1 //===- DWARFDebugLine.cpp -------------------------------------------------===//
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
7 //===----------------------------------------------------------------------===//
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/Path.h"
20 #include "llvm/Support/WithColor.h"
21 #include "llvm/Support/raw_ostream.h"
30 using namespace dwarf;
32 using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind;
36 struct ContentDescriptor {
37 dwarf::LineNumberEntryFormat Type;
41 using ContentDescriptors = SmallVector<ContentDescriptor, 4>;
43 } // end anonmyous namespace
45 void DWARFDebugLine::ContentTypeTracker::trackContentType(
46 dwarf::LineNumberEntryFormat ContentType) {
47 switch (ContentType) {
48 case dwarf::DW_LNCT_timestamp:
51 case dwarf::DW_LNCT_size:
54 case dwarf::DW_LNCT_MD5:
57 case dwarf::DW_LNCT_LLVM_source:
61 // We only care about values we consider optional, and new values may be
62 // added in the vendor extension range, so we do not match exhaustively.
67 DWARFDebugLine::Prologue::Prologue() { clear(); }
69 bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const {
70 uint16_t DwarfVersion = getVersion();
71 assert(DwarfVersion != 0 &&
72 "line table prologue has no dwarf version information");
73 if (DwarfVersion >= 5)
74 return FileIndex < FileNames.size();
75 return FileIndex != 0 && FileIndex <= FileNames.size();
78 const llvm::DWARFDebugLine::FileNameEntry &
79 DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const {
80 uint16_t DwarfVersion = getVersion();
81 assert(DwarfVersion != 0 &&
82 "line table prologue has no dwarf version information");
83 // In DWARF v5 the file names are 0-indexed.
84 if (DwarfVersion >= 5)
85 return FileNames[Index];
86 return FileNames[Index - 1];
89 void DWARFDebugLine::Prologue::clear() {
90 TotalLength = PrologueLength = 0;
92 MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
94 FormParams = dwarf::FormParams({0, 0, DWARF32});
95 ContentTypes = ContentTypeTracker();
96 StandardOpcodeLengths.clear();
97 IncludeDirectories.clear();
101 void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
102 DIDumpOptions DumpOptions) const {
103 OS << "Line table prologue:\n"
104 << format(" total_length: 0x%8.8" PRIx64 "\n", TotalLength)
105 << format(" version: %u\n", getVersion());
106 if (getVersion() >= 5)
107 OS << format(" address_size: %u\n", getAddressSize())
108 << format(" seg_select_size: %u\n", SegSelectorSize);
109 OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength)
110 << format(" min_inst_length: %u\n", MinInstLength)
111 << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
112 << format(" default_is_stmt: %u\n", DefaultIsStmt)
113 << format(" line_base: %i\n", LineBase)
114 << format(" line_range: %u\n", LineRange)
115 << format(" opcode_base: %u\n", OpcodeBase);
117 for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
118 OS << format("standard_opcode_lengths[%s] = %u\n",
119 LNStandardString(I + 1).data(), StandardOpcodeLengths[I]);
121 if (!IncludeDirectories.empty()) {
122 // DWARF v5 starts directory indexes at 0.
123 uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
124 for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
125 OS << format("include_directories[%3u] = ", I + DirBase);
126 IncludeDirectories[I].dump(OS, DumpOptions);
131 if (!FileNames.empty()) {
132 // DWARF v5 starts file indexes at 0.
133 uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
134 for (uint32_t I = 0; I != FileNames.size(); ++I) {
135 const FileNameEntry &FileEntry = FileNames[I];
136 OS << format("file_names[%3u]:\n", I + FileBase);
138 FileEntry.Name.dump(OS, DumpOptions);
140 << format(" dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
141 if (ContentTypes.HasMD5)
142 OS << " md5_checksum: " << FileEntry.Checksum.digest() << '\n';
143 if (ContentTypes.HasModTime)
144 OS << format(" mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
145 if (ContentTypes.HasLength)
146 OS << format(" length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
147 if (ContentTypes.HasSource) {
149 FileEntry.Source.dump(OS, DumpOptions);
156 // Parse v2-v4 directory and file tables.
