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 const llvm::DWARFDebugLine::FileNameEntry & 83 DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const { 84 uint16_t DwarfVersion = getVersion(); 85 assert(DwarfVersion != 0 && 86 "line table prologue has no dwarf version information"); 87 // In DWARF v5 the file names are 0-indexed. 88 if (DwarfVersion >= 5) 89 return FileNames[Index]; 90 return FileNames[Index - 1]; 91 } 92 93 void DWARFDebugLine::Prologue::clear() { 94 TotalLength = PrologueLength = 0; 95 SegSelectorSize = 0; 96 MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0; 97 OpcodeBase = 0; 98 FormParams = dwarf::FormParams({0, 0, DWARF32}); 99 ContentTypes = ContentTypeTracker(); 100 StandardOpcodeLengths.clear(); 101 IncludeDirectories.clear(); 102 FileNames.clear(); 103 } 104 105 void DWARFDebugLine::Prologue::dump(raw_ostream &OS, 106 DIDumpOptions DumpOptions) const { 107 if (!totalLengthIsValid()) 108 return; 109 OS << "Line table prologue:\n" 110 << format(" total_length: 0x%8.8" PRIx64 "\n", TotalLength) 111 << format(" version: %u\n", getVersion()); 112 if (!versionIsSupported(getVersion())) 113 return; 114 if (getVersion() >= 5) 115 OS << format(" address_size: %u\n", getAddressSize()) 116 << format(" seg_select_size: %u\n", SegSelectorSize); 117 OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength) 118 << format(" min_inst_length: %u\n", MinInstLength) 119 << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst) 120 << format(" default_is_stmt: %u\n", DefaultIsStmt) 121 << format(" line_base: %i\n", LineBase) 122 << format(" line_range: %u\n", LineRange) 123 << format(" opcode_base: %u\n", OpcodeBase); 124 125 for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I) 126 OS << formatv("standard_opcode_lengths[{0}] = {1}\n", 127 static_cast<dwarf::LineNumberOps>(I + 1), 128 StandardOpcodeLengths[I]); 129 130 if (!IncludeDirectories.empty()) { 131 // DWARF v5 starts directory indexes at 0. 132 uint32_t DirBase = getVersion() >= 5 ? 0 : 1; 133 for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) { 134 OS << format("include_directories[%3u] = ", I + DirBase); 135 IncludeDirectories[I].dump(OS, DumpOptions); 136 OS << '\n'; 137 } 138 } 139 140 if (!FileNames.empty()) { 141 // DWARF v5 starts file indexes at 0. 142 uint32_t FileBase = getVersion() >= 5 ? 0 : 1; 143 for (uint32_t I = 0; I != FileNames.size(); ++I) { 144 const FileNameEntry &FileEntry = FileNames[I]; 145 OS << format("file_names[%3u]:\n", I + FileBase); 146 OS << " name: "; 147 FileEntry.Name.dump(OS, DumpOptions); 148 OS << '\n' 149 << format(" dir_index: %" PRIu64 "\n", FileEntry.DirIdx); 150 if (ContentTypes.HasMD5) 151 OS << " md5_checksum: " << FileEntry.Checksum.digest() << '\n'; 152 if (ContentTypes.HasModTime) 153 OS << format(" mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime); 154 if (ContentTypes.HasLength) 155 OS << format(" length: 0x%8.8" PRIx64 "\n", FileEntry.Length); 156 if (ContentTypes.HasSource) { 157 OS << " source: "; 158 FileEntry.Source.dump(OS, DumpOptions); 159 OS << '\n'; 160 } 161 } 162 } 163 } 164 165 // Parse v2-v4 directory and file tables. 166 static Error 167 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData, 168 uint64_t *OffsetPtr, uint64_t EndPrologueOffset, 169 DWARFDebugLine::ContentTypeTracker &ContentTypes, 170 std::vector<DWARFFormValue> &IncludeDirectories, 171 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { 172 bool Terminated = false; 173 while (*OffsetPtr < EndPrologueOffset) { 174 StringRef S = DebugLineData.getCStrRef(OffsetPtr); 175 if (S.empty()) { 176 Terminated = true; 177 break; 178 } 179 DWARFFormValue Dir = 180 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, S.data()); 181 IncludeDirectories.push_back(Dir); 182 } 183 184 if (!Terminated) 185 return createStringError(errc::invalid_argument, 186 "include directories table was not null " 187 "terminated before the end of the prologue"); 188 189 Terminated = false; 190 while (*OffsetPtr < EndPrologueOffset) { 191 StringRef Name = DebugLineData.getCStrRef(OffsetPtr); 192 if (Name.empty()) { 193 Terminated = true; 194 break; 195 } 196 DWARFDebugLine::FileNameEntry FileEntry; 197 FileEntry.Name = 198 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name.data()); 199 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr); 200 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr); 201 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr); 202 FileNames.push_back(FileEntry); 203 } 204 205 ContentTypes.HasModTime = true; 206 ContentTypes.HasLength = true; 207 208 if (!Terminated) 209 return createStringError(errc::invalid_argument, 210 "file names table was not null terminated before " 211 "the end of the prologue"); 212 213 return Error::success(); 214 } 215 216 // Parse v5 directory/file entry content descriptions. 217 // Returns the descriptors, or an error if we did not find a path or ran off 218 // the end of the prologue. 219 static llvm::Expected<ContentDescriptors> 220 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, 221 DWARFDebugLine::ContentTypeTracker *ContentTypes) { 222 Error Err = Error::success(); 223 ContentDescriptors Descriptors; 224 int FormatCount = DebugLineData.getU8(OffsetPtr, &Err); 225 bool HasPath = false; 226 for (int I = 0; I != FormatCount && !Err; ++I) { 227 ContentDescriptor Descriptor; 228 Descriptor.Type = 229 dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr, &Err)); 230 Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr, &Err)); 231 if (Descriptor.Type == dwarf::DW_LNCT_path) 232 HasPath = true; 233 if (ContentTypes) 234 ContentTypes->trackContentType(Descriptor.Type); 235 Descriptors.push_back(Descriptor); 236 } 237 238 if (Err) 239 return createStringError(errc::invalid_argument, 240 "failed to parse entry content descriptors: %s", 241 toString(std::move(Err)).c_str()); 242 243 if (!HasPath) 244 return createStringError(errc::invalid_argument, 245 "failed to parse entry content descriptions" 246 " because no path was found"); 247 return Descriptors; 248 } 249 250 static Error 251 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData, 252 uint64_t *OffsetPtr, const dwarf::FormParams &FormParams, 253 const DWARFContext &Ctx, const DWARFUnit *U, 254 DWARFDebugLine::ContentTypeTracker &ContentTypes, 255 std::vector<DWARFFormValue> &IncludeDirectories, 256 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { 257 // Get the directory entry description. 