1 //===- DWARFDebugLine.cpp -------------------------------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 10 #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h" 11 #include "llvm/ADT/SmallString.h" 12 #include "llvm/ADT/SmallVector.h" 13 #include "llvm/ADT/StringRef.h" 14 #include "llvm/BinaryFormat/Dwarf.h" 15 #include "llvm/DebugInfo/DWARF/DWARFContext.h" 16 #include "llvm/DebugInfo/DWARF/DWARFFormValue.h" 17 #include "llvm/DebugInfo/DWARF/DWARFRelocMap.h" 18 #include "llvm/Support/Format.h" 19 #include "llvm/Support/Path.h" 20 #include "llvm/Support/raw_ostream.h" 21 #include <algorithm> 22 #include <cassert> 23 #include <cinttypes> 24 #include <cstdint> 25 #include <cstdio> 26 #include <utility> 27 28 using namespace llvm; 29 using namespace dwarf; 30 31 using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind; 32 33 namespace { 34 35 struct ContentDescriptor { 36 dwarf::LineNumberEntryFormat Type; 37 dwarf::Form Form; 38 }; 39 40 using ContentDescriptors = SmallVector<ContentDescriptor, 4>; 41 42 } // end anonmyous namespace 43 44 DWARFDebugLine::Prologue::Prologue() { clear(); } 45 46 void DWARFDebugLine::Prologue::clear() { 47 TotalLength = PrologueLength = 0; 48 SegSelectorSize = 0; 49 MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0; 50 OpcodeBase = 0; 51 FormParams = DWARFFormParams({0, 0, DWARF32}); 52 StandardOpcodeLengths.clear(); 53 IncludeDirectories.clear(); 54 FileNames.clear(); 55 } 56 57 void DWARFDebugLine::Prologue::dump(raw_ostream &OS) const { 58 OS << "Line table prologue:\n" 59 << format(" total_length: 0x%8.8" PRIx64 "\n", TotalLength) 60 << format(" version: %u\n", getVersion()); 61 if (getVersion() >= 5) 62 OS << format(" address_size: %u\n", getAddressSize()) 63 << format(" seg_select_size: %u\n", SegSelectorSize); 64 OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength) 65 << format(" min_inst_length: %u\n", MinInstLength) 66 << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst) 67 << format(" default_is_stmt: %u\n", DefaultIsStmt) 68 << format(" line_base: %i\n", LineBase) 69 << format(" line_range: %u\n", LineRange) 70 << format(" opcode_base: %u\n", OpcodeBase); 71 72 for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I) 73 OS << format("standard_opcode_lengths[%s] = %u\n", 74 LNStandardString(I + 1).data(), StandardOpcodeLengths[I]); 75 76 if (!IncludeDirectories.empty()) 77 for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) 78 OS << format("include_directories[%3u] = '", I + 1) 79 << IncludeDirectories[I] << "'\n"; 80 81 if (!FileNames.empty()) { 82 OS << " Dir Mod Time File Len File Name\n" 83 << " ---- ---------- ---------- -----------" 84 "----------------\n"; 85 for (uint32_t I = 0; I != FileNames.size(); ++I) { 86 const FileNameEntry &FileEntry = FileNames[I]; 87 OS << format("file_names[%3u] %4" PRIu64 " ", I + 1, FileEntry.DirIdx) 88 << format("0x%8.8" PRIx64 " 0x%8.8" PRIx64 " ", FileEntry.ModTime, 89 FileEntry.Length) 90 << FileEntry.Name << '\n'; 91 } 92 } 93 } 94 95 // Parse v2-v4 directory and file tables. 96 static void 97 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData, 98 uint32_t *OffsetPtr, uint64_t EndPrologueOffset, 99 std::vector<StringRef> &IncludeDirectories, 100 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { 101 while (*OffsetPtr < EndPrologueOffset) { 102 StringRef S = DebugLineData.getCStrRef(OffsetPtr); 103 if (S.empty()) 104 break; 105 IncludeDirectories.push_back(S); 106 } 107 108 while (*OffsetPtr < EndPrologueOffset) { 109 StringRef Name = DebugLineData.getCStrRef(OffsetPtr); 110 if (Name.empty()) 111 break; 112 DWARFDebugLine::FileNameEntry FileEntry; 113 FileEntry.Name = Name; 114 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr); 115 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr); 116 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr); 117 FileNames.push_back(FileEntry); 118 } 119 } 120 121 // Parse v5 directory/file entry content descriptions. 122 // Returns the descriptors, or an empty vector if we did not find a path or 123 // ran off the end of the prologue. 