[InstCombine] Signed saturation patterns
[llvm-core.git] / lib / DebugInfo / DWARF / DWARFDebugLine.cpp
blobdbee28ff5ab17957cbc1b003b7680bfda120ac45
1 //===- DWARFDebugLine.cpp -------------------------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
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/WithColor.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>
28 using namespace llvm;
29 using namespace dwarf;
31 using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind;
33 namespace {
35 struct ContentDescriptor {
36 dwarf::LineNumberEntryFormat Type;
37 dwarf::Form Form;
40 using ContentDescriptors = SmallVector<ContentDescriptor, 4>;
42 } // end anonmyous namespace
44 void DWARFDebugLine::ContentTypeTracker::trackContentType(
45 dwarf::LineNumberEntryFormat ContentType) {
46 switch (ContentType) {
47 case dwarf::DW_LNCT_timestamp:
48 HasModTime = true;
49 break;
50 case dwarf::DW_LNCT_size:
51 HasLength = true;
52 break;
53 case dwarf::DW_LNCT_MD5:
54 HasMD5 = true;
55 break;
56 case dwarf::DW_LNCT_LLVM_source:
57 HasSource = true;
58 break;
59 default:
60 // We only care about values we consider optional, and new values may be
61 // added in the vendor extension range, so we do not match exhaustively.
62 break;
66 DWARFDebugLine::Prologue::Prologue() { clear(); }
68 bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const {
69 uint16_t DwarfVersion = getVersion();
70 assert(DwarfVersion != 0 &&
71 "line table prologue has no dwarf version information");
72 if (DwarfVersion >= 5)
73 return FileIndex < FileNames.size();
74 return FileIndex != 0 && FileIndex <= FileNames.size();
77 const llvm::DWARFDebugLine::FileNameEntry &
78 DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const {
79 uint16_t DwarfVersion = getVersion();
80 assert(DwarfVersion != 0 &&
81 "line table prologue has no dwarf version information");
82 // In DWARF v5 the file names are 0-indexed.
83 if (DwarfVersion >= 5)
84 return FileNames[Index];
85 return FileNames[Index - 1];
88 void DWARFDebugLine::Prologue::clear() {
89 TotalLength = PrologueLength = 0;
90 SegSelectorSize = 0;
91 MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0;
92 OpcodeBase = 0;
93 FormParams = dwarf::FormParams({0, 0, DWARF32});
94 ContentTypes = ContentTypeTracker();
95 StandardOpcodeLengths.clear();
96 IncludeDirectories.clear();
97 FileNames.clear();
100 void DWARFDebugLine::Prologue::dump(raw_ostream &OS,
101 DIDumpOptions DumpOptions) const {
102 OS << "Line table prologue:\n"
103 << format(" total_length: 0x%8.8" PRIx64 "\n", TotalLength)
104 << format(" version: %u\n", getVersion());
105 if (getVersion() >= 5)
106 OS << format(" address_size: %u\n", getAddressSize())
107 << format(" seg_select_size: %u\n", SegSelectorSize);
108 OS << format(" prologue_length: 0x%8.8" PRIx64 "\n", PrologueLength)
109 << format(" min_inst_length: %u\n", MinInstLength)
110 << format(getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst)
111 << format(" default_is_stmt: %u\n", DefaultIsStmt)
112 << format(" line_base: %i\n", LineBase)
113 << format(" line_range: %u\n", LineRange)
114 << format(" opcode_base: %u\n", OpcodeBase);
116 for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I)
117 OS << format("standard_opcode_lengths[%s] = %u\n",
118 LNStandardString(I + 1).data(), StandardOpcodeLengths[I]);
120 if (!IncludeDirectories.empty()) {
121 // DWARF v5 starts directory indexes at 0.
122 uint32_t DirBase = getVersion() >= 5 ? 0 : 1;
123 for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) {
124 OS << format("include_directories[%3u] = ", I + DirBase);
125 IncludeDirectories[I].dump(OS, DumpOptions);
126 OS << '\n';
130 if (!FileNames.empty()) {
131 // DWARF v5 starts file indexes at 0.
132 uint32_t FileBase = getVersion() >= 5 ? 0 : 1;
133 for (uint32_t I = 0; I != FileNames.size(); ++I) {
134 const FileNameEntry &FileEntry = FileNames[I];
135 OS << format("file_names[%3u]:\n", I + FileBase);
136 OS << " name: ";
137 FileEntry.Name.dump(OS, DumpOptions);
138 OS << '\n'
139 << format(" dir_index: %" PRIu64 "\n", FileEntry.DirIdx);
140 if (ContentTypes.HasMD5)
141 OS << " md5_checksum: " << FileEntry.Checksum.digest() << '\n';
142 if (ContentTypes.HasModTime)
143 OS << format(" mod_time: 0x%8.8" PRIx64 "\n", FileEntry.ModTime);
144 if (ContentTypes.HasLength)
145 OS << format(" length: 0x%8.8" PRIx64 "\n", FileEntry.Length);
146 if (ContentTypes.HasSource) {
147 OS << " source: ";
148 FileEntry.Source.dump(OS, DumpOptions);
149 OS << '\n';
155 // Parse v2-v4 directory and file tables.
