.getULEB128(Cursor);
941           Prologue.FileNames.push_back(FileEntry);
942           if (Cursor && Verbose)
943             *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
944                 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
945                 << ", length=" << FileEntry.Length << ")";
946         }
947         break;
948 
949       case DW_LNE_set_discriminator:
950         State.Row.Discriminator = TableData.getULEB128(Cursor);
951         if (Cursor && Verbose)
952           *OS << " (" << State.Row.Discriminator << ")";
953         break;
954 
955       default:
956         if (Cursor && Verbose)
957           *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
958               << format(" length %" PRIx64, Len);
959         // Len doesn't include the zero opcode byte or the length itself, but
960         // it does include the sub_opcode, so we have to adjust for that.
961         TableData.skip(Cursor, Len - 1);
962         break;
963       }
964       // Make sure the length as recorded in the table and the standard length
965       // for the opcode match. If they don't, continue from the end as claimed
966       // by the table. Similarly, continue from the claimed end in the event of
967       // a parsing error.
968       uint64_t End = ExtOffset + Len;
969       if (Cursor && Cursor.tell() != End)
970         RecoverableErrorHandler(createStringError(
971             errc::illegal_byte_sequence,
972             "unexpected line op length at offset 0x%8.8" PRIx64
973             " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64,
974             ExtOffset, Len, Cursor.tell() - ExtOffset));
975       if (!Cursor && Verbose) {
976         DWARFDataExtractor::Cursor ByteCursor(OperandOffset);
977         uint8_t Byte = TableData.getU8(ByteCursor);
978         if (ByteCursor) {
979           *OS << " (<parsing error>";
980           do {
981             *OS << format(" %2.2" PRIx8, Byte);
982             Byte = TableData.getU8(ByteCursor);
983           } while (ByteCursor);
984           *OS << ")";
985         }
986 
987         // The only parse failure in this case should be if the end was reached.
988         // In that case, throw away the error, as the main Cursor's error will
989         // be sufficient.
990         consumeError(ByteCursor.takeError());
991       }
992       *OffsetPtr = End;
993     } else if (Opcode < Prologue.OpcodeBase) {
994       if (Verbose)
995         *OS << LNStandardString(Opcode);
996       switch (Opcode) {
997       // Standard Opcodes
998       case DW_LNS_copy:
999         // Takes no arguments. Append a row to the matrix using the
1000         // current values of the state-machine registers.
1001         EmitRow();
1002         break;
1003 
1004       case DW_LNS_advance_pc:
1005         // Takes a single unsigned LEB128 operand, multiplies it by the
1006         // min_inst_length field of the prologue, and adds the
1007         // result to the address register of the state machine.
1008         if (std::optional<uint64_t> Operand =
1009                 parseULEB128<uint64_t>(TableData, Cursor)) {
1010           uint64_t AddrOffset =
1011               State.advanceAddr(*Operand, Opcode, OpcodeOffset);
1012           if (Verbose)
1013             *OS << " (" << AddrOffset << ")";
1014         }
1015         break;
1016 
1017       case DW_LNS_advance_line:
1018         // Takes a single signed LEB128 operand and adds that value to
1019         // the line register of the state machine.
1020         {
1021           int64_t LineDelta = TableData.getSLEB128(Cursor);
1022           if (Cursor) {
1023             State.Row.Line += LineDelta;
1024             if (Verbose)
1025               *OS << " (" << State.Row.Line << ")";
1026           }
1027         }
1028         break;
1029 
1030       case DW_LNS_set_file:
1031         // Takes a single unsigned LEB128 operand and stores it in the file
1032         // register of the state machine.
1033         if (std::optional<uint16_t> File =
1034                 parseULEB128<uint16_t>(TableData, Cursor)) {
1035           State.Row.File = *File;
1036           if (Verbose)
1037             *OS << " (" << State.Row.File << ")";
1038         }
1039         break;
1040 
1041       case DW_LNS_set_column:
1042         // Takes a single unsigned LEB128 operand and stores it in the
1043         // column register of the state machine.
