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//===----------------------------------------------------------------------===//
35struct ContentDescriptor {
42}
// end anonymous namespace
45 return Version >= 2 && Version <= 5;
50 switch (ContentType) {
51 case dwarf::DW_LNCT_timestamp:
54 case dwarf::DW_LNCT_size:
57 case dwarf::DW_LNCT_MD5:
60 case dwarf::DW_LNCT_LLVM_source:
64 // We only care about values we consider optional, and new values may be
65 // added in the vendor extension range, so we do not match exhaustively.
74 assert(DwarfVersion != 0 &&
75 "line table prologue has no dwarf version information");
76 if (DwarfVersion >= 5)
78 return FileIndex != 0 && FileIndex <=
FileNames.size();
81std::optional<uint64_t>
86 assert(DwarfVersion != 0 &&
87 "line table prologue has no dwarf version information");
88 // In DWARF v5 the file names are 0-indexed.
89 if (DwarfVersion >= 5)
97 assert(DwarfVersion != 0 &&
98 "line table prologue has no dwarf version information");
99 // In DWARF v5 the file names are 0-indexed.
100 if (DwarfVersion >= 5)
122 OS <<
"Line table prologue:\n"
123 <<
format(
" total_length: 0x%0*" PRIx64
"\n", OffsetDumpWidth,
132 OS <<
format(
" prologue_length: 0x%0*" PRIx64
"\n", OffsetDumpWidth,
142 OS <<
formatv(
"standard_opcode_lengths[{0}] = {1}\n",
147 // DWARF v5 starts directory indexes at 0.
150 OS <<
format(
"include_directories[%3u] = ",
I + DirBase);
157 // DWARF v5 starts file indexes at 0.
161 OS <<
format(
"file_names[%3u]:\n",
I + FileBase);
163 FileEntry.
Name.
dump(OS, DumpOptions);
164 OS <<
'\n' <<
format(
" dir_index: %" PRIu64
"\n", FileEntry.
DirIdx);
168 OS <<
format(
" mod_time: 0x%8.8" PRIx64
"\n", FileEntry.
ModTime);
170 OS <<
format(
" length: 0x%8.8" PRIx64
"\n", FileEntry.
Length);
175 else if ((*Source)[0]) {
185// Parse v2-v4 directory and file tables.
190 std::vector<DWARFFormValue> &IncludeDirectories,
191 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
198 "include directories table was not null "
199 "terminated before the end of the prologue");
205 IncludeDirectories.push_back(Dir);
214 if (!Err && Name.empty())
228 "file names table was not null terminated before "
229 "the end of the prologue");
231 FileNames.push_back(FileEntry);
237// Parse v5 directory/file entry content descriptions.
238// Returns the descriptors, or an error if we did not find a path or ran off
239// the end of the prologue.
240static llvm::Expected<ContentDescriptors>
244 ContentDescriptors Descriptors;
245 int FormatCount = DebugLineData.
getU8(OffsetPtr, &Err);
246 bool HasPath =
false;
247 for (
int I = 0;
I != FormatCount && !Err; ++
I) {
248 ContentDescriptor Descriptor;
252 if (Descriptor.Type == dwarf::DW_LNCT_path)
256 Descriptors.push_back(Descriptor);
261 "failed to parse entry content descriptors: %s",
266 "failed to parse entry content descriptions"
267 " because no path was found");
276 std::vector<DWARFFormValue> &IncludeDirectories,
277 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
278 // Get the directory entry description.
284 // Get the directory entries, according to the format described above.
287 for (
auto Descriptor : *DirDescriptors) {
289 switch (Descriptor.Type) {
291 if (!
Value.extractValue(DebugLineData, OffsetPtr,
FormParams, &Ctx, U))
293 "failed to parse directory entry because "
294 "extracting the form value failed");
295 IncludeDirectories.push_back(
Value);
300 "failed to parse directory entry because "
301 "skipping the form value failed");
306 // Get the file entry description.
309 if (!FileDescriptors)
312 // Get the file entries, according to the format described above.
316 for (
auto Descriptor : *FileDescriptors) {
318 if (!
Value.extractValue(DebugLineData, OffsetPtr,
FormParams, &Ctx, U))
320 "failed to parse file entry because "
321 "extracting the form value failed");
322 switch (Descriptor.Type) {
326 case DW_LNCT_LLVM_source:
329 case DW_LNCT_directory_index:
332 case DW_LNCT_timestamp:
339 if (!
Value.getAsBlock() ||
Value.getAsBlock()->size() != 16)
342 "failed to parse file entry because the MD5 hash is invalid");
350 FileNames.push_back(FileEntry);
359 Length += 2;
// Address + Segment selector sizes.
367 const uint64_t PrologueOffset = *OffsetPtr;
378 // Treat this error as unrecoverable - we cannot be sure what any of
379 // the data represents including the length field, so cannot skip it or make
380 // any reasonable assumptions.
381 *OffsetPtr = Cursor.tell();
384 "parsing line table prologue at offset 0x%8.8" PRIx64
385 ": unsupported version %" PRIu16,
394 if (DataAddrSize == 0) {
395 if (PrologueAddrSize != 4 && PrologueAddrSize != 8) {
398 "parsing line table prologue at offset 0x%8.8" PRIx64
399 ": invalid address size %" PRIu8,
400 PrologueOffset, PrologueAddrSize));
402 }
else if (DataAddrSize != PrologueAddrSize) {
405 "parsing line table prologue at offset 0x%8.8" PRIx64
": address "
406 "size %" PRIu8
" doesn't match architecture address size %" PRIu8,
407 PrologueOffset, PrologueAddrSize, DataAddrSize));
426 // If the opcode base is 0, we cannot read the standard opcode lengths (of
427 // which there are supposed to be one fewer than the opcode base). Assume
428 // there are no standard opcodes and continue parsing.
431 "parsing line table prologue at offset 0x%8.8" PRIx64
432 " found opcode base of 0. Assuming no standard opcodes",
442 *OffsetPtr = Cursor.tell();
443 // A corrupt file name or directory table does not prevent interpretation of
444 // the main line program, so check the cursor state now so that its errors can
445 // be handled separately.