124 static ContentDescriptors 125 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint32_t *OffsetPtr, 126 uint64_t EndPrologueOffset) { 127 ContentDescriptors Descriptors; 128 int FormatCount = DebugLineData.getU8(OffsetPtr); 129 bool HasPath = false; 130 for (int I = 0; I != FormatCount; ++I) { 131 if (*OffsetPtr >= EndPrologueOffset) 132 return ContentDescriptors(); 133 ContentDescriptor Descriptor; 134 Descriptor.Type = 135 dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr)); 136 Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr)); 137 if (Descriptor.Type == dwarf::DW_LNCT_path) 138 HasPath = true; 139 Descriptors.push_back(Descriptor); 140 } 141 return HasPath ? Descriptors : ContentDescriptors(); 142 } 143 144 static bool 145 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData, 146 uint32_t *OffsetPtr, uint64_t EndPrologueOffset, 147 const DWARFFormParams &FormParams, const DWARFUnit *U, 148 std::vector<StringRef> &IncludeDirectories, 149 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { 150 // Get the directory entry description. 151 ContentDescriptors DirDescriptors = 152 parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset); 153 if (DirDescriptors.empty()) 154 return false; 155 156 // Get the directory entries, according to the format described above. 157 int DirEntryCount = DebugLineData.getU8(OffsetPtr); 158 for (int I = 0; I != DirEntryCount; ++I) { 159 if (*OffsetPtr >= EndPrologueOffset) 160 return false; 161 for (auto Descriptor : DirDescriptors) { 162 DWARFFormValue Value(Descriptor.Form); 163 switch (Descriptor.Type) { 164 case DW_LNCT_path: 165 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, U)) 166 return false; 167 IncludeDirectories.push_back(Value.getAsCString().getValue()); 168 break; 169 default: 170 if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams)) 171 return false; 172 } 173 } 174 } 175 176 // Get the file entry description. 177 ContentDescriptors FileDescriptors = 178 parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset); 179 if (FileDescriptors.empty()) 180 return false; 181 182 // Get the file entries, according to the format described above. 183 int FileEntryCount = DebugLineData.getU8(OffsetPtr); 184 for (int I = 0; I != FileEntryCount; ++I) { 185 if (*OffsetPtr >= EndPrologueOffset) 186 return false; 187 DWARFDebugLine::FileNameEntry FileEntry; 188 for (auto Descriptor : FileDescriptors) { 189 DWARFFormValue Value(Descriptor.Form); 190 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, U)) 191 return false; 192 switch (Descriptor.Type) { 193 case DW_LNCT_path: 194 FileEntry.Name = Value.getAsCString().getValue(); 195 break; 196 case DW_LNCT_directory_index: 197 FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue(); 198 break; 199 case DW_LNCT_timestamp: 200 FileEntry.ModTime = Value.getAsUnsignedConstant().getValue(); 201 break; 202 case DW_LNCT_size: 203 FileEntry.Length = Value.getAsUnsignedConstant().getValue(); 204 break; 205 // FIXME: Add MD5 206 default: 207 break; 208 } 209 } 210 FileNames.push_back(FileEntry); 211 } 212 return true; 213 } 214 215 bool DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData, 216 uint32_t *OffsetPtr, const DWARFUnit *U) { 217 const uint64_t PrologueOffset = *OffsetPtr; 218 219 clear(); 220 TotalLength = DebugLineData.getU32(OffsetPtr); 221 if (TotalLength == UINT32_MAX) { 222 FormParams.Format = dwarf::DWARF64; 223 TotalLength = DebugLineData.getU64(OffsetPtr); 224 } else if (TotalLength >= 0xffffff00) { 225 return false; 226 } 227 FormParams.Version = DebugLineData.getU16(OffsetPtr); 228 if (getVersion() < 2) 229 return false; 230 231 if (getVersion() >= 5) { 232 FormParams.AddrSize = DebugLineData.getU8(OffsetPtr); 233 assert((DebugLineData.getAddressSize() == 0 || 234 DebugLineData.getAddressSize() == getAddressSize()) && 235 "Line table header and data extractor disagree"); 236 SegSelectorSize = DebugLineData.getU8(OffsetPtr); 237 } 238 239 PrologueLength = DebugLineData.getUnsigned(OffsetPtr, sizeofPrologueLength()); 240 