1044         if (std::optional<uint16_t> Column =
1045                 parseULEB128<uint16_t>(TableData, Cursor)) {
1046           State.Row.Column = *Column;
1047           if (Verbose)
1048             *OS << " (" << State.Row.Column << ")";
1049         }
1050         break;
1051 
1052       case DW_LNS_negate_stmt:
1053         // Takes no arguments. Set the is_stmt register of the state
1054         // machine to the logical negation of its current value.
1055         State.Row.IsStmt = !State.Row.IsStmt;
1056         break;
1057 
1058       case DW_LNS_set_basic_block:
1059         // Takes no arguments. Set the basic_block register of the
1060         // state machine to true
1061         State.Row.BasicBlock = true;
1062         break;
1063 
1064       case DW_LNS_const_add_pc:
1065         // Takes no arguments. Add to the address register of the state
1066         // machine the address increment value corresponding to special
1067         // opcode 255. The motivation for DW_LNS_const_add_pc is this:
1068         // when the statement program needs to advance the address by a
1069         // small amount, it can use a single special opcode, which occupies
1070         // a single byte. When it needs to advance the address by up to
1071         // twice the range of the last special opcode, it can use
1072         // DW_LNS_const_add_pc followed by a special opcode, for a total
1073         // of two bytes. Only if it needs to advance the address by more
1074         // than twice that range will it need to use both DW_LNS_advance_pc
1075         // and a special opcode, requiring three or more bytes.
1076         {
1077           uint64_t AddrOffset =
1078               State.advanceAddrForOpcode(Opcode, OpcodeOffset).AddrDelta;
1079           if (Verbose)
1080             *OS << format(" (0x%16.16" PRIx64 ")", AddrOffset);
1081         }
1082         break;
1083 
1084       case DW_LNS_fixed_advance_pc:
1085         // Takes a single uhalf operand. Add to the address register of
1086         // the state machine the value of the (unencoded) operand. This
1087         // is the only extended opcode that takes an argument that is not
1088         // a variable length number. The motivation for DW_LNS_fixed_advance_pc
1089         // is this: existing assemblers cannot emit DW_LNS_advance_pc or
1090         // special opcodes because they cannot encode LEB128 numbers or
1091         // judge when the computation of a special opcode overflows and
1092         // requires the use of DW_LNS_advance_pc. Such assemblers, however,
1093         // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
1094         {
1095           uint16_t PCOffset =
1096               TableData.getRelocatedValue(Cursor, 2);
1097           if (Cursor) {
1098             State.Row.Address.Address += PCOffset;
1099             if (Verbose)
1100               *OS << format(" (0x%4.4" PRIx16 ")", PCOffset);
1101           }
1102         }
1103         break;
1104 
1105       case DW_LNS_set_prologue_end:
1106         // Takes no arguments. Set the prologue_end register of the
1107         // state machine to true
1108         State.Row.PrologueEnd = true;
1109         break;
1110 
1111       case DW_LNS_set_epilogue_begin:
1112         // Takes no arguments. Set the basic_block register of the
1113         // state machine to true
1114         State.Row.EpilogueBegin = true;
1115         break;
1116 
1117       case DW_LNS_set_isa:
1118         // Takes a single unsigned LEB128 operand and stores it in the
1119         // ISA register of the state machine.
1120         if (std::optional<uint8_t> Isa =
1121                 parseULEB128<uint8_t>(TableData, Cursor)) {
1122           State.Row.Isa = *Isa;
1123           if (Verbose)
1124             *OS << " (" << (uint64_t)State.Row.Isa << ")";
1125         }
1126         break;
1127 
1128       default:
1129         // Handle any unknown standard opcodes here. We know the lengths
1130         // of such opcodes because they are specified in the prologue
1131         // as a multiple of LEB128 operands for each opcode.
1132         {
1133           assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
1134           if (Verbose)
1135             *OS << "Unrecognized standard opcode";
1136           uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
1137           std::vector<uint64_t> Operands;
1138           for (uint8_t I = 0; I < OpcodeLength; ++I) {
1139             if (std::optional<uint64_t> Value =
1140                     parseULEB128<uint64_t>(TableData, Cursor))
1141               Operands.push_back(*Value);
1142             else
1143               break;
1144           }
1145           if (Verbose && !Operands.empty()) {
1146             *OS << " (operands: ";
1147             bool First = true;
1148             for (uint64_t Value : Operands) {
1149               if (!First)
1150                 *OS << ", ";
1151               First = false;
1152               *OS << format("0x%16.16" PRIx64, Value);
1153             }
1154             if (Verbose)
1155               *OS << ')';
1156           }
1157         }
1158         break;
1159       }
1160 
1161       *OffsetPtr = Cursor.tell();
1162     } else {
1163       // Special Opcodes.