449 "parsing line table prologue at offset 0x%8.8" PRIx64
": %s",
450 PrologueOffset,
toString(Cursor.takeError()).c_str());
462 "parsing line table prologue at 0x%8.8" PRIx64
463 " found an invalid directory or file table description at"
465 PrologueOffset, *OffsetPtr),
470 assert(*OffsetPtr <= EndPrologueOffset);
471 if (*OffsetPtr != EndPrologueOffset) {
474 "unknown data in line table prologue at offset 0x%8.8" PRIx64
475 ": parsing ended (at offset 0x%8.8" PRIx64
476 ") before reaching the prologue end at offset 0x%8.8" PRIx64,
477 PrologueOffset, *OffsetPtr, EndPrologueOffset));
509 <<
"Address Line Column File ISA Discriminator OpIndex "
512 <<
"------------------ ------ ------ ------ --- ------------- ------- "
551 // Terminate the table with a final blank line to clearly delineate it from
562DWARFDebugLine::ParsingState::ParsingState(
567void DWARFDebugLine::ParsingState::resetRowAndSequence(
uint64_t Offset) {
568 Row.reset(LineTable->Prologue.DefaultIsStmt);
570 Sequence.StmtSeqOffset =
Offset;
573void DWARFDebugLine::ParsingState::appendRowToMatrix() {
574 unsigned RowNumber = LineTable->Rows.size();
575 if (Sequence.Empty) {
576 // Record the beginning of instruction sequence.
577 Sequence.Empty =
false;
578 Sequence.LowPC = Row.Address.Address;
579 Sequence.FirstRowIndex = RowNumber;
583 // Record the end of instruction sequence.
594const DWARFDebugLine::LineTable *
596 LineTableConstIter Pos = LineTableMap.find(
Offset);
597 if (Pos != LineTableMap.end())
607 "offset 0x%8.8" PRIx64
608 " is not a valid debug line section offset",
611 std::pair<LineTableIter, bool> Pos =
616 LT->parse(DebugLineData, &
Offset, Ctx, U, RecoverableErrorHandler))
617 return std::move(Err);
624 LineTableMap.erase(
Offset);
629 if (Opcode < OpcodeBase)
634DWARFDebugLine::ParsingState::AddrOpIndexDelta
635DWARFDebugLine::ParsingState::advanceAddrOpIndex(uint64_t OperationAdvance,
637 uint64_t OpcodeOffset) {
639 // For versions less than 4, the MaxOpsPerInst member is set to 0, as the
640 // maximum_operations_per_instruction field wasn't introduced until DWARFv4.
641 // Don't warn about bad values in this situation.
646 "line table program at offset 0x%8.8" PRIx64
647 " contains a %s opcode at offset 0x%8.8" PRIx64
648 ", but the prologue maximum_operations_per_instruction value is 0"
649 ", which is invalid. Assuming a value of 1 instead",
650 LineTableOffset, OpcodeName.
data(), OpcodeOffset));
651 // Although we are able to correctly parse line number programs with
652 // MaxOpsPerInst > 1, the rest of DWARFDebugLine and its
653 // users have not been updated to handle line information for all operations
654 // in a multi-operation instruction, so warn about potentially incorrect
659 "line table program at offset 0x%8.8" PRIx64
660 " contains a %s opcode at offset 0x%8.8" PRIx64
661 ", but the prologue maximum_operations_per_instruction value is %" PRId8
662 ", which is experimentally supported, so line number information "
664 LineTableOffset, OpcodeName.
data(), OpcodeOffset,
669 "line table program at offset 0x%8.8" PRIx64
670 " contains a %s opcode at offset 0x%8.8" PRIx64
671 ", but the prologue minimum_instruction_length value "
672 "is 0, which prevents any address advancing",
673 LineTableOffset, OpcodeName.
data(), OpcodeOffset));
674 ReportAdvanceAddrProblem =
false;
676 // Advances the address and op_index according to DWARFv5, section 6.2.5.1:
678 // new address = address +
679 // minimum_instruction_length *
680 // ((op_index + operation advance) / maximum_operations_per_instruction)
683 // (op_index + operation advance) % maximum_operations_per_instruction
685 // For versions less than 4, the MaxOpsPerInst member is set to 0, as the
686 // maximum_operations_per_instruction field wasn't introduced until DWARFv4.
687 uint8_t MaxOpsPerInst =
690 uint64_t AddrOffset = ((
Row.
OpIndex + OperationAdvance) / MaxOpsPerInst) *
696 int16_t OpIndexDelta =
static_cast<int16_t
>(
Row.
OpIndex) - PrevOpIndex;
698 return {AddrOffset, OpIndexDelta};
701DWARFDebugLine::ParsingState::OpcodeAdvanceResults
702DWARFDebugLine::ParsingState::advanceForOpcode(uint8_t Opcode,
703 uint64_t OpcodeOffset) {
704 assert(Opcode == DW_LNS_const_add_pc ||
707 StringRef OpcodeName =
711 "line table program at offset 0x%8.8" PRIx64
712 " contains a %s opcode at offset 0x%8.8" PRIx64
713 ", but the prologue line_range value is 0. The "
714 "address and line will not be adjusted",
715 LineTableOffset, OpcodeName.
data(), OpcodeOffset));
716 ReportBadLineRange =
false;
719 uint8_t OpcodeValue = Opcode;
720 if (Opcode == DW_LNS_const_add_pc)
723 uint64_t OperationAdvance =
727 AddrOpIndexDelta Advance =
728 advanceAddrOpIndex(OperationAdvance, Opcode, OpcodeOffset);
729 return {Advance.AddrOffset, Advance.OpIndexDelta, AdjustedOpcode};
732DWARFDebugLine::ParsingState::SpecialOpcodeDelta
733DWARFDebugLine::ParsingState::handleSpecialOpcode(uint8_t Opcode,
734 uint64_t OpcodeOffset) {
735 // A special opcode value is chosen based on the amount that needs
736 // to be added to the line and address registers. The maximum line
737 // increment for a special opcode is the value of the line_base
738 // field in the header, plus the value of the line_range field,
739 // minus 1 (line base + line range - 1). If the desired line
740 // increment is greater than the maximum line increment, a standard
741 // opcode must be used instead of a special opcode. The "address
742 // advance" is calculated by dividing the desired address increment
743 // by the minimum_instruction_length field from the header. The
744 // special opcode is then calculated using the following formula:
746 // opcode = (desired line increment - line_base) +
747 // (line_range * address advance) + opcode_base
749 // If the resulting opcode is greater than 255, a standard opcode
750 // must be used instead.
