1 //===- DWARFDebugLine.cpp -------------------------------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 9 #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h" 10 #include "llvm/ADT/Optional.h" 11 #include "llvm/ADT/SmallString.h" 12 #include "llvm/ADT/SmallVector.h" 13 #include "llvm/ADT/StringRef.h" 14 #include "llvm/BinaryFormat/Dwarf.h" 15 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h" 16 #include "llvm/DebugInfo/DWARF/DWARFRelocMap.h" 17 #include "llvm/Support/Errc.h" 18 #include "llvm/Support/Format.h" 19 #include "llvm/Support/FormatVariadic.h" 20 #include "llvm/Support/WithColor.h" 21 #include "llvm/Support/raw_ostream.h" 22 #include <algorithm> 23 #include <cassert> 24 #include <cinttypes> 25 #include <cstdint> 26 #include <cstdio> 27 #include <utility> 28 29 using namespace llvm; 30 using namespace dwarf; 31 32 using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind; 33 34 namespace { 35 36 struct ContentDescriptor { 37 dwarf::LineNumberEntryFormat Type; 38 dwarf::Form Form; 39 }; 40 41 using ContentDescriptors = SmallVector<ContentDescriptor, 4>; 42 43 } // end anonymous namespace 44 45 static bool versionIsSupported(uint16_t Version) { 46 return Version >= 2 && Version <= 5; 47 } 48 49 void DWARFDebugLine::ContentTypeTracker::trackContentType( 50 dwarf::LineNumberEntryFormat ContentType) { 51 switch (ContentType) { 52 case dwarf::DW_LNCT_timestamp: 53 HasModTime = true; 54 break; 55 case dwarf::DW_LNCT_size: 56 HasLength = true; 57 break; 58 case dwarf::DW_LNCT_MD5: 59 HasMD5 = true; 60 break; 61 case dwarf::DW_LNCT_LLVM_source: 62 HasSource = true; 63 break; 64 default: 65 // We only care about values we consider optional, and new values may be 66 // added in the vendor extension range, so we do not match exhaustively. 67 break; 68 } 69 } 70 71 DWARFDebugLine::Prologue::Prologue() { clear(); } 72 73 bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const { 74 uint16_t DwarfVersion = getVersion(); 75 assert(DwarfVersion != 0 && 76 "line table prologue has no dwarf version information"); 77 if (DwarfVersion >= 5) 78 return FileIndex < FileNames.size(); 79 return FileIndex != 0 && FileIndex <= FileNames.size(); 80 } 81 82 Optional<uint64_t> DWARFDebugLine::Prologue::getLastValidFileIndex() const { 83 if (FileNames.empty()) 84 return None; 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 & 95 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 } 104 105 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 } 116 117 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 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> 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 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().getValue(); 327 break; 328 case DW_LNCT_timestamp: 329 FileEntry.ModTime = Value.getAsUnsignedConstant().getValue(); 330 break; 331 case DW_LNCT_size: 332 FileEntry.Length = Value.getAsUnsignedConstant().getValue(); 333 break; 334 case DW_LNCT_MD5: 335 if (!Value.getAsBlock() || Value.getAsBlock().getValue().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().getValue().begin(), 16, 340 FileEntry.Checksum.Bytes.begin()); 341 break; 342 default: 343 break; 344 } 345 } 346 FileNames.push_back(FileEntry); 347 } 348 return Error::success(); 349 } 350 351 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 } 358 359 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 } 461 462 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); } 463 464 void DWARFDebugLine::Row::postAppend() { 465 Discriminator = 0; 466 BasicBlock = false; 467 PrologueEnd = false; 468 EpilogueBegin = false; 469 } 470 471 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 } 485 486 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 } 493 494 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 } 502 503 DWARFDebugLine::Sequence::Sequence() { reset(); } 504 505 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 } 513 514 DWARFDebugLine::LineTable::LineTable() { clear(); } 515 516 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 } 532 533 void DWARFDebugLine::LineTable::clear() { 534 Prologue.clear(); 535 Rows.clear(); 536 Sequences.clear(); 537 } 538 539 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 } 545 546 void DWARFDebugLine::ParsingState::resetRowAndSequence() { 547 Row.reset(LineTable->Prologue.DefaultIsStmt); 548 Sequence.reset(); 549 } 550 551 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 * 573 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 } 579 580 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 } 599 600 static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) { 601 assert(Opcode != 0); 602 if (Opcode < OpcodeBase) 603 return LNStandardString(Opcode); 604 return "special"; 605 } 606 607 uint64_t DWARFDebugLine::ParsingState::advanceAddr(uint64_t OperationAdvance, 608 uint8_t Opcode, 609 uint64_t OpcodeOffset) { 610 StringRef OpcodeName = getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase); 611 // For versions less than 4, the MaxOpsPerInst member is set to 0, as the 612 // maximum_operations_per_instruction field wasn't introduced until DWARFv4. 