1164       ParsingState::AddrAndLineDelta Delta =
1165           State.handleSpecialOpcode(Opcode, OpcodeOffset);
1166 
1167       if (Verbose)
1168         *OS << "address += " << Delta.Address << ",  line += " << Delta.Line;
1169       EmitRow();
1170       *OffsetPtr = Cursor.tell();
1171     }
1172 
1173     // When a row is added to the matrix, it is also dumped, which includes a
1174     // new line already, so don't add an extra one.
1175     if (Verbose && Rows.size() == RowCount)
1176       *OS << "\n";
1177 
1178     // Most parse failures other than when parsing extended opcodes are due to
1179     // failures to read ULEBs. Bail out of parsing, since we don't know where to
1180     // continue reading from as there is no stated length for such byte
1181     // sequences. Print the final trailing new line if needed before doing so.
1182     if (!Cursor && Opcode != 0) {
1183       if (Verbose)
1184         *OS << "\n";
1185       return Cursor.takeError();
1186     }
1187 
1188     if (!Cursor)
1189       RecoverableErrorHandler(Cursor.takeError());
1190   }
1191 
1192   if (!State.Sequence.Empty)
1193     RecoverableErrorHandler(createStringError(
1194         errc::illegal_byte_sequence,
1195         "last sequence in debug line table at offset 0x%8.8" PRIx64
1196         " is not terminated",
1197         DebugLineOffset));
1198 
1199   // Sort all sequences so that address lookup will work faster.
1200   if (!Sequences.empty()) {
1201     llvm::sort(Sequences, Sequence::orderByHighPC);
1202     // Note: actually, instruction address ranges of sequences should not
1203     // overlap (in shared objects and executables). If they do, the address
1204     // lookup would still work, though, but result would be ambiguous.
1205     // We don't report warning in this case. For example,
1206     // sometimes .so compiled from multiple object files contains a few
1207     // rudimentary sequences for address ranges [0x0, 0xsomething).
1208   }
1209 
1210   // Terminate the table with a final blank line to clearly delineate it from
1211   // later dumps.
1212   if (OS)
1213     *OS << "\n";
1214 
1215   return Error::success();
1216 }
1217 
findRowInSeq(const DWARFDebugLine::Sequence & Seq,object::SectionedAddress Address) const1218 uint32_t DWARFDebugLine::LineTable::findRowInSeq(
1219     const DWARFDebugLine::Sequence &Seq,
1220     object::SectionedAddress Address) const {
1221   if (!Seq.containsPC(Address))
1222     return UnknownRowIndex;
1223   assert(Seq.SectionIndex == Address.SectionIndex);
1224   // In some cases, e.g. first instruction in a function, the compiler generates
1225   // two entries, both with the same address. We want the last one.
1226   //
1227   // In general we want a non-empty range: the last row whose address is less
1228   // than or equal to Address. This can be computed as upper_bound - 1.
1229   DWARFDebugLine::Row Row;
1230   Row.Address = Address;
1231   RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
1232   RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
1233   assert(FirstRow->Address.Address <= Row.Address.Address &&
1234          Row.Address.Address < LastRow[-1].Address.Address);
1235   RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row,
1236                                     DWARFDebugLine::Row::orderByAddress) -
1237                    1;
1238   assert(Seq.SectionIndex == RowPos->Address.SectionIndex);
1239   return RowPos - Rows.begin();
1240 }
1241 
lookupAddress(object::SectionedAddress Address) const1242 uint32_t DWARFDebugLine::LineTable::lookupAddress(
1243     object::SectionedAddress Address) const {
1244 
1245   // Search for relocatable addresses
1246   uint32_t Result = lookupAddressImpl(Address);
1247 
1248   if (Result != UnknownRowIndex ||
1249       Address.SectionIndex == object::SectionedAddress::UndefSection)
1250     return Result;
1251 
1252   // Search for absolute addresses
1253   Address.SectionIndex = object::SectionedAddress::UndefSection;
1254   return lookupAddressImpl(Address);
1255 }
1256 
lookupAddressImpl(object::SectionedAddress Address) const1257 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl(
1258     object::SectionedAddress Address) const {
1259   // First, find an instruction sequence containing the given address.