const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr; 241 MinInstLength = DebugLineData.getU8(OffsetPtr); 242 if (getVersion() >= 4) 243 MaxOpsPerInst = DebugLineData.getU8(OffsetPtr); 244 DefaultIsStmt = DebugLineData.getU8(OffsetPtr); 245 LineBase = DebugLineData.getU8(OffsetPtr); 246 LineRange = DebugLineData.getU8(OffsetPtr); 247 OpcodeBase = DebugLineData.getU8(OffsetPtr); 248 249 StandardOpcodeLengths.reserve(OpcodeBase - 1); 250 for (uint32_t I = 1; I < OpcodeBase; ++I) { 251 uint8_t OpLen = DebugLineData.getU8(OffsetPtr); 252 StandardOpcodeLengths.push_back(OpLen); 253 } 254 255 if (getVersion() >= 5) { 256 if (!parseV5DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset, 257 getFormParams(), U, IncludeDirectories, 258 FileNames)) { 259 fprintf(stderr, 260 "warning: parsing line table prologue at 0x%8.8" PRIx64 261 " found an invalid directory or file table description at" 262 " 0x%8.8" PRIx64 "\n", PrologueOffset, (uint64_t)*OffsetPtr); 263 return false; 264 } 265 } else 266 parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset, 267 IncludeDirectories, FileNames); 268 269 if (*OffsetPtr != EndPrologueOffset) { 270 fprintf(stderr, 271 "warning: parsing line table prologue at 0x%8.8" PRIx64 272 " should have ended at 0x%8.8" PRIx64 273 " but it ended at 0x%8.8" PRIx64 "\n", 274 PrologueOffset, EndPrologueOffset, (uint64_t)*OffsetPtr); 275 return false; 276 } 277 return true; 278 } 279 280 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); } 281 282 void DWARFDebugLine::Row::postAppend() { 283 BasicBlock = false; 284 PrologueEnd = false; 285 EpilogueBegin = false; 286 } 287 288 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) { 289 Address = 0; 290 Line = 1; 291 Column = 0; 292 File = 1; 293 Isa = 0; 294 Discriminator = 0; 295 IsStmt = DefaultIsStmt; 296 BasicBlock = false; 297 EndSequence = false; 298 PrologueEnd = false; 299 EpilogueBegin = false; 300 } 301 302 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) { 303 OS << "Address Line Column File ISA Discriminator Flags\n" 304 << "------------------ ------ ------ ------ --- ------------- " 305 "-------------\n"; 306 } 307 308 void DWARFDebugLine::Row::dump(raw_ostream &OS) const { 309 OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column) 310 << format(" %6u %3u %13u ", File, Isa, Discriminator) 311 << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "") 312 << (PrologueEnd ? " prologue_end" : "") 313 << (EpilogueBegin ? " epilogue_begin" : "") 314 << (EndSequence ? " end_sequence" : "") << '\n'; 315 } 316 317 DWARFDebugLine::Sequence::Sequence() { reset(); } 318 319 void DWARFDebugLine::Sequence::reset() { 320 LowPC = 0; 321 HighPC = 0; 322 FirstRowIndex = 0; 323 LastRowIndex = 0; 324 Empty = true; 325 } 326 327 DWARFDebugLine::LineTable::LineTable() { clear(); } 328 329 void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const { 330 Prologue.dump(OS); 331 OS << '\n'; 332 333 if (!Rows.empty()) { 334 Row::dumpTableHeader(OS); 335 for (const Row &R : Rows) { 336 R.dump(OS); 337 } 338 } 339 } 340 341 void DWARFDebugLine::LineTable::clear() { 342 Prologue.clear(); 343 Rows.clear(); 344 Sequences.clear(); 345 } 346 347 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT) 348 : LineTable(LT) { 349 resetRowAndSequence(); 350 } 351 352 void DWARFDebugLine::ParsingState::resetRowAndSequence() { 353 Row.reset(LineTable->Prologue.DefaultIsStmt); 354 Sequence.reset(); 355 } 356 357 void DWARFDebugLine::ParsingState::appendRowToMatrix(uint32_t Offset) { 358 if (Sequence.Empty) { 359 // Record the beginning of instruction sequence. 360 Sequence.Empty = false; 361 Sequence.LowPC = Row.Address; 362 Sequence.FirstRowIndex = RowNumber; 363 } 364 ++RowNumber; 365 LineTable->appendRow(Row); 366 if (Row.EndSequence) { 367 // Record the end of instruction sequence. 