158 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
159 uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
160 DWARFDebugLine::ContentTypeTracker &ContentTypes,
161 std::vector<DWARFFormValue> &IncludeDirectories,
162 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
163 while (*OffsetPtr < EndPrologueOffset) {
164 StringRef S = DebugLineData.getCStrRef(OffsetPtr);
168 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, S.data());
169 IncludeDirectories.push_back(Dir);
172 while (*OffsetPtr < EndPrologueOffset) {
173 StringRef Name = DebugLineData.getCStrRef(OffsetPtr);
176 DWARFDebugLine::FileNameEntry FileEntry;
178 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name.data());
179 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
180 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
181 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
182 FileNames.push_back(FileEntry);
185 ContentTypes.HasModTime = true;
186 ContentTypes.HasLength = true;
189 // Parse v5 directory/file entry content descriptions.
190 // Returns the descriptors, or an empty vector if we did not find a path or
191 // ran off the end of the prologue.
192 static ContentDescriptors
193 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint32_t
194 *OffsetPtr, uint64_t EndPrologueOffset, DWARFDebugLine::ContentTypeTracker
196 ContentDescriptors Descriptors;
197 int FormatCount = DebugLineData.getU8(OffsetPtr);
198 bool HasPath = false;
199 for (int I = 0; I != FormatCount; ++I) {
200 if (*OffsetPtr >= EndPrologueOffset)
201 return ContentDescriptors();
202 ContentDescriptor Descriptor;
204 dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr));
205 Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr));
206 if (Descriptor.Type == dwarf::DW_LNCT_path)
209 ContentTypes->trackContentType(Descriptor.Type);
210 Descriptors.push_back(Descriptor);
212 return HasPath ? Descriptors : ContentDescriptors();
216 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
217 uint32_t *OffsetPtr, uint64_t EndPrologueOffset,
218 const dwarf::FormParams &FormParams,
219 const DWARFContext &Ctx, const DWARFUnit *U,
220 DWARFDebugLine::ContentTypeTracker &ContentTypes,
221 std::vector<DWARFFormValue> &IncludeDirectories,
222 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
223 // Get the directory entry description.
224 ContentDescriptors DirDescriptors =
225 parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset, nullptr);
226 if (DirDescriptors.empty())
229 // Get the directory entries, according to the format described above.
230 int DirEntryCount = DebugLineData.getU8(OffsetPtr);
231 for (int I = 0; I != DirEntryCount; ++I) {
232 if (*OffsetPtr >= EndPrologueOffset)
234 for (auto Descriptor : DirDescriptors) {
235 DWARFFormValue Value(Descriptor.Form);
236 switch (Descriptor.Type) {
238 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
240 IncludeDirectories.push_back(Value);
243 if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
249 // Get the file entry description.
250 ContentDescriptors FileDescriptors =
251 parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset,
253 if (FileDescriptors.empty())
256 // Get the file entries, according to the format described above.
257 int FileEntryCount = DebugLineData.getU8(OffsetPtr);
258 for (int I = 0; I != FileEntryCount; ++I) {
259 if (*OffsetPtr >= EndPrologueOffset)
261 DWARFDebugLine::FileNameEntry FileEntry;
262 for (auto Descriptor : FileDescriptors) {
263 DWARFFormValue Value(Descriptor.Form);
264 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
266 switch (Descriptor.Type) {
268 FileEntry.Name = Value;
270 case DW_LNCT_LLVM_source:
271 FileEntry.Source = Value;
273 case DW_LNCT_directory_index:
274 FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
276 case DW_LNCT_timestamp:
277 FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
280 FileEntry.Length = Value.getAsUnsignedConstant().getValue();
283 assert(Value.getAsBlock().getValue().size() == 16);
284 std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16,
285 FileEntry.Checksum.Bytes.begin());
291 FileNames.push_back(FileEntry);
296 Error DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData,
298 const DWARFContext &Ctx,
299 const DWARFUnit *U) {
300 const uint64_t PrologueOffset = *OffsetPtr;
303 TotalLength = DebugLineData.getRelocatedValue(4, OffsetPtr);