752 // To decode a special opcode, subtract the opcode_base from the
753 // opcode itself to give the adjusted opcode. The amount to
754 // increment the address register is the result of the adjusted
755 // opcode divided by the line_range multiplied by the
756 // minimum_instruction_length field from the header. That is:
758 // address increment = (adjusted opcode / line_range) *
759 // minimum_instruction_length
761 // The amount to increment the line register is the line_base plus
762 // the result of the adjusted opcode modulo the line_range. That is:
764 // line increment = line_base + (adjusted opcode % line_range)
766 DWARFDebugLine::ParsingState::OpcodeAdvanceResults AddrAdvanceResult =
767 advanceForOpcode(Opcode, OpcodeOffset);
768 int32_t LineOffset = 0;
774 return {AddrAdvanceResult.
AddrDelta, LineOffset,
778/// Parse a ULEB128 using the specified \p Cursor. \returns the parsed value on
779/// success, or std::nullopt if \p Cursor is in a failing state.
794 assert((OS || !
Verbose) &&
"cannot have verbose output without stream");
795 const uint64_t DebugLineOffset = *OffsetPtr;
800 Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U);
809 // Ensure there is a blank line after the prologue to clearly delineate it
819 assert(DebugLineData.
size() > DebugLineOffset &&
820 "prologue parsing should handle invalid offset");
821 uint64_t BytesRemaining = DebugLineData.
size() - DebugLineOffset;
822 RecoverableErrorHandler(
824 "line table program with offset 0x%8.8" PRIx64
825 " has length 0x%8.8" PRIx64
" but only 0x%8.8" PRIx64
826 " bytes are available",
827 DebugLineOffset, ProgramLength, BytesRemaining));
828 // Continue by capping the length at the number of remaining bytes.
829 ProgramLength = BytesRemaining;
832 // Create a DataExtractor which can only see the data up to the end of the
833 // table, to prevent reading past the end.
834 const uint64_t EndOffset = DebugLineOffset + ProgramLength;
837 // See if we should tell the data extractor the address size.
844 ParsingState State(
this, DebugLineOffset, RecoverableErrorHandler);
846 *OffsetPtr = DebugLineOffset +
Prologue.getLength();
847 if (OS && *OffsetPtr < EndOffset) {
851 // *OffsetPtr points to the end of the prologue - i.e. the start of the first
852 // sequence. So initialize the first sequence offset accordingly.
853 State.resetRowAndSequence(*OffsetPtr);
855 bool TombstonedAddress =
false;
857 if (!TombstonedAddress) {
864 State.appendRowToMatrix();
867 while (*OffsetPtr < EndOffset) {
871 *OS <<
format(
"0x%08.08" PRIx64
": ", *OffsetPtr);
875 size_t RowCount =
Rows.size();
878 *OS <<
format(
"%02.02" PRIx8
" ", Opcode);
881 // Extended Opcodes always start with a zero opcode followed by
882 // a uleb128 length so you can skip ones you don't know about
886 // Tolerate zero-length; assume length is correct and soldier on.
889 *OS <<
"Badly formed extended line op (length 0)\n";
893 RecoverableErrorHandler(Cursor.takeError());
895 *OffsetPtr = Cursor.tell();
900 // OperandOffset will be the same as ExtOffset, if it was not possible to
901 // read the SubOpcode.
902 uint64_t OperandOffset = Cursor.tell();
906 case DW_LNE_end_sequence:
907 // Set the end_sequence register of the state machine to true and
908 // append a row to the matrix using the current values of the
909 // state-machine registers. Then reset the registers to the initial
910 // values specified above. Every statement program sequence must end
911 // with a DW_LNE_end_sequence instruction which creates a row whose
912 // address is that of the byte after the last target machine instruction
914 State.Row.EndSequence =
true;
915 // No need to test the Cursor is valid here, since it must be to get
916 // into this code path - if it were invalid, the default case would be
919 // Cursor now points to right after the end_sequence opcode - so points
920 // to the start of the next sequence - if one exists.
921 State.resetRowAndSequence(Cursor.tell());
924 case DW_LNE_set_address:
925 // Takes a single relocatable address as an operand. The size of the
926 // operand is the size appropriate to hold an address on the target
927 // machine. Set the address register to the value given by the
928 // relocatable address and set the op_index register to 0. All of the
929 // other statement program opcodes that affect the address register
930 // add a delta to it. This instruction stores a relocatable value into
933 // Make sure the extractor knows the address size. If not, infer it
934 // from the size of the operand.
937 uint64_t OpcodeAddressSize = Len - 1;
938 if (ExtractorAddressSize != OpcodeAddressSize &&
939 ExtractorAddressSize != 0)
942 "mismatching address size at offset 0x%8.8" PRIx64
943 " expected 0x%2.2" PRIx8
" found 0x%2.2" PRIx64,
944 ExtOffset, ExtractorAddressSize, Len - 1));
946 // Assume that the line table is correct and temporarily override the
947 // address size. If the size is unsupported, give up trying to read
948 // the address and continue to the next opcode.
949 if (OpcodeAddressSize != 1 && OpcodeAddressSize != 2 &&
950 OpcodeAddressSize != 4 && OpcodeAddressSize != 8) {
953 "address size 0x%2.2" PRIx64
954 " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64
956 OpcodeAddressSize, ExtOffset));
957 TableData.
skip(Cursor, OpcodeAddressSize);
961 Cursor, &State.Row.Address.SectionIndex);
962 State.Row.OpIndex = 0;
966 TombstonedAddress = State.Row.Address.Address ==
Tombstone;
968 // Restore the address size if the extractor already had it.