368 Sequence.HighPC = Row.Address; 369 Sequence.LastRowIndex = RowNumber; 370 if (Sequence.isValid()) 371 LineTable->appendSequence(Sequence); 372 Sequence.reset(); 373 } 374 Row.postAppend(); 375 } 376 377 const DWARFDebugLine::LineTable * 378 DWARFDebugLine::getLineTable(uint32_t Offset) const { 379 LineTableConstIter Pos = LineTableMap.find(Offset); 380 if (Pos != LineTableMap.end()) 381 return &Pos->second; 382 return nullptr; 383 } 384 385 const DWARFDebugLine::LineTable * 386 DWARFDebugLine::getOrParseLineTable(DWARFDataExtractor &DebugLineData, 387 uint32_t Offset, const DWARFUnit *U) { 388 std::pair<LineTableIter, bool> Pos = 389 LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable())); 390 LineTable *LT = &Pos.first->second; 391 if (Pos.second) { 392 if (!LT->parse(DebugLineData, &Offset, U)) 393 return nullptr; 394 } 395 return LT; 396 } 397 398 bool DWARFDebugLine::LineTable::parse(DWARFDataExtractor &DebugLineData, 399 uint32_t *OffsetPtr, const DWARFUnit *U, 400 raw_ostream *OS) { 401 const uint32_t DebugLineOffset = *OffsetPtr; 402 403 clear(); 404 405 if (!Prologue.parse(DebugLineData, OffsetPtr, U)) { 406 // Restore our offset and return false to indicate failure! 407 *OffsetPtr = DebugLineOffset; 408 return false; 409 } 410 411 if (OS) 412 Prologue.dump(*OS); 413 414 const uint32_t EndOffset = 415 DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength(); 416 417 // See if we should tell the data extractor the address size. 418 if (DebugLineData.getAddressSize() == 0) 419 DebugLineData.setAddressSize(Prologue.getAddressSize()); 420 else 421 assert(Prologue.getAddressSize() == 0 || 422 Prologue.getAddressSize() == DebugLineData.getAddressSize()); 423 424 ParsingState State(this); 425 426 while (*OffsetPtr < EndOffset) { 427 if (OS) 428 *OS << format("0x%08.08" PRIx32 ": ", *OffsetPtr); 429 430 uint8_t Opcode = DebugLineData.getU8(OffsetPtr); 431 432 if (OS) 433 *OS << format("%02.02" PRIx8 " ", Opcode); 434 435 if (Opcode == 0) { 436 // Extended Opcodes always start with a zero opcode followed by 437 // a uleb128 length so you can skip ones you don't know about 438 uint64_t Len = DebugLineData.getULEB128(OffsetPtr); 439 uint32_t ExtOffset = *OffsetPtr; 440 441 // Tolerate zero-length; assume length is correct and soldier on. 442 if (Len == 0) { 443 if (OS) 444 *OS << "Badly formed extended line op (length 0)\n"; 445 continue; 446 } 447 448 uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr); 449 if (OS) 450 *OS << LNExtendedString(SubOpcode); 451 switch (SubOpcode) { 452 case DW_LNE_end_sequence: 453 // Set the end_sequence register of the state machine to true and 454 // append a row to the matrix using the current values of the 455 // state-machine registers. Then reset the registers to the initial 456 // values specified above. Every statement program sequence must end 457 // with a DW_LNE_end_sequence instruction which creates a row whose 458 // address is that of the byte after the last target machine instruction 459 // of the sequence. 460 State.Row.EndSequence = true; 461 State.appendRowToMatrix(*OffsetPtr); 462 if (OS) { 463 *OS << "\n"; 464 OS->indent(12); 465 State.Row.dump(*OS); 466 } 467 State.resetRowAndSequence(); 468 break; 469 470 case DW_LNE_set_address: 471 // Takes a single relocatable address as an operand. The size of the 472 // operand is the size appropriate to hold an address on the target 473 // machine. Set the address register to the value given by the 474 // relocatable address. All of the other statement program opcodes 475 // that affect the address register add a delta to it. This instruction 476 // stores a relocatable value into it instead. 477 // 478 // Make sure the extractor knows the address size. If not, infer it 479 // from the size of the operand. 480 if (DebugLineData.getAddressSize() == 0) 481 DebugLineData.setAddressSize(Len - 1); 482 else 483 assert(DebugLineData.getAddressSize() == Len - 1); 484 State.Row.Address = DebugLineData.getRelocatedAddress(OffsetPtr); 485 if (OS) 486 *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address); 487 break; 488 489 case DW_LNE_define_file: 490 // Takes 4 arguments. The first is a null terminated string containing 491 // a source file name. The second is an unsigned LEB128 number 492 // representing the directory index of the directory in which the file 493 // was found. The third is an unsigned LEB128 number representing the 494 // time of last modification of the file. The fourth is an unsigned 495 // LEB128 number representing the length in bytes of the file. The time 496 // and length fields may contain LEB128(0) if the information is not 497 // available. 