613 // Don't warn about bad values in this situation. 614 if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 && 615 LineTable->Prologue.MaxOpsPerInst != 1) 616 ErrorHandler(createStringError( 617 errc::not_supported, 618 "line table program at offset 0x%8.8" PRIx64 619 " contains a %s opcode at offset 0x%8.8" PRIx64 620 ", but the prologue maximum_operations_per_instruction value is %" PRId8 621 ", which is unsupported. Assuming a value of 1 instead", 622 LineTableOffset, OpcodeName.data(), OpcodeOffset, 623 LineTable->Prologue.MaxOpsPerInst)); 624 if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0) 625 ErrorHandler( 626 createStringError(errc::invalid_argument, 627 "line table program at offset 0x%8.8" PRIx64 628 " contains a %s opcode at offset 0x%8.8" PRIx64 629 ", but the prologue minimum_instruction_length value " 630 "is 0, which prevents any address advancing", 631 LineTableOffset, OpcodeName.data(), OpcodeOffset)); 632 ReportAdvanceAddrProblem = false; 633 uint64_t AddrOffset = OperationAdvance * LineTable->Prologue.MinInstLength; 634 Row.Address.Address += AddrOffset; 635 return AddrOffset; 636 } 637 638 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode 639 DWARFDebugLine::ParsingState::advanceAddrForOpcode(uint8_t Opcode, 640 uint64_t OpcodeOffset) { 641 assert(Opcode == DW_LNS_const_add_pc || 642 Opcode >= LineTable->Prologue.OpcodeBase); 643 if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) { 644 StringRef OpcodeName = 645 getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase); 646 ErrorHandler( 647 createStringError(errc::not_supported, 648 "line table program at offset 0x%8.8" PRIx64 649 " contains a %s opcode at offset 0x%8.8" PRIx64 650 ", but the prologue line_range value is 0. The " 651 "address and line will not be adjusted", 652 LineTableOffset, OpcodeName.data(), OpcodeOffset)); 653 ReportBadLineRange = false; 654 } 655 656 uint8_t OpcodeValue = Opcode; 657 if (Opcode == DW_LNS_const_add_pc) 658 OpcodeValue = 255; 659 uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase; 660 uint64_t OperationAdvance = 661 LineTable->Prologue.LineRange != 0 662 ? AdjustedOpcode / LineTable->Prologue.LineRange 663 : 0; 664 uint64_t AddrOffset = advanceAddr(OperationAdvance, Opcode, OpcodeOffset); 665 return {AddrOffset, AdjustedOpcode}; 666 } 667 668 DWARFDebugLine::ParsingState::AddrAndLineDelta 669 DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode, 670 uint64_t OpcodeOffset) { 671 // A special opcode value is chosen based on the amount that needs 672 // to be added to the line and address registers. The maximum line 673 // increment for a special opcode is the value of the line_base 674 // field in the header, plus the value of the line_range field, 675 // minus 1 (line base + line range - 1). If the desired line 676 // increment is greater than the maximum line increment, a standard 677 // opcode must be used instead of a special opcode. The "address 678 // advance" is calculated by dividing the desired address increment 679 // by the minimum_instruction_length field from the header. The 680 // special opcode is then calculated using the following formula: 681 // 682 // opcode = (desired line increment - line_base) + 683 // (line_range * address advance) + opcode_base 684 // 685 // If the resulting opcode is greater than 255, a standard opcode 686 // must be used instead. 687 // 688 // To decode a special opcode, subtract the opcode_base from the 689 // opcode itself to give the adjusted opcode. The amount to 690 // increment the address register is the result of the adjusted 691 // opcode divided by the line_range multiplied by the 692 // minimum_instruction_length field from the header. That is: 693 // 694 // address increment = (adjusted opcode / line_range) * 695 // minimum_instruction_length 696 // 697 // The amount to increment the line register is the line_base plus 698 // the result of the adjusted opcode modulo the line_range. That is: 699 // 700 // line increment = line_base + (adjusted opcode % line_range) 701 702 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode AddrAdvanceResult = 703 advanceAddrForOpcode(Opcode, OpcodeOffset); 704 int32_t LineOffset = 0; 705 if (LineTable->Prologue.LineRange != 0) 706 LineOffset = 707 LineTable->Prologue.LineBase + 708 (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange); 709 Row.Line += LineOffset; 710 return {AddrAdvanceResult.AddrDelta, LineOffset}; 711 } 712 713 /// Parse a ULEB128 using the specified \p Cursor. \returns the parsed value on 714 /// success, or None if \p Cursor is in a failing state. 