969 if (ExtractorAddressSize != 0)
976 State.Row.Address.Address);
982 case DW_LNE_define_file:
983 // Takes 4 arguments. The first is a null terminated string containing
984 // a source file name. The second is an unsigned LEB128 number
985 // representing the directory index of the directory in which the file
986 // was found. The third is an unsigned LEB128 number representing the
987 // time of last modification of the file. The fourth is an unsigned
988 // LEB128 number representing the length in bytes of the file. The time
989 // and length fields may contain LEB128(0) if the information is not
992 // The directory index represents an entry in the include_directories
993 // section of the statement program prologue. The index is LEB128(0)
994 // if the file was found in the current directory of the compilation,
995 // LEB128(1) if it was found in the first directory in the
996 // include_directories section, and so on. The directory index is
997 // ignored for file names that represent full path names.
999 // The files are numbered, starting at 1, in the order in which they
1000 // appear; the names in the prologue come before names defined by
1001 // the DW_LNE_define_file instruction. These numbers are used in the
1002 // the file register of the state machine.
1005 const char *Name = TableData.
getCStr(Cursor);
1011 Prologue.FileNames.push_back(FileEntry);
1013 *OS <<
" (" << Name <<
", dir=" << FileEntry.
DirIdx <<
", mod_time="
1015 <<
", length=" << FileEntry.
Length <<
")";
1019 case DW_LNE_set_discriminator:
1020 State.Row.Discriminator = TableData.
getULEB128(Cursor);
1022 *OS <<
" (" << State.Row.Discriminator <<
")";
1027 *OS <<
format(
"Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
1028 <<
format(
" length %" PRIx64, Len);
1029 // Len doesn't include the zero opcode byte or the length itself, but
1030 // it does include the sub_opcode, so we have to adjust for that.
1031 TableData.
skip(Cursor, Len - 1);
1034 // Make sure the length as recorded in the table and the standard length
1035 // for the opcode match. If they don't, continue from the end as claimed
1036 // by the table. Similarly, continue from the claimed end in the event of
1039 if (Cursor && Cursor.tell() != End)
1042 "unexpected line op length at offset 0x%8.8" PRIx64
1043 " expected 0x%2.2" PRIx64
" found 0x%2.2" PRIx64,
1044 ExtOffset, Len, Cursor.tell() - ExtOffset));
1049 *OS <<
" (<parsing error>";
1051 *OS <<
format(
" %2.2" PRIx8, Byte);
1052 Byte = TableData.
getU8(ByteCursor);
1053 }
while (ByteCursor);
1057 // The only parse failure in this case should be if the end was reached.
1058 // In that case, throw away the error, as the main Cursor's error will
1063 }
else if (Opcode <
Prologue.OpcodeBase) {
1069 // Takes no arguments. Append a row to the matrix using the
1070 // current values of the state-machine registers.
1074 case DW_LNS_advance_pc:
1075 // Takes a single unsigned LEB128 operand as the operation advance
1076 // and modifies the address and op_index registers of the state machine
1077 // according to that.
1078 if (std::optional<uint64_t> Operand =
1081 State.advanceAddrOpIndex(*Operand, Opcode, OpcodeOffset);
1088 case DW_LNS_advance_line:
1089 // Takes a single signed LEB128 operand and adds that value to
1090 // the line register of the state machine.
1092 int64_t LineDelta = TableData.
getSLEB128(Cursor);
1094 State.Row.Line += LineDelta;
1096 *OS <<
" (" << State.Row.Line <<
")";
1101 case DW_LNS_set_file:
1102 // Takes a single unsigned LEB128 operand and stores it in the file
1103 // register of the state machine.
1104 if (std::optional<uint16_t> File =
1106 State.Row.File = *File;
1108 *OS <<
" (" << State.Row.File <<
")";
1112 case DW_LNS_set_column:
1113 // Takes a single unsigned LEB128 operand and stores it in the
1114 // column register of the state machine.
1115 if (std::optional<uint16_t> Column =
1117 State.Row.Column = *Column;
1119 *OS <<
" (" << State.Row.Column <<
")";
1123 case DW_LNS_negate_stmt:
1124 // Takes no arguments. Set the is_stmt register of the state
1125 // machine to the logical negation of its current value.
1126 State.Row.IsStmt = !State.Row.IsStmt;
1129 case DW_LNS_set_basic_block:
1130 // Takes no arguments. Set the basic_block register of the
1131 // state machine to true
1132 State.Row.BasicBlock =
true;
1135 case DW_LNS_const_add_pc:
1136 // Takes no arguments. Advance the address and op_index registers of
1137 // the state machine by the increments corresponding to special
1138 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
1139 // when the statement program needs to advance the address by a
1140 // small amount, it can use a single special opcode, which occupies
1141 // a single byte. When it needs to advance the address by up to
1142 // twice the range of the last special opcode, it can use
1143 // DW_LNS_const_add_pc followed by a special opcode, for a total
1144 // of two bytes. Only if it needs to advance the address by more
1145 // than twice that range will it need to use both DW_LNS_advance_pc
1146 // and a special opcode, requiring three or more bytes.
1149 State.advanceForOpcode(Opcode, OpcodeOffset);
1151 *OS <<
format(
" (addr += 0x%16.16" PRIx64
", op-index += %" PRIu8
1157 case DW_LNS_fixed_advance_pc:
1158 // Takes a single uhalf operand. Add to the address register of
1159 // the state machine the value of the (unencoded) operand and set
1160 // the op_index register to 0. This is the only extended opcode that
1161 // takes an argument that is not a variable length number.
1162 // The motivation for DW_LNS_fixed_advance_pc is this: existing
1163 // assemblers cannot emit DW_LNS_advance_pc or special opcodes because
1164 // they cannot encode LEB128 numbers or judge when the computation
1165 // of a special opcode overflows and requires the use of
1166 // DW_LNS_advance_pc. Such assemblers, however, can use
1167 // DW_LNS_fixed_advance_pc instead, sacrificing compression.