258 llvm::Expected<ContentDescriptors> DirDescriptors = 259 parseV5EntryFormat(DebugLineData, OffsetPtr, nullptr); 260 if (!DirDescriptors) 261 return DirDescriptors.takeError(); 262 263 // Get the directory entries, according to the format described above. 264 uint64_t DirEntryCount = DebugLineData.getULEB128(OffsetPtr); 265 for (uint64_t I = 0; I != DirEntryCount; ++I) { 266 for (auto Descriptor : *DirDescriptors) { 267 DWARFFormValue Value(Descriptor.Form); 268 switch (Descriptor.Type) { 269 case DW_LNCT_path: 270 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U)) 271 return createStringError(errc::invalid_argument, 272 "failed to parse directory entry because " 273 "extracting the form value failed."); 274 IncludeDirectories.push_back(Value); 275 break; 276 default: 277 if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams)) 278 return createStringError(errc::invalid_argument, 279 "failed to parse directory entry because " 280 "skipping the form value failed."); 281 } 282 } 283 } 284 285 // Get the file entry description. 286 llvm::Expected<ContentDescriptors> FileDescriptors = 287 parseV5EntryFormat(DebugLineData, OffsetPtr, &ContentTypes); 288 if (!FileDescriptors) 289 return FileDescriptors.takeError(); 290 291 // Get the file entries, according to the format described above. 292 uint64_t FileEntryCount = DebugLineData.getULEB128(OffsetPtr); 293 for (uint64_t I = 0; I != FileEntryCount; ++I) { 294 DWARFDebugLine::FileNameEntry FileEntry; 295 for (auto Descriptor : *FileDescriptors) { 296 DWARFFormValue Value(Descriptor.Form); 297 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U)) 298 return createStringError(errc::invalid_argument, 299 "failed to parse file entry because " 300 "extracting the form value failed."); 301 switch (Descriptor.Type) { 302 case DW_LNCT_path: 303 FileEntry.Name = Value; 304 break; 305 case DW_LNCT_LLVM_source: 306 FileEntry.Source = Value; 307 break; 308 case DW_LNCT_directory_index: 309 FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue(); 310 break; 311 case DW_LNCT_timestamp: 312 FileEntry.ModTime = Value.getAsUnsignedConstant().getValue(); 313 break; 314 case DW_LNCT_size: 315 FileEntry.Length = Value.getAsUnsignedConstant().getValue(); 316 break; 317 case DW_LNCT_MD5: 318 if (!Value.getAsBlock() || Value.getAsBlock().getValue().size() != 16) 319 return createStringError( 320 errc::invalid_argument, 321 "failed to parse file entry because the MD5 hash is invalid"); 322 std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16, 323 FileEntry.Checksum.Bytes.begin()); 324 break; 325 default: 326 break; 327 } 328 } 329 FileNames.push_back(FileEntry); 330 } 331 return Error::success(); 332 } 333 334 uint64_t DWARFDebugLine::Prologue::getLength() const { 335 uint64_t Length = PrologueLength + sizeofTotalLength() + 336 sizeof(getVersion()) + sizeofPrologueLength(); 337 if (getVersion() >= 5) 338 Length += 2; // Address + Segment selector sizes. 339 return Length; 340 } 341 342 Error DWARFDebugLine::Prologue::parse( 343 const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, 344 function_ref<void(Error)> RecoverableErrorHandler, const DWARFContext &Ctx, 345 const DWARFUnit *U) { 346 const uint64_t PrologueOffset = *OffsetPtr; 347 348 clear(); 349 Error Err = Error::success(); 350 std::tie(TotalLength, FormParams.Format) = 351 DebugLineData.getInitialLength(OffsetPtr, &Err); 352 if (Err) 353 return createStringError( 354 errc::invalid_argument, 355 "parsing line table prologue at offset 0x%8.8" PRIx64 ": %s", 356 PrologueOffset, toString(std::move(Err)).c_str()); 357 358 FormParams.Version = DebugLineData.getU16(OffsetPtr); 359 if (!versionIsSupported(getVersion())) 360 // Treat this error as unrecoverable - we cannot be sure what any of 361 // the data represents including the length field, so cannot skip it or make 362 // any reasonable assumptions. 363 return createStringError( 364 errc::not_supported, 365 "parsing line table prologue at offset 0x%8.8" PRIx64 366 ": unsupported version %" PRIu16, 367 PrologueOffset, getVersion()); 368 369 if (getVersion() >= 5) { 370 FormParams.AddrSize = DebugLineData.getU8(OffsetPtr); 371 assert((DebugLineData.getAddressSize() == 0 || 372 DebugLineData.getAddressSize() == getAddressSize()) && 373 "Line table header and data extractor disagree"); 374 SegSelectorSize = DebugLineData.getU8(OffsetPtr); 375 } 376 377 PrologueLength = 378 DebugLineData.getRelocatedValue(sizeofPrologueLength(), OffsetPtr); 379 const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr; 380 MinInstLength = DebugLineData.getU8(OffsetPtr); 381 if (getVersion() >= 4) 382 MaxOpsPerInst = DebugLineData.getU8(OffsetPtr); 383 DefaultIsStmt = DebugLineData.getU8(OffsetPtr); 384 LineBase = DebugLineData.getU8(OffsetPtr); 385 LineRange = DebugLineData.getU8(OffsetPtr); 386 OpcodeBase = DebugLineData.getU8(OffsetPtr); 387 388 if (OpcodeBase == 0) { 389 // If the opcode base is 0, we cannot read the standard opcode lengths (of 390 // which there are supposed to be one fewer than the opcode base). Assume 391 // there are no standard opcodes and continue parsing. 392 RecoverableErrorHandler(createStringError( 393 errc::invalid_argument, 394 "parsing line table prologue at offset 0x%8.8" PRIx64 395 " found opcode base of 0. Assuming no standard opcodes", 396 PrologueOffset)); 397 } else { 398 StandardOpcodeLengths.reserve(OpcodeBase - 1); 399 for (uint32_t I = 1; I < OpcodeBase; ++I) { 400 uint8_t OpLen = DebugLineData.getU8(OffsetPtr); 401 StandardOpcodeLengths.push_back(OpLen); 402 } 403 } 404 405 auto ReportInvalidDirFileTable = [&](Error E) { 406 RecoverableErrorHandler(joinErrors( 407 createStringError( 408 errc::invalid_argument, 409 "parsing line table prologue at 0x%8.8" PRIx64 410 " found an invalid directory or file table description at" 411 " 0x%8.8" PRIx64, 412 PrologueOffset, *OffsetPtr), 413 std::move(E))); 414 // Skip to the end of the prologue, since the chances are that the parser 415 // did not read the whole table. This prevents the length check below from 416 // executing. 