304 if (TotalLength == UINT32_MAX) {
305 FormParams.Format = dwarf::DWARF64;
306 TotalLength = DebugLineData.getU64(OffsetPtr);
307 } else if (TotalLength >= 0xfffffff0) {
308 return createStringError(errc::invalid_argument,
309 "parsing line table prologue at offset 0x%8.8" PRIx64
310 " unsupported reserved unit length found of value 0x%8.8" PRIx64,
311 PrologueOffset, TotalLength);
313 FormParams.Version = DebugLineData.getU16(OffsetPtr);
314 if (getVersion() < 2)
315 return createStringError(errc::not_supported,
316 "parsing line table prologue at offset 0x%8.8" PRIx64
317 " found unsupported version 0x%2.2" PRIx16,
318 PrologueOffset, getVersion());
320 if (getVersion() >= 5) {
321 FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
322 assert((DebugLineData.getAddressSize() == 0 ||
323 DebugLineData.getAddressSize() == getAddressSize()) &&
324 "Line table header and data extractor disagree");
325 SegSelectorSize = DebugLineData.getU8(OffsetPtr);
329 DebugLineData.getRelocatedValue(sizeofPrologueLength(), OffsetPtr);
330 const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
331 MinInstLength = DebugLineData.getU8(OffsetPtr);
332 if (getVersion() >= 4)
333 MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
334 DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
335 LineBase = DebugLineData.getU8(OffsetPtr);
336 LineRange = DebugLineData.getU8(OffsetPtr);
337 OpcodeBase = DebugLineData.getU8(OffsetPtr);
339 StandardOpcodeLengths.reserve(OpcodeBase - 1);
340 for (uint32_t I = 1; I < OpcodeBase; ++I) {
341 uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
342 StandardOpcodeLengths.push_back(OpLen);
345 if (getVersion() >= 5) {
346 if (!parseV5DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
347 FormParams, Ctx, U, ContentTypes,
348 IncludeDirectories, FileNames)) {
349 return createStringError(errc::invalid_argument,
350 "parsing line table prologue at 0x%8.8" PRIx64
351 " found an invalid directory or file table description at"
353 PrologueOffset, (uint64_t)*OffsetPtr);
356 parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
357 ContentTypes, IncludeDirectories, FileNames);
359 if (*OffsetPtr != EndPrologueOffset)
360 return createStringError(errc::invalid_argument,
361 "parsing line table prologue at 0x%8.8" PRIx64
362 " should have ended at 0x%8.8" PRIx64
363 " but it ended at 0x%8.8" PRIx64,
364 PrologueOffset, EndPrologueOffset, (uint64_t)*OffsetPtr);
365 return Error::success();
368 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
370 void DWARFDebugLine::Row::postAppend() {
374 EpilogueBegin = false;
377 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
379 Address.SectionIndex = object::SectionedAddress::UndefSection;
385 IsStmt = DefaultIsStmt;
389 EpilogueBegin = false;
392 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) {
393 OS << "Address Line Column File ISA Discriminator Flags\n"
394 << "------------------ ------ ------ ------ --- ------------- "
398 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
399 OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column)
400 << format(" %6u %3u %13u ", File, Isa, Discriminator)
401 << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
402 << (PrologueEnd ? " prologue_end" : "")
403 << (EpilogueBegin ? " epilogue_begin" : "")
404 << (EndSequence ? " end_sequence" : "") << '\n';
407 DWARFDebugLine::Sequence::Sequence() { reset(); }
409 void DWARFDebugLine::Sequence::reset() {
412 SectionIndex = object::SectionedAddress::UndefSection;
418 DWARFDebugLine::LineTable::LineTable() { clear(); }
420 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
421 DIDumpOptions DumpOptions) const {
422 Prologue.dump(OS, DumpOptions);
426 Row::dumpTableHeader(OS);
427 for (const Row &R : Rows) {
433 void DWARFDebugLine::LineTable::clear() {
439 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT)
441 resetRowAndSequence();
444 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
445 Row.reset(LineTable->Prologue.DefaultIsStmt);
449 void DWARFDebugLine::ParsingState::appendRowToMatrix() {
450 unsigned RowNumber = LineTable->Rows.size();
451 if (Sequence.Empty) {
452 // Record the beginning of instruction sequence.
453 Sequence.Empty = false;
454 Sequence.LowPC = Row.Address.Address;
455 Sequence.FirstRowIndex = RowNumber;
457 LineTable->appendRow(Row);
458 if (Row.EndSequence) {
459 // Record the end of instruction sequence.