1171 State.Row.Address.Address += PCOffset;
1172 State.Row.OpIndex = 0;
1174 *OS <<
format(
" (addr += 0x%4.4" PRIx16
", op-index = 0)",
1180 case DW_LNS_set_prologue_end:
1181 // Takes no arguments. Set the prologue_end register of the
1182 // state machine to true
1183 State.Row.PrologueEnd =
true;
1186 case DW_LNS_set_epilogue_begin:
1187 // Takes no arguments. Set the basic_block register of the
1188 // state machine to true
1189 State.Row.EpilogueBegin =
true;
1192 case DW_LNS_set_isa:
1193 // Takes a single unsigned LEB128 operand and stores it in the
1194 // ISA register of the state machine.
1195 if (std::optional<uint8_t> Isa =
1197 State.Row.Isa = *Isa;
1199 *OS <<
" (" << (
uint64_t)State.Row.Isa <<
")";
1204 // Handle any unknown standard opcodes here. We know the lengths
1205 // of such opcodes because they are specified in the prologue
1206 // as a multiple of LEB128 operands for each opcode.
1210 *OS <<
"Unrecognized standard opcode";
1212 std::vector<uint64_t> Operands;
1214 if (std::optional<uint64_t>
Value =
1216 Operands.push_back(*
Value);
1220 if (
Verbose && !Operands.empty()) {
1221 *OS <<
" (operands: ";
1236 *OffsetPtr = Cursor.tell();
1240 State.handleSpecialOpcode(Opcode, OpcodeOffset);
1243 *OS <<
"address += " << Delta.
Address <<
", line += " << Delta.
Line
1244 <<
", op-index += " << Delta.
OpIndex;
1246 *OffsetPtr = Cursor.tell();
1249 // When a row is added to the matrix, it is also dumped, which includes a
1250 // new line already, so don't add an extra one.
1254 // Most parse failures other than when parsing extended opcodes are due to
1255 // failures to read ULEBs. Bail out of parsing, since we don't know where to
1256 // continue reading from as there is no stated length for such byte
1257 // sequences. Print the final trailing new line if needed before doing so.
1258 if (!Cursor && Opcode != 0) {
1261 return Cursor.takeError();
1265 RecoverableErrorHandler(Cursor.takeError());
1268 if (!State.Sequence.Empty)
1271 "last sequence in debug line table at offset 0x%8.8" PRIx64
1272 " is not terminated",
1275 // Sort all sequences so that address lookup will work faster.
1278 // Note: actually, instruction address ranges of sequences should not
1279 // overlap (in shared objects and executables). If they do, the address
1280 // lookup would still work, though, but result would be ambiguous.
1281 // We don't report warning in this case. For example,
1282 // sometimes .so compiled from multiple object files contains a few
1283 // rudimentary sequences for address ranges [0x0, 0xsomething).
1284 // Address ranges may also overlap when using ICF.
1287 // Terminate the table with a final blank line to clearly delineate it from
1295uint32_t DWARFDebugLine::LineTable::findRowInSeq(
1299 return UnknownRowIndex;
1301 // In some cases, e.g. first instruction in a function, the compiler generates
1302 // two entries, both with the same address. We want the last one.
1304 // In general we want a non-empty range: the last row whose address is less
1305 // than or equal to Address. This can be computed as upper_bound - 1.
1307 // TODO: This function, and its users, needs to be update to return multiple
1308 // rows for bundles with multiple op-indexes.
1315 RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1,
Row,
1319 return RowPos - Rows.begin();
1324 bool *IsApproximateLine)
const {
1326 // Search for relocatable addresses
1333 // Search for absolute addresses
1335 return lookupAddressImpl(
Address, IsApproximateLine);
1340 bool *IsApproximateLine)
const {
1341 assert((!IsApproximateLine || !*IsApproximateLine) &&
1342 "Make sure IsApproximateLine is appropriately "
1343 "initialized, if provided");
1344 // First, find an instruction sequence containing the given address.
1350 if (It == Sequences.end() || It->SectionIndex !=
Address.SectionIndex)
1351 return UnknownRowIndex;
1354 if (RowIndex == UnknownRowIndex || !IsApproximateLine)
1357 // Approximation will only be attempted if a valid RowIndex exists.
1358 uint32_t ApproxRowIndex = RowIndex;
1359 // Approximation Loop
1360 for (; ApproxRowIndex >= It->FirstRowIndex; --ApproxRowIndex) {
1361 if (Rows[ApproxRowIndex].Line)
1362 return ApproxRowIndex;
1363 *IsApproximateLine =
true;
1365 // Approximation Loop fails to find the valid ApproxRowIndex
1366 if (ApproxRowIndex < It->FirstRowIndex)
1367 *IsApproximateLine =
false;
1374 std::vector<uint32_t> &Result,
1375 std::optional<uint64_t> StmtSequenceOffset)
const {
1377 // Search for relocatable addresses
1378 if (lookupAddressRangeImpl(
Address,
Size, Result, StmtSequenceOffset))
1384 // Search for absolute addresses
1386 return lookupAddressRangeImpl(
Address,
Size, Result, StmtSequenceOffset);
1389bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
1391 std::vector<uint32_t> &Result,
1392 std::optional<uint64_t> StmtSequenceOffset)
const {
1393 if (Sequences.empty())
1396 // First, find an instruction sequence containing the given address.
1400 SequenceIter LastSeq = Sequences.end();
1401 SequenceIter SeqPos;
1403 if (StmtSequenceOffset) {
1404 // If we have a statement sequence offset, find the specific sequence.
1405 // Linear search for sequence with matching StmtSeqOffset
1406 SeqPos = std::find_if(Sequences.begin(), LastSeq,
1408 return S.StmtSeqOffset == *StmtSequenceOffset;
1411 // If sequence not found, return false
1412 if (SeqPos == LastSeq)
1415 // Set LastSeq to the next sequence since we only want the one matching
1416 // sequence (sequences are guaranteed to have unique StmtSeqOffset)
1417 LastSeq = SeqPos + 1;
1419 // No specific sequence requested, find first sequence containing address
1420 SeqPos = std::upper_bound(Sequences.begin(), LastSeq, Sequence,
1422 if (SeqPos == LastSeq)
1426 // If the start sequence doesn't contain the address, nothing to do
1427 if (!SeqPos->containsPC(Address))
1430 SequenceIter StartPos = SeqPos;
1432 // Process sequences that overlap with the desired range
1433 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
1435 // For the first sequence, we need to find which row in the sequence is the
1436 // first in our range.