1260   DWARFDebugLine::Sequence Sequence;
1261   Sequence.SectionIndex = Address.SectionIndex;
1262   Sequence.HighPC = Address.Address;
1263   SequenceIter It = llvm::upper_bound(Sequences, Sequence,
1264                                       DWARFDebugLine::Sequence::orderByHighPC);
1265   if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex)
1266     return UnknownRowIndex;
1267   return findRowInSeq(*It, Address);
1268 }
1269 
lookupAddressRange(object::SectionedAddress Address,uint64_t Size,std::vector<uint32_t> & Result) const1270 bool DWARFDebugLine::LineTable::lookupAddressRange(
1271     object::SectionedAddress Address, uint64_t Size,
1272     std::vector<uint32_t> &Result) const {
1273 
1274   // Search for relocatable addresses
1275   if (lookupAddressRangeImpl(Address, Size, Result))
1276     return true;
1277 
1278   if (Address.SectionIndex == object::SectionedAddress::UndefSection)
1279     return false;
1280 
1281   // Search for absolute addresses
1282   Address.SectionIndex = object::SectionedAddress::UndefSection;
1283   return lookupAddressRangeImpl(Address, Size, Result);
1284 }
1285 
lookupAddressRangeImpl(object::SectionedAddress Address,uint64_t Size,std::vector<uint32_t> & Result) const1286 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
1287     object::SectionedAddress Address, uint64_t Size,
1288     std::vector<uint32_t> &Result) const {
1289   if (Sequences.empty())
1290     return false;
1291   uint64_t EndAddr = Address.Address + Size;
1292   // First, find an instruction sequence containing the given address.
1293   DWARFDebugLine::Sequence Sequence;
1294   Sequence.SectionIndex = Address.SectionIndex;
1295   Sequence.HighPC = Address.Address;
1296   SequenceIter LastSeq = Sequences.end();
1297   SequenceIter SeqPos = llvm::upper_bound(
1298       Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC);
1299   if (SeqPos == LastSeq || !SeqPos->containsPC(Address))
1300     return false;
1301 
1302   SequenceIter StartPos = SeqPos;
1303 
1304   // Add the rows from the first sequence to the vector, starting with the
1305   // index we just calculated
1306 
1307   while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
1308     const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
1309     // For the first sequence, we need to find which row in the sequence is the
1310     // first in our range.
1311     uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
1312     if (SeqPos == StartPos)
1313       FirstRowIndex = findRowInSeq(CurSeq, Address);
1314 
1315     // Figure out the last row in the range.
1316     uint32_t LastRowIndex =
1317         findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
1318     if (LastRowIndex == UnknownRowIndex)
1319       LastRowIndex = CurSeq.LastRowIndex - 1;
1320 
1321     assert(FirstRowIndex != UnknownRowIndex);
1322     assert(LastRowIndex != UnknownRowIndex);
1323 
1324     for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
1325       Result.push_back(I);
1326     }
1327 
1328     ++SeqPos;
1329   }
1330 
1331   return true;
1332 }
1333 
1334 std::optional<StringRef>
getSourceByIndex(uint64_t FileIndex,FileLineInfoKind Kind) const1335 DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
1336                                             FileLineInfoKind Kind) const {
1337   if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex))
1338     return std::nullopt;
1339   const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex);
1340   if (auto E = dwarf::toString(Entry.Source))
1341     return StringRef(*E);
1342   return std::nullopt;
1343 }
1344 
isPathAbsoluteOnWindowsOrPosix(const Twine & Path)1345 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
1346   // Debug info can contain paths from any OS, not necessarily
1347   // an OS we're currently running on. Moreover different compilation units can
1348   // be compiled on different operating systems and linked together later.