715 template <typename T> 716 static Optional<T> parseULEB128(DWARFDataExtractor &Data, 717 DataExtractor::Cursor &Cursor) { 718 T Value = Data.getULEB128(Cursor); 719 if (Cursor) 720 return Value; 721 return None; 722 } 723 724 Error DWARFDebugLine::LineTable::parse( 725 DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, 726 const DWARFContext &Ctx, const DWARFUnit *U, 727 function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS, 728 bool Verbose) { 729 assert((OS || !Verbose) && "cannot have verbose output without stream"); 730 const uint64_t DebugLineOffset = *OffsetPtr; 731 732 clear(); 733 734 Error PrologueErr = 735 Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U); 736 737 if (OS) { 738 DIDumpOptions DumpOptions; 739 DumpOptions.Verbose = Verbose; 740 Prologue.dump(*OS, DumpOptions); 741 } 742 743 if (PrologueErr) { 744 // Ensure there is a blank line after the prologue to clearly delineate it 745 // from later dumps. 746 if (OS) 747 *OS << "\n"; 748 return PrologueErr; 749 } 750 751 uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength(); 752 if (!DebugLineData.isValidOffsetForDataOfSize(DebugLineOffset, 753 ProgramLength)) { 754 assert(DebugLineData.size() > DebugLineOffset && 755 "prologue parsing should handle invalid offset"); 756 uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset; 757 RecoverableErrorHandler( 758 createStringError(errc::invalid_argument, 759 "line table program with offset 0x%8.8" PRIx64 760 " has length 0x%8.8" PRIx64 " but only 0x%8.8" PRIx64 761 " bytes are available", 762 DebugLineOffset, ProgramLength, BytesRemaining)); 763 // Continue by capping the length at the number of remaining bytes. 764 ProgramLength = BytesRemaining; 765 } 766 767 // Create a DataExtractor which can only see the data up to the end of the 768 // table, to prevent reading past the end. 769 const uint64_t EndOffset = DebugLineOffset + ProgramLength; 770 DWARFDataExtractor TableData(DebugLineData, EndOffset); 771 772 // See if we should tell the data extractor the address size. 773 if (TableData.getAddressSize() == 0) 774 TableData.setAddressSize(Prologue.getAddressSize()); 775 else 776 assert(Prologue.getAddressSize() == 0 || 777 Prologue.getAddressSize() == TableData.getAddressSize()); 778 779 ParsingState State(this, DebugLineOffset, RecoverableErrorHandler); 780 781 *OffsetPtr = DebugLineOffset + Prologue.getLength(); 782 if (OS && *OffsetPtr < EndOffset) { 783 *OS << '\n'; 784 Row::dumpTableHeader(*OS, /*Indent=*/Verbose ? 12 : 0); 785 } 786 bool TombstonedAddress = false; 787 auto EmitRow = [&] { 788 if (!TombstonedAddress) { 789 if (Verbose) { 790 *OS << "\n"; 791 OS->indent(12); 792 } 793 if (OS) 794 State.Row.dump(*OS); 795 State.appendRowToMatrix(); 796 } 797 }; 798 while (*OffsetPtr < EndOffset) { 799 DataExtractor::Cursor Cursor(*OffsetPtr); 800 801 if (Verbose) 802 *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr); 803 804 uint64_t OpcodeOffset = *OffsetPtr; 805 uint8_t Opcode = TableData.getU8(Cursor); 806 size_t RowCount = Rows.size(); 807 808 if (Cursor && Verbose) 809 *OS << format("%02.02" PRIx8 " ", Opcode); 810 811 if (Opcode == 0) { 812 // Extended Opcodes always start with a zero opcode followed by 813 // a uleb128 length so you can skip ones you don't know about 814 uint64_t Len = TableData.getULEB128(Cursor); 815 uint64_t ExtOffset = Cursor.tell(); 816 817 // Tolerate zero-length; assume length is correct and soldier on. 818 if (Len == 0) { 819 if (Cursor && Verbose) 820 *OS << "Badly formed extended line op (length 0)\n"; 821 if (!Cursor) { 822 if (Verbose) 823 *OS << "\n"; 824 RecoverableErrorHandler(Cursor.takeError()); 825 } 826 *OffsetPtr = Cursor.tell(); 827 continue; 828 } 829 830 uint8_t SubOpcode = TableData.getU8(Cursor); 831 // OperandOffset will be the same as ExtOffset, if it was not possible to 832 // read the SubOpcode. 833 uint64_t OperandOffset = Cursor.tell(); 834 if (Verbose) 835 *OS << LNExtendedString(SubOpcode); 836 switch (SubOpcode) { 837 case DW_LNE_end_sequence: 838 // Set the end_sequence register of the state machine to true and 839 // append a row to the matrix using the current values of the 840 // state-machine registers. Then reset the registers to the initial 841 // values specified above. Every statement program sequence must end 842 // with a DW_LNE_end_sequence instruction which creates a row whose 843 // address is that of the byte after the last target machine instruction 844 // of the sequence. 845 State.Row.EndSequence = true; 846 // No need to test the Cursor is valid here, since it must be to get 847 // into this code path - if it were invalid, the default case would be 848 // followed. 849 EmitRow(); 850 State.resetRowAndSequence(); 851 break; 852 853 case DW_LNE_set_address: 854 // Takes a single relocatable address as an operand. The size of the 855 // operand is the size appropriate to hold an address on the target 856 // machine. Set the address register to the value given by the 857 // relocatable address. All of the other statement program opcodes 858 // that affect the address register add a delta to it. This instruction 859 // stores a relocatable value into it instead. 