417 if (*OffsetPtr < EndPrologueOffset) 418 *OffsetPtr = EndPrologueOffset; 419 }; 420 if (getVersion() >= 5) { 421 if (Error E = 422 parseV5DirFileTables(DebugLineData, OffsetPtr, FormParams, Ctx, U, 423 ContentTypes, IncludeDirectories, FileNames)) 424 ReportInvalidDirFileTable(std::move(E)); 425 } else if (Error E = parseV2DirFileTables(DebugLineData, OffsetPtr, 426 EndPrologueOffset, ContentTypes, 427 IncludeDirectories, FileNames)) 428 ReportInvalidDirFileTable(std::move(E)); 429 430 if (*OffsetPtr != EndPrologueOffset) { 431 RecoverableErrorHandler(createStringError( 432 errc::invalid_argument, 433 "parsing line table prologue at 0x%8.8" PRIx64 434 " should have ended at 0x%8.8" PRIx64 " but it ended at 0x%8.8" PRIx64, 435 PrologueOffset, EndPrologueOffset, *OffsetPtr)); 436 *OffsetPtr = EndPrologueOffset; 437 } 438 return Error::success(); 439 } 440 441 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); } 442 443 void DWARFDebugLine::Row::postAppend() { 444 Discriminator = 0; 445 BasicBlock = false; 446 PrologueEnd = false; 447 EpilogueBegin = false; 448 } 449 450 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) { 451 Address.Address = 0; 452 Address.SectionIndex = object::SectionedAddress::UndefSection; 453 Line = 1; 454 Column = 0; 455 File = 1; 456 Isa = 0; 457 Discriminator = 0; 458 IsStmt = DefaultIsStmt; 459 BasicBlock = false; 460 EndSequence = false; 461 PrologueEnd = false; 462 EpilogueBegin = false; 463 } 464 465 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) { 466 OS << "Address Line Column File ISA Discriminator Flags\n" 467 << "------------------ ------ ------ ------ --- ------------- " 468 "-------------\n"; 469 } 470 471 void DWARFDebugLine::Row::dump(raw_ostream &OS) const { 472 OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column) 473 << format(" %6u %3u %13u ", File, Isa, Discriminator) 474 << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "") 475 << (PrologueEnd ? " prologue_end" : "") 476 << (EpilogueBegin ? " epilogue_begin" : "") 477 << (EndSequence ? " end_sequence" : "") << '\n'; 478 } 479 480 DWARFDebugLine::Sequence::Sequence() { reset(); } 481 482 void DWARFDebugLine::Sequence::reset() { 483 LowPC = 0; 484 HighPC = 0; 485 SectionIndex = object::SectionedAddress::UndefSection; 486 FirstRowIndex = 0; 487 LastRowIndex = 0; 488 Empty = true; 489 } 490 491 DWARFDebugLine::LineTable::LineTable() { clear(); } 492 493 void DWARFDebugLine::LineTable::dump(raw_ostream &OS, 494 DIDumpOptions DumpOptions) const { 495 Prologue.dump(OS, DumpOptions); 496 497 if (!Rows.empty()) { 498 OS << '\n'; 499 Row::dumpTableHeader(OS); 500 for (const Row &R : Rows) { 501 R.dump(OS); 502 } 503 } 504 505 // Terminate the table with a final blank line to clearly delineate it from 506 // later dumps. 507 OS << '\n'; 508 } 509 510 void DWARFDebugLine::LineTable::clear() { 511 Prologue.clear(); 512 Rows.clear(); 513 Sequences.clear(); 514 } 515 516 DWARFDebugLine::ParsingState::ParsingState( 517 struct LineTable *LT, uint64_t TableOffset, 518 function_ref<void(Error)> ErrorHandler) 519 : LineTable(LT), LineTableOffset(TableOffset), ErrorHandler(ErrorHandler) { 520 resetRowAndSequence(); 521 } 522 523 void DWARFDebugLine::ParsingState::resetRowAndSequence() { 524 Row.reset(LineTable->Prologue.DefaultIsStmt); 525 Sequence.reset(); 526 } 527 528 void DWARFDebugLine::ParsingState::appendRowToMatrix() { 529 unsigned RowNumber = LineTable->Rows.size(); 530 if (Sequence.Empty) { 531 // Record the beginning of instruction sequence. 532 Sequence.Empty = false; 533 Sequence.LowPC = Row.Address.Address; 534 Sequence.FirstRowIndex = RowNumber; 535 } 536 LineTable->appendRow(Row); 537 if (Row.EndSequence) { 538 // Record the end of instruction sequence. 539 Sequence.HighPC = Row.Address.Address; 540 Sequence.LastRowIndex = RowNumber + 1; 541 Sequence.SectionIndex = Row.Address.SectionIndex; 542 if (Sequence.isValid()) 543 LineTable->appendSequence(Sequence); 544 Sequence.reset(); 545 } 546 Row.postAppend(); 547 } 548 549 const DWARFDebugLine::LineTable * 550 DWARFDebugLine::getLineTable(uint64_t Offset) const { 551 LineTableConstIter Pos = LineTableMap.find(Offset); 552 if (Pos != LineTableMap.end()) 553 return &Pos->second; 554 return nullptr; 555 } 556 557 Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable( 558 DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx, 559 const DWARFUnit *U, function_ref<void(Error)> RecoverableErrorHandler) { 560 if (!DebugLineData.isValidOffset(Offset)) 561 return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx64 562 " is not a valid debug line section offset", 563 Offset); 564 565 std::pair<LineTableIter, bool> Pos = 566 LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable())); 567 LineTable *LT = &Pos.first->second; 568 if (Pos.second) { 569 if (Error Err = 570 LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorHandler)) 571 return std::move(Err); 572 return LT; 573 } 574 return LT; 575 } 576 577 static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) { 578 assert(Opcode != 0); 579 if (Opcode < OpcodeBase) 580 return LNStandardString(Opcode); 581 return "special"; 582 } 583 584 uint64_t DWARFDebugLine::ParsingState::advanceAddr(uint64_t OperationAdvance, 585 uint8_t Opcode, 586 uint64_t OpcodeOffset) { 587 StringRef OpcodeName = getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase); 588 // For versions less than 4, the MaxOpsPerInst member is set to 0, as the 589 // maximum_operations_per_instruction field wasn't introduced until DWARFv4. 590 // Don't warn about bad values in this situation. 