1438 if (SeqPos == StartPos)
1439 FirstRowIndex = findRowInSeq(CurSeq, Address);
1441 // Figure out the last row in the range.
1443 findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
1444 if (LastRowIndex == UnknownRowIndex)
1447 assert(FirstRowIndex != UnknownRowIndex);
1448 assert(LastRowIndex != UnknownRowIndex);
1450 for (uint32_t
I = FirstRowIndex;
I <= LastRowIndex; ++
I) {
1460std::optional<StringRef>
1461DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
1462 FileLineInfoKind Kind)
const {
1464 return std::nullopt;
1467 return StringRef(*
E);
1468 return std::nullopt;
1472 // Debug info can contain paths from any OS, not necessarily
1473 // an OS we're currently running on. Moreover different compilation units can
1474 // be compiled on different operating systems and linked together later.
1482 if (Kind == FileLineInfoKind::None || !
hasFileAtIndex(FileIndex))
1489 if (Kind == FileLineInfoKind::RawValue ||
1491 Result = std::string(FileName);
1494 if (Kind == FileLineInfoKind::BaseNameOnly) {
1501 // Be defensive about the contents of Entry.
1503 // DirIdx 0 is the compilation directory, so don't include it for
1505 if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) &&
1513 // For absolute paths only, include the compilation directory of compile unit,
1514 // unless v5 DirIdx == 0 (IncludeDir indicates the compilation directory). We
1515 // know that FileName is not absolute, the only way to have an absolute path
1516 // at this point would be if IncludeDir is absolute.
1517 if (Kind == FileLineInfoKind::AbsoluteFilePath &&
1522 assert((Kind == FileLineInfoKind::AbsoluteFilePath ||
1523 Kind == FileLineInfoKind::RelativeFilePath) &&
1524 "invalid FileLineInfo Kind");
1526 // sys::path::append skips empty strings.
1528 Result = std::string(FilePath);
1534 FileLineInfoKind Kind,
DILineInfo &Result)
const {
1535 // Get the index of row we're looking for in the line table.
1538 if (RowIndex == -1U)
1540 // Take file number and line/column from the row.
1541 const auto &
Row =
Rows[RowIndex];
1547 Result.Source = getSourceByIndex(
Row.
File, Kind);
1554 if (Entry.DirIdx <
Prologue.IncludeDirectories.size()) {
1561 if (0 < Entry.DirIdx && Entry.DirIdx <=
Prologue.IncludeDirectories.size()) {
1569// We want to supply the Unit associated with a .debug_line[.dwo] table when
1570// we dump it, if possible, but still dump the table even if there isn't a Unit.
1571// Therefore, collect up handles on all the Units that point into the
1572// line-table section.
1576 for (
const auto &U : Units)
1577 if (
auto CUDIE = U->getUnitDIE())
1579 LineToUnit.insert(std::make_pair(*StmtOffset, &*U));
1586 : DebugLineData(
Data), Context(
C) {
1588 if (!DebugLineData.isValidOffset(Offset))
1600 assert(DebugLineData.isValidOffset(Offset) &&
1601 "parsing should have terminated");
1605 if (
Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1606 RecoverableErrorHandler, OS,
Verbose))
1607 UnrecoverableErrorHandler(std::move(Err));
1608 moveToNextTable(OldOffset, LT.Prologue);
1615 assert(DebugLineData.isValidOffset(Offset) &&
1616 "parsing should have terminated");
1620 if (
Error Err = LT.Prologue.parse(DebugLineData, &Offset,
1621 RecoverableErrorHandler, Context, U))
1622 UnrecoverableErrorHandler(std::move(Err));
1623 moveToNextTable(OldOffset, LT.Prologue);
1628 auto It = LineToUnit.find(
Offset);
1629 if (It != LineToUnit.end())
1635bool DWARFDebugLine::SectionParser::hasValidVersion(
uint64_t Offset) {
1641 // Ignore any error here.
1642 // If this is not the end of the section parseNext() will still be
1643 // attempted, where this error will occur again (and can be handled).
1650void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset,
1652 // If the length field is not valid, we don't know where the next table is, so
1653 // cannot continue to parse. Mark the parser as done, and leave the Offset
1654 // value as it currently is. This will be the end of the bad length field.
1655 if (!
P.totalLengthIsValid()) {
1660 Offset = OldOffset +
P.TotalLength +
P.sizeofTotalLength();
1666 // Heuristic: If the version is valid, then this is probably a line table.
1667 // Otherwise, the offset might need alignment (to a 4 or 8 byte boundary).
1668 if (hasValidVersion(
Offset))
1671 // ARM C/C++ Compiler aligns each line table to word boundaries and pads out
1672 // the .debug_line section to a word multiple. Note that in the specification
1673 // this does not seem forbidden since each unit has a DW_AT_stmt_list.
1674 for (
unsigned Align : {4, 8}) {
1677 // This is almost certainly not another line table but some alignment
1678 // padding. This assumes the alignments tested are ordered, and are
1679 // smaller than the header size (which is true for 4 and 8).
1683 if (hasValidVersion(AlignedOffset)) {
assert(UImm &&(UImm !=~static_cast< T >(0)) &&"Invalid immediate!")
static GCRegistry::Add< CoreCLRGC > E("coreclr", "CoreCLR-compatible GC")
static Error parseV5DirFileTables(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, const dwarf::FormParams &FormParams, const DWARFContext &Ctx, const DWARFUnit *U, DWARFDebugLine::ContentTypeTracker &ContentTypes, std::vector< DWARFFormValue > &IncludeDirectories, std::vector< DWARFDebugLine::FileNameEntry > &FileNames)
static Error parseV2DirFileTables(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, DWARFDebugLine::ContentTypeTracker &ContentTypes, std::vector< DWARFFormValue > &IncludeDirectories, std::vector< DWARFDebugLine::FileNameEntry > &FileNames)
static llvm::Expected< ContentDescriptors > parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, DWARFDebugLine::ContentTypeTracker *ContentTypes)
static DWARFDebugLine::SectionParser::LineToUnitMap buildLineToUnitMap(DWARFUnitVector::iterator_range Units)
static bool versionIsSupported(uint16_t Version)
static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase)
static std::optional< T > parseULEB128(DWARFDataExtractor &Data, DataExtractor::Cursor &Cursor)
Parse a ULEB128 using the specified Cursor.