460 Sequence.HighPC = Row.Address.Address;
461 Sequence.LastRowIndex = RowNumber + 1;
462 Sequence.SectionIndex = Row.Address.SectionIndex;
463 if (Sequence.isValid())
464 LineTable->appendSequence(Sequence);
470 const DWARFDebugLine::LineTable *
471 DWARFDebugLine::getLineTable(uint32_t Offset) const {
472 LineTableConstIter Pos = LineTableMap.find(Offset);
473 if (Pos != LineTableMap.end())
478 Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable(
479 DWARFDataExtractor &DebugLineData, uint32_t Offset, const DWARFContext &Ctx,
480 const DWARFUnit *U, std::function<void(Error)> RecoverableErrorCallback) {
481 if (!DebugLineData.isValidOffset(Offset))
482 return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx32
483 " is not a valid debug line section offset",
486 std::pair<LineTableIter, bool> Pos =
487 LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
488 LineTable *LT = &Pos.first->second;
491 LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorCallback))
492 return std::move(Err);
498 Error DWARFDebugLine::LineTable::parse(
499 DWARFDataExtractor &DebugLineData, uint32_t *OffsetPtr,
500 const DWARFContext &Ctx, const DWARFUnit *U,
501 std::function<void(Error)> RecoverableErrorCallback, raw_ostream *OS) {
502 const uint32_t DebugLineOffset = *OffsetPtr;
506 Error PrologueErr = Prologue.parse(DebugLineData, OffsetPtr, Ctx, U);
509 // The presence of OS signals verbose dumping.
510 DIDumpOptions DumpOptions;
511 DumpOptions.Verbose = true;
512 Prologue.dump(*OS, DumpOptions);
518 const uint32_t EndOffset =
519 DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength();
521 // See if we should tell the data extractor the address size.
522 if (DebugLineData.getAddressSize() == 0)
523 DebugLineData.setAddressSize(Prologue.getAddressSize());
525 assert(Prologue.getAddressSize() == 0 ||
526 Prologue.getAddressSize() == DebugLineData.getAddressSize());
528 ParsingState State(this);
530 while (*OffsetPtr < EndOffset) {
532 *OS << format("0x%08.08" PRIx32 ": ", *OffsetPtr);
534 uint8_t Opcode = DebugLineData.getU8(OffsetPtr);
537 *OS << format("%02.02" PRIx8 " ", Opcode);
540 // Extended Opcodes always start with a zero opcode followed by
541 // a uleb128 length so you can skip ones you don't know about
542 uint64_t Len = DebugLineData.getULEB128(OffsetPtr);
543 uint32_t ExtOffset = *OffsetPtr;
545 // Tolerate zero-length; assume length is correct and soldier on.
548 *OS << "Badly formed extended line op (length 0)\n";
552 uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr);
554 *OS << LNExtendedString(SubOpcode);
556 case DW_LNE_end_sequence:
557 // Set the end_sequence register of the state machine to true and
558 // append a row to the matrix using the current values of the
559 // state-machine registers. Then reset the registers to the initial
560 // values specified above. Every statement program sequence must end
561 // with a DW_LNE_end_sequence instruction which creates a row whose
562 // address is that of the byte after the last target machine instruction
564 State.Row.EndSequence = true;
565 State.appendRowToMatrix();
571 State.resetRowAndSequence();
574 case DW_LNE_set_address:
575 // Takes a single relocatable address as an operand. The size of the
576 // operand is the size appropriate to hold an address on the target
577 // machine. Set the address register to the value given by the
578 // relocatable address. All of the other statement program opcodes
579 // that affect the address register add a delta to it. This instruction
580 // stores a relocatable value into it instead.
582 // Make sure the extractor knows the address size. If not, infer it
583 // from the size of the operand.
584 if (DebugLineData.getAddressSize() == 0)
585 DebugLineData.setAddressSize(Len - 1);
586 else if (DebugLineData.getAddressSize() != Len - 1) {
587 return createStringError(errc::invalid_argument,
588 "mismatching address size at offset 0x%8.8" PRIx32
589 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64,
590 ExtOffset, DebugLineData.getAddressSize(),
593 State.Row.Address.Address = DebugLineData.getRelocatedAddress(
594 OffsetPtr, &State.Row.Address.SectionIndex);
596 *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address.Address);
599 case DW_LNE_define_file:
600 // Takes 4 arguments. The first is a null terminated string containing
601 // a source file name. The second is an unsigned LEB128 number
602 // representing the directory index of the directory in which the file
603 // was found. The third is an unsigned LEB128 number representing the
604 // time of last modification of the file. The fourth is an unsigned
605 // LEB128 number representing the length in bytes of the file. The time
606 // and length fields may contain LEB128(0) if the information is not
609 // The directory index represents an entry in the include_directories
610 // section of the statement program prologue. The index is LEB128(0)
611 // if the file was found in the current directory of the compilation,
612 // LEB128(1) if it was found in the first directory in the
613 // include_directories section, and so on. The directory index is
614 // ignored for file names that represent full path names.