591 if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 && 592 LineTable->Prologue.MaxOpsPerInst != 1) 593 ErrorHandler(createStringError( 594 errc::not_supported, 595 "line table program at offset 0x%8.8" PRIx64 596 " contains a %s opcode at offset 0x%8.8" PRIx64 597 ", but the prologue maximum_operations_per_instruction value is %" PRId8 598 ", which is unsupported. Assuming a value of 1 instead", 599 LineTableOffset, OpcodeName.data(), OpcodeOffset, 600 LineTable->Prologue.MaxOpsPerInst)); 601 if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0) 602 ErrorHandler( 603 createStringError(errc::invalid_argument, 604 "line table program at offset 0x%8.8" PRIx64 605 " contains a %s opcode at offset 0x%8.8" PRIx64 606 ", but the prologue minimum_instruction_length value " 607 "is 0, which prevents any address advancing", 608 LineTableOffset, OpcodeName.data(), OpcodeOffset)); 609 ReportAdvanceAddrProblem = false; 610 uint64_t AddrOffset = OperationAdvance * LineTable->Prologue.MinInstLength; 611 Row.Address.Address += AddrOffset; 612 return AddrOffset; 613 } 614 615 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode 616 DWARFDebugLine::ParsingState::advanceAddrForOpcode(uint8_t Opcode, 617 uint64_t OpcodeOffset) { 618 assert(Opcode == DW_LNS_const_add_pc || 619 Opcode >= LineTable->Prologue.OpcodeBase); 620 if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) { 621 StringRef OpcodeName = 622 getOpcodeName(Opcode, LineTable->Prologue.OpcodeBase); 623 ErrorHandler( 624 createStringError(errc::not_supported, 625 "line table program at offset 0x%8.8" PRIx64 626 " contains a %s opcode at offset 0x%8.8" PRIx64 627 ", but the prologue line_range value is 0. The " 628 "address and line will not be adjusted", 629 LineTableOffset, OpcodeName.data(), OpcodeOffset)); 630 ReportBadLineRange = false; 631 } 632 633 uint8_t OpcodeValue = Opcode; 634 if (Opcode == DW_LNS_const_add_pc) 635 OpcodeValue = 255; 636 uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase; 637 uint64_t OperationAdvance = 638 LineTable->Prologue.LineRange != 0 639 ? AdjustedOpcode / LineTable->Prologue.LineRange 640 : 0; 641 uint64_t AddrOffset = advanceAddr(OperationAdvance, Opcode, OpcodeOffset); 642 return {AddrOffset, AdjustedOpcode}; 643 } 644 645 DWARFDebugLine::ParsingState::AddrAndLineDelta 646 DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode, 647 uint64_t OpcodeOffset) { 648 // A special opcode value is chosen based on the amount that needs 649 // to be added to the line and address registers. The maximum line 650 // increment for a special opcode is the value of the line_base 651 // field in the header, plus the value of the line_range field, 652 // minus 1 (line base + line range - 1). If the desired line 653 // increment is greater than the maximum line increment, a standard 654 // opcode must be used instead of a special opcode. The "address 655 // advance" is calculated by dividing the desired address increment 656 // by the minimum_instruction_length field from the header. The 657 // special opcode is then calculated using the following formula: 658 // 659 // opcode = (desired line increment - line_base) + 660 // (line_range * address advance) + opcode_base 661 // 662 // If the resulting opcode is greater than 255, a standard opcode 663 // must be used instead. 664 // 665 // To decode a special opcode, subtract the opcode_base from the 666 // opcode itself to give the adjusted opcode. The amount to 667 // increment the address register is the result of the adjusted 668 // opcode divided by the line_range multiplied by the 669 // minimum_instruction_length field from the header. That is: 670 // 671 // address increment = (adjusted opcode / line_range) * 672 // minimum_instruction_length 673 // 674 // The amount to increment the line register is the line_base plus 675 // the result of the adjusted opcode modulo the line_range. That is: 676 // 677 // line increment = line_base + (adjusted opcode % line_range) 678 679 DWARFDebugLine::ParsingState::AddrAndAdjustedOpcode AddrAdvanceResult = 680 advanceAddrForOpcode(Opcode, OpcodeOffset); 681 int32_t LineOffset = 0; 682 if (LineTable->Prologue.LineRange != 0) 683 LineOffset = 684 LineTable->Prologue.LineBase + 685 (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange); 686 Row.Line += LineOffset; 687 return {AddrAdvanceResult.AddrDelta, LineOffset}; 688 } 689 690 Error DWARFDebugLine::LineTable::parse( 691 DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, 692 const DWARFContext &Ctx, const DWARFUnit *U, 693 function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS) { 694 const uint64_t DebugLineOffset = *OffsetPtr; 695 696 clear(); 697 698 Error PrologueErr = 699 Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U); 700 701 if (OS) { 702 // The presence of OS signals verbose dumping. 703 DIDumpOptions DumpOptions; 704 DumpOptions.Verbose = true; 705 Prologue.dump(*OS, DumpOptions); 706 } 707 708 if (PrologueErr) 709 return PrologueErr; 710 711 uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength(); 712 if (!DebugLineData.isValidOffsetForDataOfSize(DebugLineOffset, 713 ProgramLength)) { 714 assert(DebugLineData.size() > DebugLineOffset && 715 "prologue parsing should handle invalid offset"); 716 uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset; 717 RecoverableErrorHandler( 718 createStringError(errc::invalid_argument, 719 "line table program with offset 0x%8.8" PRIx64 720 " has length 0x%8.8" PRIx64 " but only 0x%8.8" PRIx64 721 " bytes are available", 722 DebugLineOffset, ProgramLength, BytesRemaining)); 723 // Continue by capping the length at the number of remaining bytes. 