1349   return sys::path::is_absolute(Path, sys::path::Style::posix) ||
1350          sys::path::is_absolute(Path, sys::path::Style::windows);
1351 }
1352 
getFileNameByIndex(uint64_t FileIndex,StringRef CompDir,FileLineInfoKind Kind,std::string & Result,sys::path::Style Style) const1353 bool DWARFDebugLine::Prologue::getFileNameByIndex(
1354     uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind,
1355     std::string &Result, sys::path::Style Style) const {
1356   if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
1357     return false;
1358   const FileNameEntry &Entry = getFileNameEntry(FileIndex);
1359   auto E = dwarf::toString(Entry.Name);
1360   if (!E)
1361     return false;
1362   StringRef FileName = *E;
1363   if (Kind == FileLineInfoKind::RawValue ||
1364       isPathAbsoluteOnWindowsOrPosix(FileName)) {
1365     Result = std::string(FileName);
1366     return true;
1367   }
1368   if (Kind == FileLineInfoKind::BaseNameOnly) {
1369     Result = std::string(llvm::sys::path::filename(FileName));
1370     return true;
1371   }
1372 
1373   SmallString<16> FilePath;
1374   StringRef IncludeDir;
1375   // Be defensive about the contents of Entry.
1376   if (getVersion() >= 5) {
1377     // DirIdx 0 is the compilation directory, so don't include it for
1378     // relative names.
1379     if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) &&
1380         Entry.DirIdx < IncludeDirectories.size())
1381       IncludeDir = dwarf::toStringRef(IncludeDirectories[Entry.DirIdx]);
1382   } else {
1383     if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size())
1384       IncludeDir = dwarf::toStringRef(IncludeDirectories[Entry.DirIdx - 1]);
1385   }
1386 
1387   // For absolute paths only, include the compilation directory of compile unit,
1388   // unless v5 DirIdx == 0 (IncludeDir indicates the compilation directory). We
1389   // know that FileName is not absolute, the only way to have an absolute path
1390   // at this point would be if IncludeDir is absolute.
1391   if (Kind == FileLineInfoKind::AbsoluteFilePath &&
1392       (getVersion() < 5 || Entry.DirIdx != 0) && !CompDir.empty() &&
1393       !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1394     sys::path::append(FilePath, Style, CompDir);
1395 
1396   assert((Kind == FileLineInfoKind::AbsoluteFilePath ||
1397           Kind == FileLineInfoKind::RelativeFilePath) &&
1398          "invalid FileLineInfo Kind");
1399 
1400   // sys::path::append skips empty strings.
1401   sys::path::append(FilePath, Style, IncludeDir, FileName);
1402   Result = std::string(FilePath.str());
1403   return true;
1404 }
1405 
getFileLineInfoForAddress(object::SectionedAddress Address,const char * CompDir,FileLineInfoKind Kind,DILineInfo & Result) const1406 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1407     object::SectionedAddress Address, const char *CompDir,
1408     FileLineInfoKind Kind, DILineInfo &Result) const {
1409   // Get the index of row we're looking for in the line table.
1410   uint32_t RowIndex = lookupAddress(Address);
1411   if (RowIndex == -1U)
1412     return false;
1413   // Take file number and line/column from the row.
1414   const auto &Row = Rows[RowIndex];
1415   if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1416     return false;
1417   Result.Line = Row.Line;
1418   Result.Column = Row.Column;
1419   Result.Discriminator = Row.Discriminator;
1420   Result.Source = getSourceByIndex(Row.File, Kind);
1421   return true;
1422 }
1423 
getDirectoryForEntry(const FileNameEntry & Entry,std::string & Directory) const1424 bool DWARFDebugLine::LineTable::getDirectoryForEntry(
1425     const FileNameEntry &Entry, std::string &Directory) const {
1426   if (Prologue.getVersion() >= 5) {
1427     if (Entry.DirIdx < Prologue.IncludeDirectories.size()) {
1428       Directory =
1429           dwarf::toString(Prologue.IncludeDirectories[Entry.DirIdx], "");
1430       return true;
1431     }
1432     return false;
1433   }
1434   if (0 < Entry.DirIdx && Entry.DirIdx <= Prologue.IncludeDirectories.size()) {
1435     Directory =
1436         dwarf::toString(Prologue.IncludeDirectories[Entry.DirIdx - 1], "");
1437     return true;
1438   }
1439   return false;
1440 }
1441 
1442 // We want to supply the Unit associated with a .debug_line[.dwo] table when
1443 // we dump it, if possible, but still dump the table even if there isn't a Unit.