616 // The files are numbered, starting at 1, in the order in which they
617 // appear; the names in the prologue come before names defined by
618 // the DW_LNE_define_file instruction. These numbers are used in the
619 // the file register of the state machine.
621 FileNameEntry FileEntry;
622 const char *Name = DebugLineData.getCStr(OffsetPtr);
624 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name);
625 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
626 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
627 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
628 Prologue.FileNames.push_back(FileEntry);
630 *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
631 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
632 << ", length=" << FileEntry.Length << ")";
636 case DW_LNE_set_discriminator:
637 State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr);
639 *OS << " (" << State.Row.Discriminator << ")";
644 *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
645 << format(" length %" PRIx64, Len);
646 // Len doesn't include the zero opcode byte or the length itself, but
647 // it does include the sub_opcode, so we have to adjust for that.
648 (*OffsetPtr) += Len - 1;
651 // Make sure the stated and parsed lengths are the same.
652 // Otherwise we have an unparseable line-number program.
653 if (*OffsetPtr - ExtOffset != Len)
654 return createStringError(errc::illegal_byte_sequence,
655 "unexpected line op length at offset 0x%8.8" PRIx32
656 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx32,
657 ExtOffset, Len, *OffsetPtr - ExtOffset);
658 } else if (Opcode < Prologue.OpcodeBase) {
660 *OS << LNStandardString(Opcode);
664 // Takes no arguments. Append a row to the matrix using the
665 // current values of the state-machine registers.
672 State.appendRowToMatrix();
675 case DW_LNS_advance_pc:
676 // Takes a single unsigned LEB128 operand, multiplies it by the
677 // min_inst_length field of the prologue, and adds the
678 // result to the address register of the state machine.
680 uint64_t AddrOffset =
681 DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength;
682 State.Row.Address.Address += AddrOffset;
684 *OS << " (" << AddrOffset << ")";
688 case DW_LNS_advance_line:
689 // Takes a single signed LEB128 operand and adds that value to
690 // the line register of the state machine.
691 State.Row.Line += DebugLineData.getSLEB128(OffsetPtr);
693 *OS << " (" << State.Row.Line << ")";
696 case DW_LNS_set_file:
697 // Takes a single unsigned LEB128 operand and stores it in the file
698 // register of the state machine.
699 State.Row.File = DebugLineData.getULEB128(OffsetPtr);
701 *OS << " (" << State.Row.File << ")";
704 case DW_LNS_set_column:
705 // Takes a single unsigned LEB128 operand and stores it in the
706 // column register of the state machine.
707 State.Row.Column = DebugLineData.getULEB128(OffsetPtr);
709 *OS << " (" << State.Row.Column << ")";
712 case DW_LNS_negate_stmt:
713 // Takes no arguments. Set the is_stmt register of the state
714 // machine to the logical negation of its current value.
715 State.Row.IsStmt = !State.Row.IsStmt;
718 case DW_LNS_set_basic_block:
719 // Takes no arguments. Set the basic_block register of the
720 // state machine to true
721 State.Row.BasicBlock = true;
724 case DW_LNS_const_add_pc:
725 // Takes no arguments. Add to the address register of the state
726 // machine the address increment value corresponding to special
727 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
728 // when the statement program needs to advance the address by a
729 // small amount, it can use a single special opcode, which occupies
730 // a single byte. When it needs to advance the address by up to
731 // twice the range of the last special opcode, it can use
732 // DW_LNS_const_add_pc followed by a special opcode, for a total
733 // of two bytes. Only if it needs to advance the address by more
734 // than twice that range will it need to use both DW_LNS_advance_pc
735 // and a special opcode, requiring three or more bytes.