724 ProgramLength = BytesRemaining; 725 } 726 727 const uint64_t EndOffset = DebugLineOffset + ProgramLength; 728 729 // See if we should tell the data extractor the address size. 730 if (DebugLineData.getAddressSize() == 0) 731 DebugLineData.setAddressSize(Prologue.getAddressSize()); 732 else 733 assert(Prologue.getAddressSize() == 0 || 734 Prologue.getAddressSize() == DebugLineData.getAddressSize()); 735 736 ParsingState State(this, DebugLineOffset, RecoverableErrorHandler); 737 738 while (*OffsetPtr < EndOffset) { 739 if (OS) 740 *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr); 741 742 uint64_t OpcodeOffset = *OffsetPtr; 743 uint8_t Opcode = DebugLineData.getU8(OffsetPtr); 744 745 if (OS) 746 *OS << format("%02.02" PRIx8 " ", Opcode); 747 748 if (Opcode == 0) { 749 // Extended Opcodes always start with a zero opcode followed by 750 // a uleb128 length so you can skip ones you don't know about 751 uint64_t Len = DebugLineData.getULEB128(OffsetPtr); 752 uint64_t ExtOffset = *OffsetPtr; 753 754 // Tolerate zero-length; assume length is correct and soldier on. 755 if (Len == 0) { 756 if (OS) 757 *OS << "Badly formed extended line op (length 0)\n"; 758 continue; 759 } 760 761 uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr); 762 if (OS) 763 *OS << LNExtendedString(SubOpcode); 764 switch (SubOpcode) { 765 case DW_LNE_end_sequence: 766 // Set the end_sequence register of the state machine to true and 767 // append a row to the matrix using the current values of the 768 // state-machine registers. Then reset the registers to the initial 769 // values specified above. Every statement program sequence must end 770 // with a DW_LNE_end_sequence instruction which creates a row whose 771 // address is that of the byte after the last target machine instruction 772 // of the sequence. 773 State.Row.EndSequence = true; 774 if (OS) { 775 *OS << "\n"; 776 OS->indent(12); 777 State.Row.dump(*OS); 778 } 779 State.appendRowToMatrix(); 780 State.resetRowAndSequence(); 781 break; 782 783 case DW_LNE_set_address: 784 // Takes a single relocatable address as an operand. The size of the 785 // operand is the size appropriate to hold an address on the target 786 // machine. Set the address register to the value given by the 787 // relocatable address. All of the other statement program opcodes 788 // that affect the address register add a delta to it. This instruction 789 // stores a relocatable value into it instead. 790 // 791 // Make sure the extractor knows the address size. If not, infer it 792 // from the size of the operand. 793 { 794 uint8_t ExtractorAddressSize = DebugLineData.getAddressSize(); 795 uint64_t OpcodeAddressSize = Len - 1; 796 if (ExtractorAddressSize != OpcodeAddressSize && 797 ExtractorAddressSize != 0) 798 RecoverableErrorHandler(createStringError( 799 errc::invalid_argument, 800 "mismatching address size at offset 0x%8.8" PRIx64 801 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64, 802 ExtOffset, ExtractorAddressSize, Len - 1)); 803 804 // Assume that the line table is correct and temporarily override the 805 // address size. If the size is unsupported, give up trying to read 806 // the address and continue to the next opcode. 807 if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 && 808 OpcodeAddressSize != 4 && OpcodeAddressSize != 8) { 809 RecoverableErrorHandler(createStringError( 810 errc::invalid_argument, 811 "address size 0x%2.2" PRIx64 812 " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64 813 " is unsupported", 814 OpcodeAddressSize, ExtOffset)); 815 *OffsetPtr += OpcodeAddressSize; 816 } else { 817 DebugLineData.setAddressSize(OpcodeAddressSize); 818 State.Row.Address.Address = DebugLineData.getRelocatedAddress( 819 OffsetPtr, &State.Row.Address.SectionIndex); 820 821 // Restore the address size if the extractor already had it. 822 if (ExtractorAddressSize != 0) 823 DebugLineData.setAddressSize(ExtractorAddressSize); 824 } 825 826 if (OS) 827 *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address.Address); 828 } 829 break; 830 831 case DW_LNE_define_file: 832 // Takes 4 arguments. The first is a null terminated string containing 833 // a source file name. The second is an unsigned LEB128 number 834 // representing the directory index of the directory in which the file 835 // was found. The third is an unsigned LEB128 number representing the 836 // time of last modification of the file. The fourth is an unsigned 837 // LEB128 number representing the length in bytes of the file. The time 838 // and length fields may contain LEB128(0) if the information is not 839 // available. 840 // 841 // The directory index represents an entry in the include_directories 842 // section of the statement program prologue. The index is LEB128(0) 843 // if the file was found in the current directory of the compilation, 844 // LEB128(1) if it was found in the first directory in the 845 // include_directories section, and so on. The directory index is 846 // ignored for file names that represent full path names. 847 // 848 // The files are numbered, starting at 1, in the order in which they 849 // appear; the names in the prologue come before names defined by 850 // the DW_LNE_define_file instruction. These numbers are used in the 851 // the file register of the state machine. 852 { 853 FileNameEntry FileEntry; 854 const char *Name = DebugLineData.getCStr(OffsetPtr); 855 FileEntry.Name = 856 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name); 857 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr); 858 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr); 859 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr); 860 Prologue.FileNames.push_back(FileEntry); 861 if (OS) 862 *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time=" 863 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime) 864 << ", length=" << FileEntry.Length << ")"; 865 } 866 break; 867 868 case DW_LNE_set_discriminator: 869 State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr); 870 if (OS) 871 *OS << " (" << State.Row.Discriminator << ")"; 872 break; 873 874 default: 875 if (OS) 876 *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode) 877 << format(" length %" PRIx64, Len); 878 // Len doesn't include the zero opcode byte or the length itself, but 879 // it does include the sub_opcode, so we have to adjust for that. 880 (*OffsetPtr) += Len - 1; 881 break; 882 } 883 // Make sure the length as recorded in the table and the standard length 884 // for the opcode match. If they don't, continue from the end as claimed 885 // by the table. 