498 // 499 // The directory index represents an entry in the include_directories 500 // section of the statement program prologue. The index is LEB128(0) 501 // if the file was found in the current directory of the compilation, 502 // LEB128(1) if it was found in the first directory in the 503 // include_directories section, and so on. The directory index is 504 // ignored for file names that represent full path names. 505 // 506 // The files are numbered, starting at 1, in the order in which they 507 // appear; the names in the prologue come before names defined by 508 // the DW_LNE_define_file instruction. These numbers are used in the 509 // the file register of the state machine. 510 { 511 FileNameEntry FileEntry; 512 FileEntry.Name = DebugLineData.getCStr(OffsetPtr); 513 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr); 514 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr); 515 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr); 516 Prologue.FileNames.push_back(FileEntry); 517 if (OS) 518 *OS << " (" << FileEntry.Name.str() 519 << ", dir=" << FileEntry.DirIdx << ", mod_time=" 520 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime) 521 << ", length=" << FileEntry.Length << ")"; 522 } 523 break; 524 525 case DW_LNE_set_discriminator: 526 State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr); 527 if (OS) 528 *OS << " (" << State.Row.Discriminator << ")"; 529 break; 530 531 default: 532 if (OS) 533 *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode) 534 << format(" length %" PRIx64, Len); 535 // Len doesn't include the zero opcode byte or the length itself, but 536 // it does include the sub_opcode, so we have to adjust for that. 537 (*OffsetPtr) += Len - 1; 538 break; 539 } 540 // Make sure the stated and parsed lengths are the same. 541 // Otherwise we have an unparseable line-number program. 542 if (*OffsetPtr - ExtOffset != Len) { 543 fprintf(stderr, "Unexpected line op length at offset 0x%8.8" PRIx32 544 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx32 "\n", 545 ExtOffset, Len, *OffsetPtr - ExtOffset); 546 // Skip the rest of the line-number program. 547 *OffsetPtr = EndOffset; 548 return false; 549 } 550 } else if (Opcode < Prologue.OpcodeBase) { 551 if (OS) 552 *OS << LNStandardString(Opcode); 553 switch (Opcode) { 554 // Standard Opcodes 555 case DW_LNS_copy: 556 // Takes no arguments. Append a row to the matrix using the 557 // current values of the state-machine registers. Then set 558 // the basic_block register to false. 559 State.appendRowToMatrix(*OffsetPtr); 560 if (OS) { 561 *OS << "\n"; 562 OS->indent(12); 563 State.Row.dump(*OS); 564 *OS << "\n"; 565 } 566 break; 567 568 case DW_LNS_advance_pc: 569 // Takes a single unsigned LEB128 operand, multiplies it by the 570 // min_inst_length field of the prologue, and adds the 571 // result to the address register of the state machine. 572 { 573 uint64_t AddrOffset = 574 DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength; 575 State.Row.Address += AddrOffset; 576 if (OS) 577 *OS << " (" << AddrOffset << ")"; 578 } 579 break; 580 581 case DW_LNS_advance_line: 582 // Takes a single signed LEB128 operand and adds that value to 583 // the line register of the state machine. 584 State.Row.Line += DebugLineData.getSLEB128(OffsetPtr); 585 if (OS) 586 *OS << " (" << State.Row.Line << ")"; 587 break; 588 589 case DW_LNS_set_file: 590 // Takes a single unsigned LEB128 operand and stores it in the file 591 // register of the state machine. 592 State.Row.File = DebugLineData.getULEB128(OffsetPtr); 593 if (OS) 594 *OS << " (" << State.Row.File << ")"; 595 break; 596 597 case DW_LNS_set_column: 598 // Takes a single unsigned LEB128 operand and stores it in the 599 // column register of the state machine. 