This file contains constants used for implementing Dwarf debug support.
static fatal_error_handler_t ErrorHandler
This file defines the SmallString class.
This file defines the SmallVector class.
LocallyHashedType DenseMapInfo< LocallyHashedType >::Tombstone
DWARFContext This data structure is the top level entity that deals with dwarf debug information pars...
LLVM_ABI void skip(function_ref< void(Error)> RecoverableErrorHandler, function_ref< void(Error)> UnrecoverableErrorHandler)
Skip the current line table and go to the following line table (if present) immediately.
std::map< uint64_t, DWARFUnit * > LineToUnitMap
LLVM_ABI LineTable parseNext(function_ref< void(Error)> RecoverableErrorHandler, function_ref< void(Error)> UnrecoverableErrorHandler, raw_ostream *OS=nullptr, bool Verbose=false)
Get the next line table from the section.
LLVM_ABI SectionParser(DWARFDataExtractor &Data, const DWARFContext &C, DWARFUnitVector::iterator_range Units)
LLVM_ABI void clearLineTable(uint64_t Offset)
LLVM_ABI Expected< const LineTable * > getOrParseLineTable(DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx, const DWARFUnit *U, function_ref< void(Error)> RecoverableErrorHandler)
LLVM_ABI const LineTable * getLineTable(uint64_t Offset) const
llvm::iterator_range< UnitVector::iterator > iterator_range
Lightweight error class with error context and mandatory checking.
static ErrorSuccess success()
Create a success value.
Tagged union holding either a T or a Error.
Error takeError()
Take ownership of the stored error.
SmallString - A SmallString is just a SmallVector with methods and accessors that make it work better...
This is a 'vector' (really, a variable-sized array), optimized for the case when the array is small.
StringRef - Represent a constant reference to a string, i.e.
constexpr bool empty() const
empty - Check if the string is empty.
constexpr const char * data() const
data - Get a pointer to the start of the string (which may not be null terminated).
Twine - A lightweight data structure for efficiently representing the concatenation of temporary valu...
LLVM Value Representation.
An efficient, type-erasing, non-owning reference to a callable.
This class implements an extremely fast bulk output stream that can only output to a stream.
raw_ostream & indent(unsigned NumSpaces)
indent - Insert 'NumSpaces' spaces.
LLVM_ABI StringRef LNExtendedString(unsigned Encoding)
LLVM_ABI StringRef FormatString(DwarfFormat Format)
LLVM_ABI StringRef LNStandardString(unsigned Standard)
@ C
The default llvm calling convention, compatible with C.
bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path)
Calculates the starting offsets for various sections within the .debug_names section.
std::optional< const char * > toString(const std::optional< DWARFFormValue > &V)
Take an optional DWARFFormValue and try to extract a string value from it.
LineNumberOps
Line Number Standard Opcode Encodings.
std::optional< uint64_t > toSectionOffset(const std::optional< DWARFFormValue > &V)
Take an optional DWARFFormValue and try to extract an section offset.
StringRef toStringRef(const std::optional< DWARFFormValue > &V, StringRef Default={})
Take an optional DWARFFormValue and try to extract a string value from it.
uint8_t getDwarfOffsetByteSize(DwarfFormat Format)
The size of a reference determined by the DWARF 32/64-bit format.
uint64_t computeTombstoneAddress(uint8_t AddressByteSize)
LLVM_ABI StringRef filename(StringRef path LLVM_LIFETIME_BOUND, Style style=Style::native)
Get filename.
LLVM_ABI bool is_absolute(const Twine &path, Style style=Style::native)
Is path absolute?
LLVM_ABI void append(SmallVectorImpl< char > &path, const Twine &a, const Twine &b="", const Twine &c="", const Twine &d="")
Append to path.
This is an optimization pass for GlobalISel generic memory operations.
FunctionAddr VTableAddr Value
void stable_sort(R &&Range)
SmallVectorImpl< T >::const_pointer c_str(SmallVectorImpl< T > &str)
auto uninitialized_copy(R &&Src, IterTy Dst)
auto upper_bound(R &&Range, T &&Value)
Provide wrappers to std::upper_bound which take ranges instead of having to pass begin/end explicitly...
Error createStringError(std::error_code EC, char const *Fmt, const Ts &... Vals)
Create formatted StringError object.
auto formatv(bool Validate, const char *Fmt, Ts &&...Vals)
FunctionAddr VTableAddr uintptr_t uintptr_t Version
Error joinErrors(Error E1, Error E2)
Concatenate errors.
format_object< Ts... > format(const char *Fmt, const Ts &... Vals)
These are helper functions used to produce formatted output.
@ First
Helpers to iterate all locations in the MemoryEffectsBase class.
FunctionAddr VTableAddr uintptr_t uintptr_t Data
uint64_t alignTo(uint64_t Size, Align A)
Returns a multiple of A needed to store Size bytes.
std::string toString(const APInt &I, unsigned Radix, bool Signed, bool formatAsCLiteral=false, bool UpperCase=true, bool InsertSeparators=false)
void consumeError(Error Err)
Consume a Error without doing anything.
Container for dump options that control which debug information will be dumped.
A format-neutral container for source line information.
Tracks which optional content types are present in a DWARF file name entry format.
bool HasLength
Whether filename entries provide a file size.
bool HasSource
For v5, whether filename entries provide source text.
bool HasModTime
Whether filename entries provide a modification timestamp.
bool HasMD5
For v5, whether filename entries provide an MD5 checksum.
LLVM_ABI void trackContentType(dwarf::LineNumberEntryFormat ContentType)
Update tracked content types with ContentType.