886 uint64_t End = ExtOffset + Len; 887 if (*OffsetPtr != End) { 888 RecoverableErrorHandler(createStringError( 889 errc::illegal_byte_sequence, 890 "unexpected line op length at offset 0x%8.8" PRIx64 891 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64, 892 ExtOffset, Len, *OffsetPtr - ExtOffset)); 893 *OffsetPtr = End; 894 } 895 } else if (Opcode < Prologue.OpcodeBase) { 896 if (OS) 897 *OS << LNStandardString(Opcode); 898 switch (Opcode) { 899 // Standard Opcodes 900 case DW_LNS_copy: 901 // Takes no arguments. Append a row to the matrix using the 902 // current values of the state-machine registers. 903 if (OS) { 904 *OS << "\n"; 905 OS->indent(12); 906 State.Row.dump(*OS); 907 *OS << "\n"; 908 } 909 State.appendRowToMatrix(); 910 break; 911 912 case DW_LNS_advance_pc: 913 // Takes a single unsigned LEB128 operand, multiplies it by the 914 // min_inst_length field of the prologue, and adds the 915 // result to the address register of the state machine. 916 { 917 uint64_t AddrOffset = State.advanceAddr( 918 DebugLineData.getULEB128(OffsetPtr), Opcode, OpcodeOffset); 919 if (OS) 920 *OS << " (" << AddrOffset << ")"; 921 } 922 break; 923 924 case DW_LNS_advance_line: 925 // Takes a single signed LEB128 operand and adds that value to 926 // the line register of the state machine. 927 State.Row.Line += DebugLineData.getSLEB128(OffsetPtr); 928 if (OS) 929 *OS << " (" << State.Row.Line << ")"; 930 break; 931 932 case DW_LNS_set_file: 933 // Takes a single unsigned LEB128 operand and stores it in the file 934 // register of the state machine. 935 State.Row.File = DebugLineData.getULEB128(OffsetPtr); 936 if (OS) 937 *OS << " (" << State.Row.File << ")"; 938 break; 939 940 case DW_LNS_set_column: 941 // Takes a single unsigned LEB128 operand and stores it in the 942 // column register of the state machine. 943 State.Row.Column = DebugLineData.getULEB128(OffsetPtr); 944 if (OS) 945 *OS << " (" << State.Row.Column << ")"; 946 break; 947 948 case DW_LNS_negate_stmt: 949 // Takes no arguments. Set the is_stmt register of the state 950 // machine to the logical negation of its current value. 951 State.Row.IsStmt = !State.Row.IsStmt; 952 break; 953 954 case DW_LNS_set_basic_block: 955 // Takes no arguments. Set the basic_block register of the 956 // state machine to true 957 State.Row.BasicBlock = true; 958 break; 959 960 case DW_LNS_const_add_pc: 961 // Takes no arguments. Add to the address register of the state 962 // machine the address increment value corresponding to special 963 // opcode 255. The motivation for DW_LNS_const_add_pc is this: 964 // when the statement program needs to advance the address by a 965 // small amount, it can use a single special opcode, which occupies 966 // a single byte. When it needs to advance the address by up to 967 // twice the range of the last special opcode, it can use 968 // DW_LNS_const_add_pc followed by a special opcode, for a total 969 // of two bytes. Only if it needs to advance the address by more 970 // than twice that range will it need to use both DW_LNS_advance_pc 971 // and a special opcode, requiring three or more bytes. 972 { 973 uint64_t AddrOffset = 974 State.advanceAddrForOpcode(Opcode, OpcodeOffset).AddrDelta; 975 if (OS) 976 *OS << format(" (0x%16.16" PRIx64 ")", AddrOffset); 977 } 978 break; 979 980 case DW_LNS_fixed_advance_pc: 981 // Takes a single uhalf operand. Add to the address register of 982 // the state machine the value of the (unencoded) operand. This 983 // is the only extended opcode that takes an argument that is not 984 // a variable length number. The motivation for DW_LNS_fixed_advance_pc 985 // is this: existing assemblers cannot emit DW_LNS_advance_pc or 986 // special opcodes because they cannot encode LEB128 numbers or 987 // judge when the computation of a special opcode overflows and 988 // requires the use of DW_LNS_advance_pc. Such assemblers, however, 989 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression. 990 { 991 uint16_t PCOffset = DebugLineData.getRelocatedValue(2, OffsetPtr); 992 State.Row.Address.Address += PCOffset; 993 if (OS) 994 *OS 995 << format(" (0x%4.4" PRIx16 ")", PCOffset); 996 } 997 break; 998 999 case DW_LNS_set_prologue_end: 1000 // Takes no arguments. Set the prologue_end register of the 1001 // state machine to true 1002 State.Row.PrologueEnd = true; 1003 break; 1004 1005 case DW_LNS_set_epilogue_begin: 1006 // Takes no arguments. Set the basic_block register of the 1007 // state machine to true 1008 State.Row.EpilogueBegin = true; 1009 break; 1010 1011 case DW_LNS_set_isa: 1012 // Takes a single unsigned LEB128 operand and stores it in the 1013 // column register of the state machine. 1014 State.Row.Isa = DebugLineData.getULEB128(OffsetPtr); 1015 if (OS) 1016 *OS << " (" << (uint64_t)State.Row.Isa << ")"; 1017 break; 1018 1019 default: 1020 // Handle any unknown standard opcodes here. We know the lengths 1021 // of such opcodes because they are specified in the prologue 1022 // as a multiple of LEB128 operands for each opcode. 1023 { 1024 assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size()); 1025 uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1]; 1026 for (uint8_t I = 0; I < OpcodeLength; ++I) { 1027 uint64_t Value = DebugLineData.getULEB128(OffsetPtr); 1028 if (OS) 1029 *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n", 1030 Value); 1031 } 1032 } 1033 break; 1034 } 1035 } else { 1036 // Special Opcodes. 1037 ParsingState::AddrAndLineDelta Delta = 1038 State.handleSpecialOpcode(Opcode, OpcodeOffset); 1039 1040 if (OS) { 1041 *OS << "address += " << Delta.Address << ", line += " << Delta.Line 1042 << "\n"; 1043 OS->indent(12); 1044 State.Row.dump(*OS); 1045 } 1046 1047 State.appendRowToMatrix(); 1048 } 1049 if(OS) 1050 *OS << "\n"; 1051 } 1052 1053 if (!State.Sequence.Empty) 1054 RecoverableErrorHandler(createStringError( 1055 errc::illegal_byte_sequence, 1056 "last sequence in debug line table at offset 0x%8.8" PRIx64 1057 " is not terminated", 1058 DebugLineOffset)); 1059 1060 // Sort all sequences so that address lookup will work faster. 