600 State.Row.Column = DebugLineData.getULEB128(OffsetPtr); 601 if (OS) 602 *OS << " (" << State.Row.Column << ")"; 603 break; 604 605 case DW_LNS_negate_stmt: 606 // Takes no arguments. Set the is_stmt register of the state 607 // machine to the logical negation of its current value. 608 State.Row.IsStmt = !State.Row.IsStmt; 609 break; 610 611 case DW_LNS_set_basic_block: 612 // Takes no arguments. Set the basic_block register of the 613 // state machine to true 614 State.Row.BasicBlock = true; 615 break; 616 617 case DW_LNS_const_add_pc: 618 // Takes no arguments. Add to the address register of the state 619 // machine the address increment value corresponding to special 620 // opcode 255. The motivation for DW_LNS_const_add_pc is this: 621 // when the statement program needs to advance the address by a 622 // small amount, it can use a single special opcode, which occupies 623 // a single byte. When it needs to advance the address by up to 624 // twice the range of the last special opcode, it can use 625 // DW_LNS_const_add_pc followed by a special opcode, for a total 626 // of two bytes. Only if it needs to advance the address by more 627 // than twice that range will it need to use both DW_LNS_advance_pc 628 // and a special opcode, requiring three or more bytes. 629 { 630 uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase; 631 uint64_t AddrOffset = 632 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength; 633 State.Row.Address += AddrOffset; 634 if (OS) 635 *OS 636 << format(" (0x%16.16" PRIx64 ")", AddrOffset); 637 } 638 break; 639 640 case DW_LNS_fixed_advance_pc: 641 // Takes a single uhalf operand. Add to the address register of 642 // the state machine the value of the (unencoded) operand. This 643 // is the only extended opcode that takes an argument that is not 644 // a variable length number. The motivation for DW_LNS_fixed_advance_pc 645 // is this: existing assemblers cannot emit DW_LNS_advance_pc or 646 // special opcodes because they cannot encode LEB128 numbers or 647 // judge when the computation of a special opcode overflows and 648 // requires the use of DW_LNS_advance_pc. Such assemblers, however, 649 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression. 650 { 651 uint16_t PCOffset = DebugLineData.getU16(OffsetPtr); 652 State.Row.Address += PCOffset; 653 if (OS) 654 *OS 655 << format(" (0x%16.16" PRIx64 ")", PCOffset); 656 } 657 break; 658 659 case DW_LNS_set_prologue_end: 660 // Takes no arguments. Set the prologue_end register of the 661 // state machine to true 662 State.Row.PrologueEnd = true; 663 break; 664 665 case DW_LNS_set_epilogue_begin: 666 // Takes no arguments. Set the basic_block register of the 667 // state machine to true 668 State.Row.EpilogueBegin = true; 669 break; 670 671 case DW_LNS_set_isa: 672 // Takes a single unsigned LEB128 operand and stores it in the 673 // column register of the state machine. 674 State.Row.Isa = DebugLineData.getULEB128(OffsetPtr); 675 if (OS) 676 *OS << " (" << State.Row.Isa << ")"; 677 break; 678 679 default: 680 // Handle any unknown standard opcodes here. We know the lengths 681 // of such opcodes because they are specified in the prologue 682 // as a multiple of LEB128 operands for each opcode. 683 { 684 assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size()); 685 uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1]; 686 for (uint8_t I = 0; I < OpcodeLength; ++I) { 687 uint64_t Value = DebugLineData.getULEB128(OffsetPtr); 688 if (OS) 689 *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n", 690 Value); 691 } 692 } 693 break; 694 } 695 } else { 696 // Special Opcodes 697 698 // A special opcode value is chosen based on the amount that needs 699 // to be added to the line and address registers. The maximum line 700 // increment for a special opcode is the value of the line_base 701 // field in the header, plus the value of the line_range field, 702 // minus 1 (line base + line range - 1). If the desired line 703 // increment is greater than the maximum line increment, a standard 704 // opcode must be used instead of a special opcode. The "address 705 // advance" is calculated by dividing the desired address increment 706 // by the minimum_instruction_length field from the header. The 707 // special opcode is then calculated using the following formula: 708 // 709 // opcode = (desired line increment - line_base) + 710 // (line_range * address advance) + opcode_base 711 // 712 // If the resulting opcode is greater than 255, a standard opcode 713 // must be used instead. 