860 // 861 // Make sure the extractor knows the address size. If not, infer it 862 // from the size of the operand. 863 { 864 uint8_t ExtractorAddressSize = TableData.getAddressSize(); 865 uint64_t OpcodeAddressSize = Len - 1; 866 if (ExtractorAddressSize != OpcodeAddressSize && 867 ExtractorAddressSize != 0) 868 RecoverableErrorHandler(createStringError( 869 errc::invalid_argument, 870 "mismatching address size at offset 0x%8.8" PRIx64 871 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64, 872 ExtOffset, ExtractorAddressSize, Len - 1)); 873 874 // Assume that the line table is correct and temporarily override the 875 // address size. If the size is unsupported, give up trying to read 876 // the address and continue to the next opcode. 877 if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 && 878 OpcodeAddressSize != 4 && OpcodeAddressSize != 8) { 879 RecoverableErrorHandler(createStringError( 880 errc::invalid_argument, 881 "address size 0x%2.2" PRIx64 882 " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64 883 " is unsupported", 884 OpcodeAddressSize, ExtOffset)); 885 TableData.skip(Cursor, OpcodeAddressSize); 886 } else { 887 TableData.setAddressSize(OpcodeAddressSize); 888 State.Row.Address.Address = TableData.getRelocatedAddress( 889 Cursor, &State.Row.Address.SectionIndex); 890 891 uint64_t Tombstone = 892 dwarf::computeTombstoneAddress(OpcodeAddressSize); 893 TombstonedAddress = State.Row.Address.Address == Tombstone; 894 895 // Restore the address size if the extractor already had it. 896 if (ExtractorAddressSize != 0) 897 TableData.setAddressSize(ExtractorAddressSize); 898 } 899 900 if (Cursor && Verbose) 901 *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address.Address); 902 } 903 break; 904 905 case DW_LNE_define_file: 906 // Takes 4 arguments. The first is a null terminated string containing 907 // a source file name. The second is an unsigned LEB128 number 908 // representing the directory index of the directory in which the file 909 // was found. The third is an unsigned LEB128 number representing the 910 // time of last modification of the file. The fourth is an unsigned 911 // LEB128 number representing the length in bytes of the file. The time 912 // and length fields may contain LEB128(0) if the information is not 913 // available. 914 // 915 // The directory index represents an entry in the include_directories 916 // section of the statement program prologue. The index is LEB128(0) 917 // if the file was found in the current directory of the compilation, 918 // LEB128(1) if it was found in the first directory in the 919 // include_directories section, and so on. The directory index is 920 // ignored for file names that represent full path names. 921 // 922 // The files are numbered, starting at 1, in the order in which they 923 // appear; the names in the prologue come before names defined by 924 // the DW_LNE_define_file instruction. These numbers are used in the 925 // the file register of the state machine. 926 { 927 FileNameEntry FileEntry; 928 const char *Name = TableData.getCStr(Cursor); 929 FileEntry.Name = 930 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name); 931 FileEntry.DirIdx = TableData.getULEB128(Cursor); 932 FileEntry.ModTime = TableData.getULEB128(Cursor); 933 FileEntry.Length = TableData.getULEB128(Cursor); 934 Prologue.FileNames.push_back(FileEntry); 935 if (Cursor && Verbose) 936 *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time=" 937 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime) 938 << ", length=" << FileEntry.Length << ")"; 939 } 940 break; 941 942 case DW_LNE_set_discriminator: 943 State.Row.Discriminator = TableData.getULEB128(Cursor); 944 if (Cursor && Verbose) 945 *OS << " (" << State.Row.Discriminator << ")"; 946 break; 947 948 default: 949 if (Cursor && Verbose) 950 *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode) 951 << format(" length %" PRIx64, Len); 952 // Len doesn't include the zero opcode byte or the length itself, but 953 // it does include the sub_opcode, so we have to adjust for that. 954 TableData.skip(Cursor, Len - 1); 955 break; 956 } 957 // Make sure the length as recorded in the table and the standard length 958 // for the opcode match. If they don't, continue from the end as claimed 959 // by the table. Similarly, continue from the claimed end in the event of 960 // a parsing error. 