156 static void
157 parseV2DirFileTables(const DWARFDataExtractor &DebugLineData,
158 uint64_t *OffsetPtr, uint64_t EndPrologueOffset,
159 DWARFDebugLine::ContentTypeTracker &ContentTypes,
160 std::vector<DWARFFormValue> &IncludeDirectories,
161 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
162 while (*OffsetPtr < EndPrologueOffset) {
163 StringRef S = DebugLineData.getCStrRef(OffsetPtr);
164 if (S.empty())
165 break;
166 DWARFFormValue Dir =
167 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, S.data());
168 IncludeDirectories.push_back(Dir);
171 while (*OffsetPtr < EndPrologueOffset) {
172 StringRef Name = DebugLineData.getCStrRef(OffsetPtr);
173 if (Name.empty())
174 break;
175 DWARFDebugLine::FileNameEntry FileEntry;
176 FileEntry.Name =
177 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name.data());
178 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
179 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
180 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
181 FileNames.push_back(FileEntry);
184 ContentTypes.HasModTime = true;
185 ContentTypes.HasLength = true;
188 // Parse v5 directory/file entry content descriptions.
189 // Returns the descriptors, or an error if we did not find a path or ran off
190 // the end of the prologue.
191 static llvm::Expected<ContentDescriptors>
192 parseV5EntryFormat(const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
193 uint64_t EndPrologueOffset,
194 DWARFDebugLine::ContentTypeTracker *ContentTypes) {
195 ContentDescriptors Descriptors;
196 int FormatCount = DebugLineData.getU8(OffsetPtr);
197 bool HasPath = false;
198 for (int I = 0; I != FormatCount; ++I) {
199 if (*OffsetPtr >= EndPrologueOffset)
200 return createStringError(
201 errc::invalid_argument,
202 "failed to parse entry content descriptions at offset "
203 "0x%8.8" PRIx64
204 " because offset extends beyond the prologue end at offset "
205 "0x%8.8" PRIx64,
206 *OffsetPtr, EndPrologueOffset);
207 ContentDescriptor Descriptor;
208 Descriptor.Type =
209 dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(OffsetPtr));
210 Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(OffsetPtr));
211 if (Descriptor.Type == dwarf::DW_LNCT_path)
212 HasPath = true;
213 if (ContentTypes)
214 ContentTypes->trackContentType(Descriptor.Type);
215 Descriptors.push_back(Descriptor);
218 if (!HasPath)
219 return createStringError(errc::invalid_argument,
220 "failed to parse entry content descriptions"
221 " because no path was found");
222 return Descriptors;
225 static Error
226 parseV5DirFileTables(const DWARFDataExtractor &DebugLineData,
227 uint64_t *OffsetPtr, uint64_t EndPrologueOffset,
228 const dwarf::FormParams &FormParams,
229 const DWARFContext &Ctx, const DWARFUnit *U,
230 DWARFDebugLine::ContentTypeTracker &ContentTypes,
231 std::vector<DWARFFormValue> &IncludeDirectories,
232 std::vector<DWARFDebugLine::FileNameEntry> &FileNames) {
233 // Get the directory entry description.
234 llvm::Expected<ContentDescriptors> DirDescriptors =
235 parseV5EntryFormat(DebugLineData, OffsetPtr, EndPrologueOffset, nullptr);
236 if (!DirDescriptors)
237 return DirDescriptors.takeError();
239 // Get the directory entries, according to the format described above.
240 int DirEntryCount = DebugLineData.getU8(OffsetPtr);
241 for (int I = 0; I != DirEntryCount; ++I) {
242 if (*OffsetPtr >= EndPrologueOffset)
243 return createStringError(
244 errc::invalid_argument,
245 "failed to parse directory entry at offset "
246 "0x%8.8" PRIx64
247 " because offset extends beyond the prologue end at offset "
248 "0x%8.8" PRIx64,
249 *OffsetPtr, EndPrologueOffset);
250 for (auto Descriptor : *DirDescriptors) {
251 DWARFFormValue Value(Descriptor.Form);
252 switch (Descriptor.Type) {
253 case DW_LNCT_path:
254 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
255 return createStringError(errc::invalid_argument,
256 "failed to parse directory entry because "
257 "extracting the form value failed.");
258 IncludeDirectories.push_back(Value);
259 break;
260 default:
261 if (!Value.skipValue(DebugLineData, OffsetPtr, FormParams))
262 return createStringError(errc::invalid_argument,
263 "failed to parse directory entry because "
264 "skipping the form value failed.");
269 // Get the file entry description.