714 // 715 // To decode a special opcode, subtract the opcode_base from the 716 // opcode itself to give the adjusted opcode. The amount to 717 // increment the address register is the result of the adjusted 718 // opcode divided by the line_range multiplied by the 719 // minimum_instruction_length field from the header. That is: 720 // 721 // address increment = (adjusted opcode / line_range) * 722 // minimum_instruction_length 723 // 724 // The amount to increment the line register is the line_base plus 725 // the result of the adjusted opcode modulo the line_range. That is: 726 // 727 // line increment = line_base + (adjusted opcode % line_range) 728 729 uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase; 730 uint64_t AddrOffset = 731 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength; 732 int32_t LineOffset = 733 Prologue.LineBase + (AdjustOpcode % Prologue.LineRange); 734 State.Row.Line += LineOffset; 735 State.Row.Address += AddrOffset; 736 737 if (OS) { 738 *OS << "address += " << ((uint32_t)AdjustOpcode) 739 << ", line += " << LineOffset << "\n"; 740 OS->indent(12); 741 State.Row.dump(*OS); 742 } 743 744 State.appendRowToMatrix(*OffsetPtr); 745 // Reset discriminator to 0. 746 State.Row.Discriminator = 0; 747 } 748 if(OS) 749 *OS << "\n"; 750 } 751 752 if (!State.Sequence.Empty) { 753 fprintf(stderr, "warning: last sequence in debug line table is not" 754 "terminated!\n"); 755 } 756 757 // Sort all sequences so that address lookup will work faster. 758 if (!Sequences.empty()) { 759 std::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC); 760 // Note: actually, instruction address ranges of sequences should not 761 // overlap (in shared objects and executables). If they do, the address 762 // lookup would still work, though, but result would be ambiguous. 763 // We don't report warning in this case. For example, 764 // sometimes .so compiled from multiple object files contains a few 765 // rudimentary sequences for address ranges [0x0, 0xsomething). 766 } 767 768 return EndOffset; 769 } 770 771 uint32_t 772 DWARFDebugLine::LineTable::findRowInSeq(const DWARFDebugLine::Sequence &Seq, 773 uint64_t Address) const { 774 if (!Seq.containsPC(Address)) 775 return UnknownRowIndex; 776 // Search for instruction address in the rows describing the sequence. 777 // Rows are stored in a vector, so we may use arithmetical operations with 778 // iterators. 779 DWARFDebugLine::Row Row; 780 Row.Address = Address; 781 RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex; 782 RowIter LastRow = Rows.begin() + Seq.LastRowIndex; 783 LineTable::RowIter RowPos = std::lower_bound( 784 FirstRow, LastRow, Row, DWARFDebugLine::Row::orderByAddress); 785 if (RowPos == LastRow) { 786 return Seq.LastRowIndex - 1; 787 } 788 uint32_t Index = Seq.FirstRowIndex + (RowPos - FirstRow); 789 if (RowPos->Address > Address) { 790 if (RowPos == FirstRow) 791 return UnknownRowIndex; 792 else 793 Index--; 794 } 795 return Index; 796 } 797 798 uint32_t DWARFDebugLine::LineTable::lookupAddress(uint64_t Address) const { 799 if (Sequences.empty()) 800 return UnknownRowIndex; 801 // First, find an instruction sequence containing the given address. 