961 uint64_t End = ExtOffset + Len; 962 if (Cursor && Cursor.tell() != End) 963 RecoverableErrorHandler(createStringError( 964 errc::illegal_byte_sequence, 965 "unexpected line op length at offset 0x%8.8" PRIx64 966 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64, 967 ExtOffset, Len, Cursor.tell() - ExtOffset)); 968 if (!Cursor && Verbose) { 969 DWARFDataExtractor::Cursor ByteCursor(OperandOffset); 970 uint8_t Byte = TableData.getU8(ByteCursor); 971 if (ByteCursor) { 972 *OS << " (<parsing error>"; 973 do { 974 *OS << format(" %2.2" PRIx8, Byte); 975 Byte = TableData.getU8(ByteCursor); 976 } while (ByteCursor); 977 *OS << ")"; 978 } 979 980 // The only parse failure in this case should be if the end was reached. 981 // In that case, throw away the error, as the main Cursor's error will 982 // be sufficient. 983 consumeError(ByteCursor.takeError()); 984 } 985 *OffsetPtr = End; 986 } else if (Opcode < Prologue.OpcodeBase) { 987 if (Verbose) 988 *OS << LNStandardString(Opcode); 989 switch (Opcode) { 990 // Standard Opcodes 991 case DW_LNS_copy: 992 // Takes no arguments. Append a row to the matrix using the 993 // current values of the state-machine registers. 994 EmitRow(); 995 break; 996 997 case DW_LNS_advance_pc: 998 // Takes a single unsigned LEB128 operand, multiplies it by the 999 // min_inst_length field of the prologue, and adds the 1000 // result to the address register of the state machine. 1001 if (Optional<uint64_t> Operand = 1002 parseULEB128<uint64_t>(TableData, Cursor)) { 1003 uint64_t AddrOffset = 1004 State.advanceAddr(*Operand, Opcode, OpcodeOffset); 1005 if (Verbose) 1006 *OS << " (" << AddrOffset << ")"; 1007 } 1008 break; 1009 1010 case DW_LNS_advance_line: 1011 // Takes a single signed LEB128 operand and adds that value to 1012 // the line register of the state machine. 1013 { 1014 int64_t LineDelta = TableData.getSLEB128(Cursor); 1015 if (Cursor) { 1016 State.Row.Line += LineDelta; 1017 if (Verbose) 1018 *OS << " (" << State.Row.Line << ")"; 1019 } 1020 } 1021 break; 1022 1023 case DW_LNS_set_file: 1024 // Takes a single unsigned LEB128 operand and stores it in the file 1025 // register of the state machine. 1026 if (Optional<uint16_t> File = 1027 parseULEB128<uint16_t>(TableData, Cursor)) { 1028 State.Row.File = *File; 1029 if (Verbose) 1030 *OS << " (" << State.Row.File << ")"; 1031 } 1032 break; 1033 1034 case DW_LNS_set_column: 1035 // Takes a single unsigned LEB128 operand and stores it in the 1036 // column register of the state machine. 1037 if (Optional<uint16_t> Column = 1038 parseULEB128<uint16_t>(TableData, Cursor)) { 1039 State.Row.Column = *Column; 1040 if (Verbose) 1041 *OS << " (" << State.Row.Column << ")"; 1042 } 1043 break; 1044 1045 case DW_LNS_negate_stmt: 1046 // Takes no arguments. Set the is_stmt register of the state 1047 // machine to the logical negation of its current value. 1048 State.Row.IsStmt = !State.Row.IsStmt; 1049 break; 1050 1051 case DW_LNS_set_basic_block: 1052 // Takes no arguments. Set the basic_block register of the 1053 // state machine to true 1054 State.Row.BasicBlock = true; 1055 break; 1056 1057 case DW_LNS_const_add_pc: 1058 // Takes no arguments. Add to the address register of the state 1059 // machine the address increment value corresponding to special 1060 // opcode 255. The motivation for DW_LNS_const_add_pc is this: 1061 // when the statement program needs to advance the address by a 1062 // small amount, it can use a single special opcode, which occupies 1063 // a single byte. When it needs to advance the address by up to 1064 // twice the range of the last special opcode, it can use 1065 // DW_LNS_const_add_pc followed by a special opcode, for a total 1066 // of two bytes. Only if it needs to advance the address by more 1067 // than twice that range will it need to use both DW_LNS_advance_pc 1068 // and a special opcode, requiring three or more bytes. 1069 { 1070 uint64_t AddrOffset = 1071 State.advanceAddrForOpcode(Opcode, OpcodeOffset).AddrDelta; 1072 if (Verbose) 1073 *OS << format(" (0x%16.16" PRIx64 ")", AddrOffset); 1074 } 1075 break; 1076 1077 case DW_LNS_fixed_advance_pc: 1078 // Takes a single uhalf operand. Add to the address register of 1079 // the state machine the value of the (unencoded) operand. This 1080 // is the only extended opcode that takes an argument that is not 1081 // a variable length number. The motivation for DW_LNS_fixed_advance_pc 1082 // is this: existing assemblers cannot emit DW_LNS_advance_pc or 1083 // special opcodes because they cannot encode LEB128 numbers or 1084 // judge when the computation of a special opcode overflows and 1085 // requires the use of DW_LNS_advance_pc. Such assemblers, however, 1086 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression. 1087 { 1088 uint16_t PCOffset = 1089 TableData.getRelocatedValue(Cursor, 2); 1090 if (Cursor) { 1091 State.Row.Address.Address += PCOffset; 1092 if (Verbose) 1093 *OS << format(" (0x%4.4" PRIx16 ")", PCOffset); 1094 } 1095 } 1096 break; 1097 1098 case DW_LNS_set_prologue_end: 1099 // Takes no arguments. Set the prologue_end register of the 1100 // state machine to true 1101 State.Row.PrologueEnd = true; 1102 break; 1103 1104 case DW_LNS_set_epilogue_begin: 1105 // Takes no arguments. Set the basic_block register of the 1106 // state machine to true 1107 State.Row.EpilogueBegin = true; 1108 break; 1109 1110 case DW_LNS_set_isa: 1111 // Takes a single unsigned LEB128 operand and stores it in the 1112 // ISA register of the state machine. 