1061 if (!Sequences.empty()) { 1062 llvm::sort(Sequences, Sequence::orderByHighPC); 1063 // Note: actually, instruction address ranges of sequences should not 1064 // overlap (in shared objects and executables). If they do, the address 1065 // lookup would still work, though, but result would be ambiguous. 1066 // We don't report warning in this case. For example, 1067 // sometimes .so compiled from multiple object files contains a few 1068 // rudimentary sequences for address ranges [0x0, 0xsomething). 1069 } 1070 1071 return Error::success(); 1072 } 1073 1074 uint32_t DWARFDebugLine::LineTable::findRowInSeq( 1075 const DWARFDebugLine::Sequence &Seq, 1076 object::SectionedAddress Address) const { 1077 if (!Seq.containsPC(Address)) 1078 return UnknownRowIndex; 1079 assert(Seq.SectionIndex == Address.SectionIndex); 1080 // In some cases, e.g. first instruction in a function, the compiler generates 1081 // two entries, both with the same address. We want the last one. 1082 // 1083 // In general we want a non-empty range: the last row whose address is less 1084 // than or equal to Address. This can be computed as upper_bound - 1. 1085 DWARFDebugLine::Row Row; 1086 Row.Address = Address; 1087 RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex; 1088 RowIter LastRow = Rows.begin() + Seq.LastRowIndex; 1089 assert(FirstRow->Address.Address <= Row.Address.Address && 1090 Row.Address.Address < LastRow[-1].Address.Address); 1091 RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row, 1092 DWARFDebugLine::Row::orderByAddress) - 1093 1; 1094 assert(Seq.SectionIndex == RowPos->Address.SectionIndex); 1095 return RowPos - Rows.begin(); 1096 } 1097 1098 uint32_t DWARFDebugLine::LineTable::lookupAddress( 1099 object::SectionedAddress Address) const { 1100 1101 // Search for relocatable addresses 1102 uint32_t Result = lookupAddressImpl(Address); 1103 1104 if (Result != UnknownRowIndex || 1105 Address.SectionIndex == object::SectionedAddress::UndefSection) 1106 return Result; 1107 1108 // Search for absolute addresses 1109 Address.SectionIndex = object::SectionedAddress::UndefSection; 1110 return lookupAddressImpl(Address); 1111 } 1112 1113 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl( 1114 object::SectionedAddress Address) const { 1115 // First, find an instruction sequence containing the given address. 1116 DWARFDebugLine::Sequence Sequence; 1117 Sequence.SectionIndex = Address.SectionIndex; 1118 Sequence.HighPC = Address.Address; 1119 SequenceIter It = llvm::upper_bound(Sequences, Sequence, 1120 DWARFDebugLine::Sequence::orderByHighPC); 1121 if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex) 1122 return UnknownRowIndex; 1123 return findRowInSeq(*It, Address); 1124 } 1125 1126 bool DWARFDebugLine::LineTable::lookupAddressRange( 1127 object::SectionedAddress Address, uint64_t Size, 1128 std::vector<uint32_t> &Result) const { 1129 1130 // Search for relocatable addresses 1131 if (lookupAddressRangeImpl(Address, Size, Result)) 1132 return true; 1133 1134 if (Address.SectionIndex == object::SectionedAddress::UndefSection) 1135 return false; 1136 1137 // Search for absolute addresses 1138 Address.SectionIndex = object::SectionedAddress::UndefSection; 1139 return lookupAddressRangeImpl(Address, Size, Result); 1140 } 1141 1142 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl( 1143 object::SectionedAddress Address, uint64_t Size, 1144 std::vector<uint32_t> &Result) const { 1145 if (Sequences.empty()) 1146 return false; 1147 uint64_t EndAddr = Address.Address + Size; 1148 // First, find an instruction sequence containing the given address. 1149 DWARFDebugLine::Sequence Sequence; 1150 Sequence.SectionIndex = Address.SectionIndex; 1151 Sequence.HighPC = Address.Address; 1152 SequenceIter LastSeq = Sequences.end(); 1153 SequenceIter SeqPos = llvm::upper_bound( 1154 Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC); 1155 if (SeqPos == LastSeq || !SeqPos->containsPC(Address)) 1156 return false; 1157 1158 SequenceIter StartPos = SeqPos; 1159 1160 // Add the rows from the first sequence to the vector, starting with the 1161 // index we just calculated 1162 1163 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) { 1164 const DWARFDebugLine::Sequence &CurSeq = *SeqPos; 1165 // For the first sequence, we need to find which row in the sequence is the 1166 // first in our range. 1167 uint32_t FirstRowIndex = CurSeq.FirstRowIndex; 1168 if (SeqPos == StartPos) 1169 FirstRowIndex = findRowInSeq(CurSeq, Address); 1170 1171 // Figure out the last row in the range. 1172 uint32_t LastRowIndex = 1173 findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex}); 1174 if (LastRowIndex == UnknownRowIndex) 1175 LastRowIndex = CurSeq.LastRowIndex - 1; 1176 1177 assert(FirstRowIndex != UnknownRowIndex); 1178 assert(LastRowIndex != UnknownRowIndex); 1179 1180 for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) { 1181 Result.push_back(I); 1182 } 1183 1184 ++SeqPos; 1185 } 1186 1187 return true; 1188 } 1189 1190 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex, 1191 FileLineInfoKind Kind) const { 1192 if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex)) 1193 return None; 1194 const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex); 1195 if (Optional<const char *> source = Entry.Source.getAsCString()) 1196 return StringRef(*source); 1197 return None; 1198 } 1199 1200 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) { 1201 // Debug info can contain paths from any OS, not necessarily 1202 // an OS we're currently running on. Moreover different compilation units can 1203 // be compiled on different operating systems and linked together later. 