802 DWARFDebugLine::Sequence Sequence; 803 Sequence.LowPC = Address; 804 SequenceIter FirstSeq = Sequences.begin(); 805 SequenceIter LastSeq = Sequences.end(); 806 SequenceIter SeqPos = std::lower_bound( 807 FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC); 808 DWARFDebugLine::Sequence FoundSeq; 809 if (SeqPos == LastSeq) { 810 FoundSeq = Sequences.back(); 811 } else if (SeqPos->LowPC == Address) { 812 FoundSeq = *SeqPos; 813 } else { 814 if (SeqPos == FirstSeq) 815 return UnknownRowIndex; 816 FoundSeq = *(SeqPos - 1); 817 } 818 return findRowInSeq(FoundSeq, Address); 819 } 820 821 bool DWARFDebugLine::LineTable::lookupAddressRange( 822 uint64_t Address, uint64_t Size, std::vector<uint32_t> &Result) const { 823 if (Sequences.empty()) 824 return false; 825 uint64_t EndAddr = Address + Size; 826 // First, find an instruction sequence containing the given address. 827 DWARFDebugLine::Sequence Sequence; 828 Sequence.LowPC = Address; 829 SequenceIter FirstSeq = Sequences.begin(); 830 SequenceIter LastSeq = Sequences.end(); 831 SequenceIter SeqPos = std::lower_bound( 832 FirstSeq, LastSeq, Sequence, DWARFDebugLine::Sequence::orderByLowPC); 833 if (SeqPos == LastSeq || SeqPos->LowPC != Address) { 834 if (SeqPos == FirstSeq) 835 return false; 836 SeqPos--; 837 } 838 if (!SeqPos->containsPC(Address)) 839 return false; 840 841 SequenceIter StartPos = SeqPos; 842 843 // Add the rows from the first sequence to the vector, starting with the 844 // index we just calculated 845 846 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) { 847 const DWARFDebugLine::Sequence &CurSeq = *SeqPos; 848 // For the first sequence, we need to find which row in the sequence is the 849 // first in our range. 850 uint32_t FirstRowIndex = CurSeq.FirstRowIndex; 851 if (SeqPos == StartPos) 852 FirstRowIndex = findRowInSeq(CurSeq, Address); 853 854 // Figure out the last row in the range. 855 uint32_t LastRowIndex = findRowInSeq(CurSeq, EndAddr - 1); 856 if (LastRowIndex == UnknownRowIndex) 857 LastRowIndex = CurSeq.LastRowIndex - 1; 858 859 assert(FirstRowIndex != UnknownRowIndex); 860 assert(LastRowIndex != UnknownRowIndex); 861 862 for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) { 863 Result.push_back(I); 864 } 865 866 ++SeqPos; 867 } 868 869 return true; 870 } 871 872 bool DWARFDebugLine::LineTable::hasFileAtIndex(uint64_t FileIndex) const { 873 return FileIndex != 0 && FileIndex <= Prologue.FileNames.size(); 874 } 875 876 bool DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex, 877 const char *CompDir, 878 FileLineInfoKind Kind, 879 std::string &Result) const { 880 if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex)) 881 return false; 882 const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1]; 883 StringRef FileName = Entry.Name; 884 if (Kind != FileLineInfoKind::AbsoluteFilePath || 885 sys::path::is_absolute(FileName)) { 886 Result = FileName; 887 return true; 888 } 889 890 SmallString<16> FilePath; 891 uint64_t IncludeDirIndex = Entry.DirIdx; 892 StringRef IncludeDir; 893 // Be defensive about the contents of Entry. 894 if (IncludeDirIndex > 0 && 895 IncludeDirIndex <= Prologue.IncludeDirectories.size()) 896 IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1]; 897 898 // We may still need to append compilation directory of compile unit. 899 // We know that FileName is not absolute, the only way to have an 900 // absolute path at this point would be if IncludeDir is absolute. 901 if (CompDir && Kind == FileLineInfoKind::AbsoluteFilePath && 902 sys::path::is_relative(IncludeDir)) 903 sys::path::append(FilePath, CompDir); 904 905 // sys::path::append skips empty strings. 906 sys::path::append(FilePath, IncludeDir, FileName); 907 Result = FilePath.str(); 908 return true; 909 } 910 911 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress( 912 uint64_t Address, const char *CompDir, FileLineInfoKind Kind, 913 DILineInfo &Result) const { 914 // Get the index of row we're looking for in the line table. 915 uint32_t RowIndex = lookupAddress(Address); 916 if (RowIndex == -1U) 917 return false; 918 // Take file number and line/column from the row. 919 const auto &Row = Rows[RowIndex]; 920 if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName)) 921 return false; 922 Result.Line = Row.Line; 923 Result.Column = Row.Column; 924 Result.Discriminator = Row.Discriminator; 925 return true; 926 } 927