LLVM_ABI uint32_t lookupAddress(object::SectionedAddress Address, bool *IsApproximateLine=nullptr) const
Returns the index of the row with file/line info for a given address, or UnknownRowIndex if there is ...
LLVM_ABI bool getDirectoryForEntry(const FileNameEntry &Entry, std::string &Directory) const
Extracts directory name by its Entry in include directories table in prologue.
const uint32_t UnknownRowIndex
Represents an invalid row.
LLVM_ABI bool getFileLineInfoForAddress(object::SectionedAddress Address, bool Approximate, const char *CompDir, DILineInfoSpecifier::FileLineInfoKind Kind, DILineInfo &Result) const
Fills the Result argument with the file and line information corresponding to Address.
LLVM_ABI Error parse(DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, const DWARFContext &Ctx, const DWARFUnit *U, function_ref< void(Error)> RecoverableErrorHandler, raw_ostream *OS=nullptr, bool Verbose=false)
Parse prologue and all rows.
bool getFileNameByIndex(uint64_t FileIndex, StringRef CompDir, DILineInfoSpecifier::FileLineInfoKind Kind, std::string &Result) const
Extracts filename by its index in filename table in prologue.
void appendSequence(const DWARFDebugLine::Sequence &S)
LLVM_ABI bool lookupAddressRange(object::SectionedAddress Address, uint64_t Size, std::vector< uint32_t > &Result, std::optional< uint64_t > StmtSequenceOffset=std::nullopt) const
Fills the Result argument with the indices of the rows that correspond to the address range specified...
void appendRow(const DWARFDebugLine::Row &R)
LLVM_ABI void dump(raw_ostream &OS, DIDumpOptions DumpOptions) const
uint8_t MaxOpsPerInst
The maximum number of individual operations that may be encoded in an instruction.
uint8_t MinInstLength
The size in bytes of the smallest target machine instruction.
LLVM_ABI bool hasFileAtIndex(uint64_t FileIndex) const
uint64_t PrologueLength
The number of bytes following the prologue_length field to the beginning of the first byte of the sta...
LLVM_ABI void dump(raw_ostream &OS, DIDumpOptions DumpOptions) const
uint32_t sizeofTotalLength() const
uint8_t SegSelectorSize
In v5, size in bytes of a segment selector.
uint16_t getVersion() const
int8_t LineBase
This parameter affects the meaning of the special opcodes. See below.
LLVM_ABI std::optional< uint64_t > getLastValidFileIndex() const
uint32_t sizeofPrologueLength() const
LLVM_ABI Error parse(DWARFDataExtractor Data, uint64_t *OffsetPtr, function_ref< void(Error)> RecoverableErrorHandler, const DWARFContext &Ctx, const DWARFUnit *U=nullptr)
uint8_t LineRange
This parameter affects the meaning of the special opcodes. See below.
std::vector< DWARFFormValue > IncludeDirectories
uint8_t OpcodeBase
The number assigned to the first special opcode.
std::vector< uint8_t > StandardOpcodeLengths
LLVM_ABI bool totalLengthIsValid() const
uint8_t getAddressSize() const
LLVM_ABI const llvm::DWARFDebugLine::FileNameEntry & getFileNameEntry(uint64_t Index) const
Get DWARF-version aware access to the file name entry at the provided index.
LLVM_ABI bool getFileNameByIndex(uint64_t FileIndex, StringRef CompDir, DILineInfoSpecifier::FileLineInfoKind Kind, std::string &Result, sys::path::Style Style=sys::path::Style::native) const
uint8_t DefaultIsStmt
The initial value of theis_stmtregister.
uint64_t TotalLength
The size in bytes of the statement information for this compilation unit (not including the total_len...
dwarf::FormParams FormParams
Version, address size (starting in v5), and DWARF32/64 format; these parameters affect interpretation...
LLVM_ABI uint64_t getLength() const
Length of the prologue in bytes.
ContentTypeTracker ContentTypes
This tracks which optional file format content types are present.
std::vector< FileNameEntry > FileNames
Standard .debug_line state machine structure.
uint8_t BasicBlock
A boolean indicating that the current instruction is the beginning of a basic block.
static bool orderByAddress(const Row &LHS, const Row &RHS)
uint32_t Line
An unsigned integer indicating a source line number.
uint16_t File
An unsigned integer indicating the identity of the source file corresponding to a machine instruction...
uint32_t Discriminator
An unsigned integer representing the DWARF path discriminator value for this location.
uint8_t EpilogueBegin
A boolean indicating that the current address is one (of possibly many) where execution should be sus...
object::SectionedAddress Address
The program-counter value corresponding to a machine instruction generated by the compiler and sectio...
LLVM_ABI void postAppend()
Called after a row is appended to the matrix.
uint8_t PrologueEnd
A boolean indicating that the current address is one (of possibly many) where execution should be sus...
uint16_t Column
An unsigned integer indicating a column number within a source line.
uint8_t EndSequence
A boolean indicating that the current address is that of the first byte after the end of a sequence o...
static LLVM_ABI void dumpTableHeader(raw_ostream &OS, unsigned Indent)
uint8_t IsStmt
A boolean indicating that the current instruction is the beginning of a statement.
LLVM_ABI void reset(bool DefaultIsStmt)
LLVM_ABI Row(bool DefaultIsStmt=false)
uint8_t Isa
An unsigned integer whose value encodes the applicable instruction set architecture for the current i...
LLVM_ABI void dump(raw_ostream &OS) const
uint8_t OpIndex
An unsigned integer representing the index of an operation within a VLIW instruction.
Represents a series of contiguous machine instructions.
uint64_t LowPC
Sequence describes instructions at address range [LowPC, HighPC) and is described by line table rows ...
static bool orderByHighPC(const Sequence &LHS, const Sequence &RHS)
bool containsPC(object::SectionedAddress PC) const
uint64_t StmtSeqOffset
The offset into the line table where this sequence begins.
uint64_t SectionIndex
If relocation information is present then this is the index of the section which contains above addre...
LLVM_ABI SmallString< 32 > digest() const
static const uint64_t UndefSection