1444 // Therefore, collect up handles on all the Units that point into the
1445 // line-table section.
1446 static DWARFDebugLine::SectionParser::LineToUnitMap
buildLineToUnitMap(DWARFUnitVector::iterator_range Units)1447 buildLineToUnitMap(DWARFUnitVector::iterator_range Units) {
1448   DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit;
1449   for (const auto &U : Units)
1450     if (auto CUDIE = U->getUnitDIE())
1451       if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list)))
1452         LineToUnit.insert(std::make_pair(*StmtOffset, &*U));
1453   return LineToUnit;
1454 }
1455 
SectionParser(DWARFDataExtractor & Data,const DWARFContext & C,DWARFUnitVector::iterator_range Units)1456 DWARFDebugLine::SectionParser::SectionParser(
1457     DWARFDataExtractor &Data, const DWARFContext &C,
1458     DWARFUnitVector::iterator_range Units)
1459     : DebugLineData(Data), Context(C) {
1460   LineToUnit = buildLineToUnitMap(Units);
1461   if (!DebugLineData.isValidOffset(Offset))
1462     Done = true;
1463 }
1464 
totalLengthIsValid() const1465 bool DWARFDebugLine::Prologue::totalLengthIsValid() const {
1466   return TotalLength != 0u;
1467 }
1468 
parseNext(function_ref<void (Error)> RecoverableErrorHandler,function_ref<void (Error)> UnrecoverableErrorHandler,raw_ostream * OS,bool Verbose)1469 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext(
1470     function_ref<void(Error)> RecoverableErrorHandler,
1471     function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS,
1472     bool Verbose) {
1473   assert(DebugLineData.isValidOffset(Offset) &&
1474          "parsing should have terminated");
1475   DWARFUnit *U = prepareToParse(Offset);
1476   uint64_t OldOffset = Offset;
1477   LineTable LT;
1478   if (Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1479                            RecoverableErrorHandler, OS, Verbose))
1480     UnrecoverableErrorHandler(std::move(Err));
1481   moveToNextTable(OldOffset, LT.Prologue);
1482   return LT;
1483 }
1484 
skip(function_ref<void (Error)> RecoverableErrorHandler,function_ref<void (Error)> UnrecoverableErrorHandler)1485 void DWARFDebugLine::SectionParser::skip(
1486     function_ref<void(Error)> RecoverableErrorHandler,
1487     function_ref<void(Error)> UnrecoverableErrorHandler) {
1488   assert(DebugLineData.isValidOffset(Offset) &&
1489          "parsing should have terminated");
1490   DWARFUnit *U = prepareToParse(Offset);
1491   uint64_t OldOffset = Offset;
1492   LineTable LT;
1493   if (Error Err = LT.Prologue.parse(DebugLineData, &Offset,
1494                                     RecoverableErrorHandler, Context, U))
1495     UnrecoverableErrorHandler(std::move(Err));
1496   moveToNextTable(OldOffset, LT.Prologue);
1497 }
1498 
prepareToParse(uint64_t Offset)1499 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) {
1500   DWARFUnit *U = nullptr;
1501   auto It = LineToUnit.find(Offset);
1502   if (It != LineToUnit.end())
1503     U = It->second;
1504   DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0);
1505   return U;
1506 }
1507 
moveToNextTable(uint64_t OldOffset,const Prologue & P)1508 void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset,
1509                                                     const Prologue &P) {
1510   // If the length field is not valid, we don't know where the next table is, so
1511   // cannot continue to parse. Mark the parser as done, and leave the Offset
1512   // value as it currently is. This will be the end of the bad length field.
1513   if (!P.totalLengthIsValid()) {
1514     Done = true;
1515     return;
1516   }
1517 
1518   Offset = OldOffset + P.TotalLength + P.sizeofTotalLength();
1519   if (!DebugLineData.isValidOffset(Offset)) {
1520     Done = true;
1521   }
1522 }
1523