737 uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase;
738 uint64_t AddrOffset =
739 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
740 State.Row.Address.Address += AddrOffset;
743 << format(" (0x%16.16" PRIx64 ")", AddrOffset);
747 case DW_LNS_fixed_advance_pc:
748 // Takes a single uhalf operand. Add to the address register of
749 // the state machine the value of the (unencoded) operand. This
750 // is the only extended opcode that takes an argument that is not
751 // a variable length number. The motivation for DW_LNS_fixed_advance_pc
752 // is this: existing assemblers cannot emit DW_LNS_advance_pc or
753 // special opcodes because they cannot encode LEB128 numbers or
754 // judge when the computation of a special opcode overflows and
755 // requires the use of DW_LNS_advance_pc. Such assemblers, however,
756 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
758 uint16_t PCOffset = DebugLineData.getRelocatedValue(2, OffsetPtr);
759 State.Row.Address.Address += PCOffset;
762 << format(" (0x%4.4" PRIx16 ")", PCOffset);
766 case DW_LNS_set_prologue_end:
767 // Takes no arguments. Set the prologue_end register of the
768 // state machine to true
769 State.Row.PrologueEnd = true;
772 case DW_LNS_set_epilogue_begin:
773 // Takes no arguments. Set the basic_block register of the
774 // state machine to true
775 State.Row.EpilogueBegin = true;
779 // Takes a single unsigned LEB128 operand and stores it in the
780 // column register of the state machine.
781 State.Row.Isa = DebugLineData.getULEB128(OffsetPtr);
783 *OS << " (" << State.Row.Isa << ")";
787 // Handle any unknown standard opcodes here. We know the lengths
788 // of such opcodes because they are specified in the prologue
789 // as a multiple of LEB128 operands for each opcode.
791 assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
792 uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
793 for (uint8_t I = 0; I < OpcodeLength; ++I) {
794 uint64_t Value = DebugLineData.getULEB128(OffsetPtr);
796 *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n",
805 // A special opcode value is chosen based on the amount that needs
806 // to be added to the line and address registers. The maximum line
807 // increment for a special opcode is the value of the line_base
808 // field in the header, plus the value of the line_range field,
809 // minus 1 (line base + line range - 1). If the desired line
810 // increment is greater than the maximum line increment, a standard
811 // opcode must be used instead of a special opcode. The "address
812 // advance" is calculated by dividing the desired address increment
813 // by the minimum_instruction_length field from the header. The
814 // special opcode is then calculated using the following formula:
816 // opcode = (desired line increment - line_base) +
817 // (line_range * address advance) + opcode_base
819 // If the resulting opcode is greater than 255, a standard opcode
820 // must be used instead.
822 // To decode a special opcode, subtract the opcode_base from the
823 // opcode itself to give the adjusted opcode. The amount to
824 // increment the address register is the result of the adjusted
825 // opcode divided by the line_range multiplied by the
826 // minimum_instruction_length field from the header. That is:
828 // address increment = (adjusted opcode / line_range) *
829 // minimum_instruction_length
831 // The amount to increment the line register is the line_base plus
832 // the result of the adjusted opcode modulo the line_range. That is:
834 // line increment = line_base + (adjusted opcode % line_range)
836 uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase;
837 uint64_t AddrOffset =
838 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
840 Prologue.LineBase + (AdjustOpcode % Prologue.LineRange);
841 State.Row.Line += LineOffset;
842 State.Row.Address.Address += AddrOffset;
845 *OS << "address += " << AddrOffset << ", line += " << LineOffset
851 State.appendRowToMatrix();
857 if (!State.Sequence.Empty)
858 RecoverableErrorCallback(
859 createStringError(errc::illegal_byte_sequence,
860 "last sequence in debug line table is not terminated!"));
862 // Sort all sequences so that address lookup will work faster.
863 if (!Sequences.empty()) {
864 llvm::sort(Sequences, Sequence::orderByHighPC);
865 // Note: actually, instruction address ranges of sequences should not
866 // overlap (in shared objects and executables). If they do, the address
867 // lookup would still work, though, but result would be ambiguous.
868 // We don't report warning in this case. For example,
869 // sometimes .so compiled from multiple object files contains a few
870 // rudimentary sequences for address ranges [0x0, 0xsomething).
873 return Error::success();
876 uint32_t DWARFDebugLine::LineTable::findRowInSeq(
877 const DWARFDebugLine::Sequence &Seq,
878 object::SectionedAddress Address) const {
879 if (!Seq.containsPC(Address))
880 return UnknownRowIndex;
881 assert(Seq.SectionIndex == Address.SectionIndex);
882 // In some cases, e.g. first instruction in a function, the compiler generates
883 // two entries, both with the same address. We want the last one.
885 // In general we want a non-empty range: the last row whose address is less
886 // than or equal to Address. This can be computed as upper_bound - 1.