1113 if (Optional<uint8_t> Isa = parseULEB128<uint8_t>(TableData, Cursor)) { 1114 State.Row.Isa = *Isa; 1115 if (Verbose) 1116 *OS << " (" << (uint64_t)State.Row.Isa << ")"; 1117 } 1118 break; 1119 1120 default: 1121 // Handle any unknown standard opcodes here. We know the lengths 1122 // of such opcodes because they are specified in the prologue 1123 // as a multiple of LEB128 operands for each opcode. 1124 { 1125 assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size()); 1126 if (Verbose) 1127 *OS << "Unrecognized standard opcode"; 1128 uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1]; 1129 std::vector<uint64_t> Operands; 1130 for (uint8_t I = 0; I < OpcodeLength; ++I) { 1131 if (Optional<uint64_t> Value = 1132 parseULEB128<uint64_t>(TableData, Cursor)) 1133 Operands.push_back(*Value); 1134 else 1135 break; 1136 } 1137 if (Verbose && !Operands.empty()) { 1138 *OS << " (operands: "; 1139 bool First = true; 1140 for (uint64_t Value : Operands) { 1141 if (!First) 1142 *OS << ", "; 1143 First = false; 1144 *OS << format("0x%16.16" PRIx64, Value); 1145 } 1146 if (Verbose) 1147 *OS << ')'; 1148 } 1149 } 1150 break; 1151 } 1152 1153 *OffsetPtr = Cursor.tell(); 1154 } else { 1155 // Special Opcodes. 1156 ParsingState::AddrAndLineDelta Delta = 1157 State.handleSpecialOpcode(Opcode, OpcodeOffset); 1158 1159 if (Verbose) 1160 *OS << "address += " << Delta.Address << ", line += " << Delta.Line; 1161 EmitRow(); 1162 *OffsetPtr = Cursor.tell(); 1163 } 1164 1165 // When a row is added to the matrix, it is also dumped, which includes a 1166 // new line already, so don't add an extra one. 1167 if (Verbose && Rows.size() == RowCount) 1168 *OS << "\n"; 1169 1170 // Most parse failures other than when parsing extended opcodes are due to 1171 // failures to read ULEBs. Bail out of parsing, since we don't know where to 1172 // continue reading from as there is no stated length for such byte 1173 // sequences. Print the final trailing new line if needed before doing so. 1174 if (!Cursor && Opcode != 0) { 1175 if (Verbose) 1176 *OS << "\n"; 1177 return Cursor.takeError(); 1178 } 1179 1180 if (!Cursor) 1181 RecoverableErrorHandler(Cursor.takeError()); 1182 } 1183 1184 if (!State.Sequence.Empty) 1185 RecoverableErrorHandler(createStringError( 1186 errc::illegal_byte_sequence, 1187 "last sequence in debug line table at offset 0x%8.8" PRIx64 1188 " is not terminated", 1189 DebugLineOffset)); 1190 1191 // Sort all sequences so that address lookup will work faster. 1192 if (!Sequences.empty()) { 1193 llvm::sort(Sequences, Sequence::orderByHighPC); 1194 // Note: actually, instruction address ranges of sequences should not 1195 // overlap (in shared objects and executables). If they do, the address 1196 // lookup would still work, though, but result would be ambiguous. 1197 // We don't report warning in this case. For example, 1198 // sometimes .so compiled from multiple object files contains a few 1199 // rudimentary sequences for address ranges [0x0, 0xsomething). 1200 } 1201 1202 // Terminate the table with a final blank line to clearly delineate it from 1203 // later dumps. 1204 if (OS) 1205 *OS << "\n"; 1206 1207 return Error::success(); 1208 } 1209 1210 uint32_t DWARFDebugLine::LineTable::findRowInSeq( 1211 const DWARFDebugLine::Sequence &Seq, 1212 object::SectionedAddress Address) const { 1213 if (!Seq.containsPC(Address)) 1214 return UnknownRowIndex; 1215 assert(Seq.SectionIndex == Address.SectionIndex); 1216 // In some cases, e.g. first instruction in a function, the compiler generates 1217 // two entries, both with the same address. We want the last one. 1218 // 1219 // In general we want a non-empty range: the last row whose address is less 1220 // than or equal to Address. This can be computed as upper_bound - 1. 1221 DWARFDebugLine::Row Row; 1222 Row.Address = Address; 1223 RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex; 1224 RowIter LastRow = Rows.begin() + Seq.LastRowIndex; 1225 assert(FirstRow->Address.Address <= Row.Address.Address && 1226 Row.Address.Address < LastRow[-1].Address.Address); 1227 RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row, 1228 DWARFDebugLine::Row::orderByAddress) - 1229 1; 1230 assert(Seq.SectionIndex == RowPos->Address.SectionIndex); 1231 return RowPos - Rows.begin(); 1232 } 1233 1234 uint32_t DWARFDebugLine::LineTable::lookupAddress( 1235 object::SectionedAddress Address) const { 1236 1237 // Search for relocatable addresses 1238 uint32_t Result = lookupAddressImpl(Address); 1239 1240 if (Result != UnknownRowIndex || 1241 Address.SectionIndex == object::SectionedAddress::UndefSection) 1242 return Result; 1243 1244 // Search for absolute addresses 1245 Address.SectionIndex = object::SectionedAddress::UndefSection; 1246 return lookupAddressImpl(Address); 1247 } 1248 1249 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl( 1250 object::SectionedAddress Address) const { 1251 // First, find an instruction sequence containing the given address. 