1204 return sys::path::is_absolute(Path, sys::path::Style::posix) || 1205 sys::path::is_absolute(Path, sys::path::Style::windows); 1206 } 1207 1208 bool DWARFDebugLine::Prologue::getFileNameByIndex( 1209 uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind, 1210 std::string &Result, sys::path::Style Style) const { 1211 if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex)) 1212 return false; 1213 const FileNameEntry &Entry = getFileNameEntry(FileIndex); 1214 Optional<const char *> Name = Entry.Name.getAsCString(); 1215 if (!Name) 1216 return false; 1217 StringRef FileName = *Name; 1218 if (Kind == FileLineInfoKind::RawValue || 1219 isPathAbsoluteOnWindowsOrPosix(FileName)) { 1220 Result = std::string(FileName); 1221 return true; 1222 } 1223 if (Kind == FileLineInfoKind::BaseNameOnly) { 1224 Result = std::string(llvm::sys::path::filename(FileName)); 1225 return true; 1226 } 1227 1228 SmallString<16> FilePath; 1229 StringRef IncludeDir; 1230 // Be defensive about the contents of Entry. 1231 if (getVersion() >= 5) { 1232 if (Entry.DirIdx < IncludeDirectories.size()) 1233 IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue(); 1234 } else { 1235 if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size()) 1236 IncludeDir = 1237 IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue(); 1238 } 1239 // For absolute paths only, include the compilation directory of compile unit. 1240 // We know that FileName is not absolute, the only way to have an absolute 1241 // path at this point would be if IncludeDir is absolute. 1242 if (Kind == FileLineInfoKind::AbsoluteFilePath && !CompDir.empty() && 1243 !isPathAbsoluteOnWindowsOrPosix(IncludeDir)) 1244 sys::path::append(FilePath, Style, CompDir); 1245 1246 assert((Kind == FileLineInfoKind::AbsoluteFilePath || 1247 Kind == FileLineInfoKind::RelativeFilePath) && 1248 "invalid FileLineInfo Kind"); 1249 1250 // sys::path::append skips empty strings. 1251 sys::path::append(FilePath, Style, IncludeDir, FileName); 1252 Result = std::string(FilePath.str()); 1253 return true; 1254 } 1255 1256 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress( 1257 object::SectionedAddress Address, const char *CompDir, 1258 FileLineInfoKind Kind, DILineInfo &Result) const { 1259 // Get the index of row we're looking for in the line table. 1260 uint32_t RowIndex = lookupAddress(Address); 1261 if (RowIndex == -1U) 1262 return false; 1263 // Take file number and line/column from the row. 1264 const auto &Row = Rows[RowIndex]; 1265 if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName)) 1266 return false; 1267 Result.Line = Row.Line; 1268 Result.Column = Row.Column; 1269 Result.Discriminator = Row.Discriminator; 1270 Result.Source = getSourceByIndex(Row.File, Kind); 1271 return true; 1272 } 1273 1274 // We want to supply the Unit associated with a .debug_line[.dwo] table when 1275 // we dump it, if possible, but still dump the table even if there isn't a Unit. 1276 // Therefore, collect up handles on all the Units that point into the 1277 // line-table section. 1278 static DWARFDebugLine::SectionParser::LineToUnitMap 1279 buildLineToUnitMap(DWARFDebugLine::SectionParser::cu_range CUs, 1280 DWARFDebugLine::SectionParser::tu_range TUs) { 1281 DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit; 1282 for (const auto &CU : CUs) 1283 if (auto CUDIE = CU->getUnitDIE()) 1284 if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list))) 1285 LineToUnit.insert(std::make_pair(*StmtOffset, &*CU)); 1286 for (const auto &TU : TUs) 1287 if (auto TUDIE = TU->getUnitDIE()) 1288 if (auto StmtOffset = toSectionOffset(TUDIE.find(DW_AT_stmt_list))) 1289 LineToUnit.insert(std::make_pair(*StmtOffset, &*TU)); 1290 return LineToUnit; 1291 } 1292 1293 DWARFDebugLine::SectionParser::SectionParser(DWARFDataExtractor &Data, 1294 const DWARFContext &C, 1295 cu_range CUs, tu_range TUs) 1296 : DebugLineData(Data), Context(C) { 1297 LineToUnit = buildLineToUnitMap(CUs, TUs); 1298 if (!DebugLineData.isValidOffset(Offset)) 1299 Done = true; 1300 } 1301 1302 bool DWARFDebugLine::Prologue::totalLengthIsValid() const { 1303 return TotalLength != 0u; 1304 } 1305 1306 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext( 1307 function_ref<void(Error)> RecoverableErrorHandler, 1308 function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS) { 1309 assert(DebugLineData.isValidOffset(Offset) && 1310 "parsing should have terminated"); 1311 DWARFUnit *U = prepareToParse(Offset); 1312 uint64_t OldOffset = Offset; 1313 LineTable LT; 1314 if (Error Err = LT.parse(DebugLineData, &Offset, Context, U, 1315 RecoverableErrorHandler, OS)) 1316 UnrecoverableErrorHandler(std::move(Err)); 1317 moveToNextTable(OldOffset, LT.Prologue); 1318 return LT; 1319 } 1320 1321 void DWARFDebugLine::SectionParser::skip( 1322 function_ref<void(Error)> RecoverableErrorHandler, 1323 function_ref<void(Error)> UnrecoverableErrorHandler) { 1324 assert(DebugLineData.isValidOffset(Offset) && 1325 "parsing should have terminated"); 1326 DWARFUnit *U = prepareToParse(Offset); 1327 uint64_t OldOffset = Offset; 1328 LineTable LT; 1329 if (Error Err = LT.Prologue.parse(DebugLineData, &Offset, 1330 RecoverableErrorHandler, Context, U)) 1331 UnrecoverableErrorHandler(std::move(Err)); 1332 moveToNextTable(OldOffset, LT.Prologue); 1333 } 1334 1335 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) { 1336 DWARFUnit *U = nullptr; 1337 auto It = LineToUnit.find(Offset); 1338 if (It != LineToUnit.end()) 1339 U = It->second; 1340 DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0); 1341 return U; 1342 } 1343 1344 void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset, 1345 const Prologue &P) { 1346 // If the length field is not valid, we don't know where the next table is, so 1347 // cannot continue to parse. Mark the parser as done, and leave the Offset 1348 // value as it currently is. This will be the end of the bad length field. 1349 if (!P.totalLengthIsValid()) { 1350 Done = true; 1351 return; 1352 } 1353 1354 Offset = OldOffset + P.TotalLength + P.sizeofTotalLength(); 1355 if (!DebugLineData.isValidOffset(Offset)) { 1356 Done = true; 1357 } 1358 } 1359