887 DWARFDebugLine::Row Row;
888 Row.Address = Address;
889 RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
890 RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
891 assert(FirstRow->Address.Address <= Row.Address.Address &&
892 Row.Address.Address < LastRow[-1].Address.Address);
893 RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row,
894 DWARFDebugLine::Row::orderByAddress) -
896 assert(Seq.SectionIndex == RowPos->Address.SectionIndex);
897 return RowPos - Rows.begin();
900 uint32_t DWARFDebugLine::LineTable::lookupAddress(
901 object::SectionedAddress Address) const {
903 // Search for relocatable addresses
904 uint32_t Result = lookupAddressImpl(Address);
906 if (Result != UnknownRowIndex ||
907 Address.SectionIndex == object::SectionedAddress::UndefSection)
910 // Search for absolute addresses
911 Address.SectionIndex = object::SectionedAddress::UndefSection;
912 return lookupAddressImpl(Address);
915 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl(
916 object::SectionedAddress Address) const {
917 // First, find an instruction sequence containing the given address.
918 DWARFDebugLine::Sequence Sequence;
919 Sequence.SectionIndex = Address.SectionIndex;
920 Sequence.HighPC = Address.Address;
921 SequenceIter It = llvm::upper_bound(Sequences, Sequence,
922 DWARFDebugLine::Sequence::orderByHighPC);
923 if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex)
924 return UnknownRowIndex;
925 return findRowInSeq(*It, Address);
928 bool DWARFDebugLine::LineTable::lookupAddressRange(
929 object::SectionedAddress Address, uint64_t Size,
930 std::vector<uint32_t> &Result) const {
932 // Search for relocatable addresses
933 if (lookupAddressRangeImpl(Address, Size, Result))
936 if (Address.SectionIndex == object::SectionedAddress::UndefSection)
939 // Search for absolute addresses
940 Address.SectionIndex = object::SectionedAddress::UndefSection;
941 return lookupAddressRangeImpl(Address, Size, Result);
944 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
945 object::SectionedAddress Address, uint64_t Size,
946 std::vector<uint32_t> &Result) const {
947 if (Sequences.empty())
949 uint64_t EndAddr = Address.Address + Size;
950 // First, find an instruction sequence containing the given address.
951 DWARFDebugLine::Sequence Sequence;
952 Sequence.SectionIndex = Address.SectionIndex;
953 Sequence.HighPC = Address.Address;
954 SequenceIter LastSeq = Sequences.end();
955 SequenceIter SeqPos = llvm::upper_bound(
956 Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC);
957 if (SeqPos == LastSeq || !SeqPos->containsPC(Address))
960 SequenceIter StartPos = SeqPos;
962 // Add the rows from the first sequence to the vector, starting with the
963 // index we just calculated
965 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
966 const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
967 // For the first sequence, we need to find which row in the sequence is the
968 // first in our range.
969 uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
970 if (SeqPos == StartPos)
971 FirstRowIndex = findRowInSeq(CurSeq, Address);
973 // Figure out the last row in the range.
974 uint32_t LastRowIndex =
975 findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
976 if (LastRowIndex == UnknownRowIndex)
977 LastRowIndex = CurSeq.LastRowIndex - 1;
979 assert(FirstRowIndex != UnknownRowIndex);
980 assert(LastRowIndex != UnknownRowIndex);
982 for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
992 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
993 FileLineInfoKind Kind) const {
994 if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex))
996 const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex);
997 if (Optional<const char *> source = Entry.Source.getAsCString())
998 return StringRef(*source);
1002 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
1003 // Debug info can contain paths from any OS, not necessarily
1004 // an OS we're currently running on. Moreover different compilation units can
1005 // be compiled on different operating systems and linked together later.
1006 return sys::path::is_absolute(Path, sys::path::Style::posix) ||
1007 sys::path::is_absolute(Path, sys::path::Style::windows);
1010 bool DWARFDebugLine::Prologue::getFileNameByIndex(uint64_t FileIndex,
1012 FileLineInfoKind Kind,
1013 std::string &Result) const {
1014 if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
1016 const FileNameEntry &Entry = getFileNameEntry(FileIndex);
1017 StringRef FileName = Entry.Name.getAsCString().getValue();
1018 if (Kind != FileLineInfoKind::AbsoluteFilePath ||
1019 isPathAbsoluteOnWindowsOrPosix(FileName)) {
1024 SmallString<16> FilePath;
1025 StringRef IncludeDir;
1026 // Be defensive about the contents of Entry.
1027 if (getVersion() >= 5) {
1028 if (Entry.DirIdx < IncludeDirectories.size())
1029 IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue();
1031 if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size())
1033 IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue();
1035 // We may still need to append compilation directory of compile unit.