1252 DWARFDebugLine::Sequence Sequence; 1253 Sequence.SectionIndex = Address.SectionIndex; 1254 Sequence.HighPC = Address.Address; 1255 SequenceIter It = llvm::upper_bound(Sequences, Sequence, 1256 DWARFDebugLine::Sequence::orderByHighPC); 1257 if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex) 1258 return UnknownRowIndex; 1259 return findRowInSeq(*It, Address); 1260 } 1261 1262 bool DWARFDebugLine::LineTable::lookupAddressRange( 1263 object::SectionedAddress Address, uint64_t Size, 1264 std::vector<uint32_t> &Result) const { 1265 1266 // Search for relocatable addresses 1267 if (lookupAddressRangeImpl(Address, Size, Result)) 1268 return true; 1269 1270 if (Address.SectionIndex == object::SectionedAddress::UndefSection) 1271 return false; 1272 1273 // Search for absolute addresses 1274 Address.SectionIndex = object::SectionedAddress::UndefSection; 1275 return lookupAddressRangeImpl(Address, Size, Result); 1276 } 1277 1278 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl( 1279 object::SectionedAddress Address, uint64_t Size, 1280 std::vector<uint32_t> &Result) const { 1281 if (Sequences.empty()) 1282 return false; 1283 uint64_t EndAddr = Address.Address + Size; 1284 // First, find an instruction sequence containing the given address. 1285 DWARFDebugLine::Sequence Sequence; 1286 Sequence.SectionIndex = Address.SectionIndex; 1287 Sequence.HighPC = Address.Address; 1288 SequenceIter LastSeq = Sequences.end(); 1289 SequenceIter SeqPos = llvm::upper_bound( 1290 Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC); 1291 if (SeqPos == LastSeq || !SeqPos->containsPC(Address)) 1292 return false; 1293 1294 SequenceIter StartPos = SeqPos; 1295 1296 // Add the rows from the first sequence to the vector, starting with the 1297 // index we just calculated 1298 1299 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) { 1300 const DWARFDebugLine::Sequence &CurSeq = *SeqPos; 1301 // For the first sequence, we need to find which row in the sequence is the 1302 // first in our range. 1303 uint32_t FirstRowIndex = CurSeq.FirstRowIndex; 1304 if (SeqPos == StartPos) 1305 FirstRowIndex = findRowInSeq(CurSeq, Address); 1306 1307 // Figure out the last row in the range. 1308 uint32_t LastRowIndex = 1309 findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex}); 1310 if (LastRowIndex == UnknownRowIndex) 1311 LastRowIndex = CurSeq.LastRowIndex - 1; 1312 1313 assert(FirstRowIndex != UnknownRowIndex); 1314 assert(LastRowIndex != UnknownRowIndex); 1315 1316 for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) { 1317 Result.push_back(I); 1318 } 1319 1320 ++SeqPos; 1321 } 1322 1323 return true; 1324 } 1325 1326 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex, 1327 FileLineInfoKind Kind) const { 1328 if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex)) 1329 return None; 1330 const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex); 1331 if (Optional<const char *> source = Entry.Source.getAsCString()) 1332 return StringRef(*source); 1333 return None; 1334 } 1335 1336 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) { 1337 // Debug info can contain paths from any OS, not necessarily 1338 // an OS we're currently running on. Moreover different compilation units can 1339 // be compiled on different operating systems and linked together later. 1340 return sys::path::is_absolute(Path, sys::path::Style::posix) || 1341 sys::path::is_absolute(Path, sys::path::Style::windows); 1342 } 1343 1344 bool DWARFDebugLine::Prologue::getFileNameByIndex( 1345 uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind, 1346 std::string &Result, sys::path::Style Style) const { 1347 if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex)) 1348 return false; 1349 const FileNameEntry &Entry = getFileNameEntry(FileIndex); 1350 Optional<const char *> Name = Entry.Name.getAsCString(); 1351 if (!Name) 1352 return false; 1353 StringRef FileName = *Name; 1354 if (Kind == FileLineInfoKind::RawValue || 1355 isPathAbsoluteOnWindowsOrPosix(FileName)) { 1356 Result = std::string(FileName); 1357 return true; 1358 } 1359 if (Kind == FileLineInfoKind::BaseNameOnly) { 1360 Result = std::string(llvm::sys::path::filename(FileName)); 1361 return true; 1362 } 1363 1364 SmallString<16> FilePath; 1365 StringRef IncludeDir; 1366 // Be defensive about the contents of Entry. 1367 if (getVersion() >= 5) { 1368 // DirIdx 0 is the compilation directory, so don't include it for 1369 // relative names. 1370 if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) && 1371 Entry.DirIdx < IncludeDirectories.size()) 1372 IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue(); 1373 } else { 1374 if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size()) 1375 IncludeDir = 1376 IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue(); 1377 } 1378 1379 // For absolute paths only, include the compilation directory of compile unit. 1380 // We know that FileName is not absolute, the only way to have an absolute 1381 // path at this point would be if IncludeDir is absolute. 1382 if (Kind == FileLineInfoKind::AbsoluteFilePath && !CompDir.empty() && 1383 !isPathAbsoluteOnWindowsOrPosix(IncludeDir)) 1384 sys::path::append(FilePath, Style, CompDir); 1385 1386 assert((Kind == FileLineInfoKind::AbsoluteFilePath || 1387 Kind == FileLineInfoKind::RelativeFilePath) && 1388 "invalid FileLineInfo Kind"); 1389 1390 // sys::path::append skips empty strings. 1391 sys::path::append(FilePath, Style, IncludeDir, FileName); 1392 Result = std::string(FilePath.str()); 1393 return true; 1394 } 1395 1396 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress( 1397 object::SectionedAddress Address, const char *CompDir, 1398 FileLineInfoKind Kind, DILineInfo &Result) const { 1399 // Get the index of row we're looking for in the line table. 1400 uint32_t RowIndex = lookupAddress(Address); 1401 if (RowIndex == -1U) 1402 return false; 1403 // Take file number and line/column from the row. 1404 const auto &Row = Rows[RowIndex]; 1405 if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName)) 1406 return false; 1407 Result.Line = Row.Line; 1408 Result.Column = Row.Column; 1409 Result.Discriminator = Row.Discriminator; 1410 Result.Source = getSourceByIndex(Row.File, Kind); 1411 return true; 1412 } 1413 1414 // We want to supply the Unit associated with a .debug_line[.dwo] table when 1415 // we dump it, if possible, but still dump the table even if there isn't a Unit. 1416 // Therefore, collect up handles on all the Units that point into the 1417 // line-table section. 1418 static DWARFDebugLine::SectionParser::LineToUnitMap 1419 buildLineToUnitMap(DWARFUnitVector::iterator_range Units) { 1420 DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit; 1421 for (const auto &U : Units) 1422 if (auto CUDIE = U->getUnitDIE()) 1423 if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list))) 1424 LineToUnit.insert(std::make_pair(*StmtOffset, &*U)); 1425 return LineToUnit; 1426 } 1427 1428 DWARFDebugLine::SectionParser::SectionParser( 1429 DWARFDataExtractor &Data, const DWARFContext &C, 1430 DWARFUnitVector::iterator_range Units) 1431 : DebugLineData(Data), Context(C) { 1432 LineToUnit = buildLineToUnitMap(Units); 1433 if (!DebugLineData.isValidOffset(Offset)) 1434 Done = true; 1435 } 1436 1437 bool DWARFDebugLine::Prologue::totalLengthIsValid() const { 1438 return TotalLength != 0u; 1439 } 1440 1441 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext( 1442 function_ref<void(Error)> RecoverableErrorHandler, 1443 function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS, 1444 bool Verbose) { 1445 assert(DebugLineData.isValidOffset(Offset) && 1446 "parsing should have terminated"); 1447 DWARFUnit *U = prepareToParse(Offset); 1448 uint64_t OldOffset = Offset; 1449 LineTable LT; 1450 if (Error Err = LT.parse(DebugLineData, &Offset, Context, U, 1451 RecoverableErrorHandler, OS, Verbose)) 1452 UnrecoverableErrorHandler(std::move(Err)); 1453 moveToNextTable(OldOffset, LT.Prologue); 1454 return LT; 1455 } 1456 1457 void DWARFDebugLine::SectionParser::skip( 1458 function_ref<void(Error)> RecoverableErrorHandler, 1459 function_ref<void(Error)> UnrecoverableErrorHandler) { 1460 assert(DebugLineData.isValidOffset(Offset) && 1461 "parsing should have terminated"); 1462 DWARFUnit *U = prepareToParse(Offset); 1463 uint64_t OldOffset = Offset; 1464 LineTable LT; 1465 if (Error Err = LT.Prologue.parse(DebugLineData, &Offset, 1466 RecoverableErrorHandler, Context, U)) 1467 UnrecoverableErrorHandler(std::move(Err)); 1468 moveToNextTable(OldOffset, LT.Prologue); 1469 } 1470 1471 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) { 1472 DWARFUnit *U = nullptr; 1473 auto It = LineToUnit.find(Offset); 1474 if (It != LineToUnit.end()) 1475 U = It->second; 1476 DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0); 1477 return U; 1478 } 1479 1480 void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset, 1481 const Prologue &P) { 1482 // If the length field is not valid, we don't know where the next table is, so 1483 // cannot continue to parse. Mark the parser as done, and leave the Offset 1484 // value as it currently is. This will be the end of the bad length field. 1485 if (!P.totalLengthIsValid()) { 1486 Done = true; 1487 return; 1488 } 1489 1490 Offset = OldOffset + P.TotalLength + P.sizeofTotalLength(); 1491 if (!DebugLineData.isValidOffset(Offset)) { 1492 Done = true; 1493 } 1494 } 1495