1036 // We know that FileName is not absolute, the only way to have an
1037 // absolute path at this point would be if IncludeDir is absolute.
1038 if (!CompDir.empty() && !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1039 sys::path::append(FilePath, CompDir);
1042 // sys::path::append skips empty strings.
1043 sys::path::append(FilePath, IncludeDir, FileName);
1044 Result = FilePath.str();
1048 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1049 object::SectionedAddress Address, const char *CompDir,
1050 FileLineInfoKind Kind, DILineInfo &Result) const {
1051 // Get the index of row we're looking for in the line table.
1052 uint32_t RowIndex = lookupAddress(Address);
1053 if (RowIndex == -1U)
1055 // Take file number and line/column from the row.
1056 const auto &Row = Rows[RowIndex];
1057 if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1059 Result.Line = Row.Line;
1060 Result.Column = Row.Column;
1061 Result.Discriminator = Row.Discriminator;
1062 Result.Source = getSourceByIndex(Row.File, Kind);
1066 // We want to supply the Unit associated with a .debug_line[.dwo] table when
1067 // we dump it, if possible, but still dump the table even if there isn't a Unit.
1068 // Therefore, collect up handles on all the Units that point into the
1069 // line-table section.
1070 static DWARFDebugLine::SectionParser::LineToUnitMap
1071 buildLineToUnitMap(DWARFDebugLine::SectionParser::cu_range CUs,
1072 DWARFDebugLine::SectionParser::tu_range TUs) {
1073 DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit;
1074 for (const auto &CU : CUs)
1075 if (auto CUDIE = CU->getUnitDIE())
1076 if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list)))
1077 LineToUnit.insert(std::make_pair(*StmtOffset, &*CU));
1078 for (const auto &TU : TUs)
1079 if (auto TUDIE = TU->getUnitDIE())
1080 if (auto StmtOffset = toSectionOffset(TUDIE.find(DW_AT_stmt_list)))
1081 LineToUnit.insert(std::make_pair(*StmtOffset, &*TU));
1085 DWARFDebugLine::SectionParser::SectionParser(DWARFDataExtractor &Data,
1086 const DWARFContext &C,
1087 cu_range CUs, tu_range TUs)
1088 : DebugLineData(Data), Context(C) {
1089 LineToUnit = buildLineToUnitMap(CUs, TUs);
1090 if (!DebugLineData.isValidOffset(Offset))
1094 bool DWARFDebugLine::Prologue::totalLengthIsValid() const {
1095 return TotalLength == 0xffffffff || TotalLength < 0xfffffff0;
1098 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext(
1099 function_ref<void(Error)> RecoverableErrorCallback,
1100 function_ref<void(Error)> UnrecoverableErrorCallback, raw_ostream *OS) {
1101 assert(DebugLineData.isValidOffset(Offset) &&
1102 "parsing should have terminated");
1103 DWARFUnit *U = prepareToParse(Offset);
1104 uint32_t OldOffset = Offset;
1106 if (Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1107 RecoverableErrorCallback, OS))
1108 UnrecoverableErrorCallback(std::move(Err));
1109 moveToNextTable(OldOffset, LT.Prologue);
1113 void DWARFDebugLine::SectionParser::skip(
1114 function_ref<void(Error)> ErrorCallback) {
1115 assert(DebugLineData.isValidOffset(Offset) &&
1116 "parsing should have terminated");
1117 DWARFUnit *U = prepareToParse(Offset);
1118 uint32_t OldOffset = Offset;
1120 if (Error Err = LT.Prologue.parse(DebugLineData, &Offset, Context, U))
1121 ErrorCallback(std::move(Err));
1122 moveToNextTable(OldOffset, LT.Prologue);
1125 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint32_t Offset) {
1126 DWARFUnit *U = nullptr;
1127 auto It = LineToUnit.find(Offset);
1128 if (It != LineToUnit.end())
1130 DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0);
1134 void DWARFDebugLine::SectionParser::moveToNextTable(uint32_t OldOffset,
1135 const Prologue &P) {
1136 // If the length field is not valid, we don't know where the next table is, so
1137 // cannot continue to parse. Mark the parser as done, and leave the Offset
1138 // value as it currently is. This will be the end of the bad length field.
1139 if (!P.totalLengthIsValid()) {
1144 Offset = OldOffset + P.TotalLength + P.sizeofTotalLength();
1145 if (!DebugLineData.isValidOffset(Offset)) {