270 llvm::Expected<ContentDescriptors> FileDescriptors = parseV5EntryFormat(
271 DebugLineData, OffsetPtr, EndPrologueOffset, &ContentTypes);
272 if (!FileDescriptors)
273 return FileDescriptors.takeError();
275 // Get the file entries, according to the format described above.
276 int FileEntryCount = DebugLineData.getU8(OffsetPtr);
277 for (int I = 0; I != FileEntryCount; ++I) {
278 if (*OffsetPtr >= EndPrologueOffset)
279 return createStringError(
280 errc::invalid_argument,
281 "failed to parse file entry at offset "
282 "0x%8.8" PRIx64
283 " because offset extends beyond the prologue end at offset "
284 "0x%8.8" PRIx64,
285 *OffsetPtr, EndPrologueOffset);
286 DWARFDebugLine::FileNameEntry FileEntry;
287 for (auto Descriptor : *FileDescriptors) {
288 DWARFFormValue Value(Descriptor.Form);
289 if (!Value.extractValue(DebugLineData, OffsetPtr, FormParams, &Ctx, U))
290 return createStringError(errc::invalid_argument,
291 "failed to parse file entry because "
292 "extracting the form value failed.");
293 switch (Descriptor.Type) {
294 case DW_LNCT_path:
295 FileEntry.Name = Value;
296 break;
297 case DW_LNCT_LLVM_source:
298 FileEntry.Source = Value;
299 break;
300 case DW_LNCT_directory_index:
301 FileEntry.DirIdx = Value.getAsUnsignedConstant().getValue();
302 break;
303 case DW_LNCT_timestamp:
304 FileEntry.ModTime = Value.getAsUnsignedConstant().getValue();
305 break;
306 case DW_LNCT_size:
307 FileEntry.Length = Value.getAsUnsignedConstant().getValue();
308 break;
309 case DW_LNCT_MD5:
310 if (!Value.getAsBlock() || Value.getAsBlock().getValue().size() != 16)
311 return createStringError(
312 errc::invalid_argument,
313 "failed to parse file entry because the MD5 hash is invalid");
314 std::uninitialized_copy_n(Value.getAsBlock().getValue().begin(), 16,
315 FileEntry.Checksum.Bytes.begin());
316 break;
317 default:
318 break;
321 FileNames.push_back(FileEntry);
323 return Error::success();
326 Error DWARFDebugLine::Prologue::parse(const DWARFDataExtractor &DebugLineData,
327 uint64_t *OffsetPtr,
328 const DWARFContext &Ctx,
329 const DWARFUnit *U) {
330 const uint64_t PrologueOffset = *OffsetPtr;
332 clear();
333 TotalLength = DebugLineData.getRelocatedValue(4, OffsetPtr);
334 if (TotalLength == dwarf::DW_LENGTH_DWARF64) {
335 FormParams.Format = dwarf::DWARF64;
336 TotalLength = DebugLineData.getU64(OffsetPtr);
337 } else if (TotalLength >= dwarf::DW_LENGTH_lo_reserved) {
338 return createStringError(errc::invalid_argument,
339 "parsing line table prologue at offset 0x%8.8" PRIx64
340 " unsupported reserved unit length found of value 0x%8.8" PRIx64,
341 PrologueOffset, TotalLength);
343 FormParams.Version = DebugLineData.getU16(OffsetPtr);
344 if (getVersion() < 2)
345 return createStringError(errc::not_supported,
346 "parsing line table prologue at offset 0x%8.8" PRIx64
347 " found unsupported version 0x%2.2" PRIx16,
348 PrologueOffset, getVersion());
350 if (getVersion() >= 5) {
351 FormParams.AddrSize = DebugLineData.getU8(OffsetPtr);
352 assert((DebugLineData.getAddressSize() == 0 ||
353 DebugLineData.getAddressSize() == getAddressSize()) &&
354 "Line table header and data extractor disagree");
355 SegSelectorSize = DebugLineData.getU8(OffsetPtr);
358 PrologueLength =
359 DebugLineData.getRelocatedValue(sizeofPrologueLength(), OffsetPtr);
360 const uint64_t EndPrologueOffset = PrologueLength + *OffsetPtr;
361 MinInstLength = DebugLineData.getU8(OffsetPtr);
362 if (getVersion() >= 4)
363 MaxOpsPerInst = DebugLineData.getU8(OffsetPtr);
364 DefaultIsStmt = DebugLineData.getU8(OffsetPtr);
365 LineBase = DebugLineData.getU8(OffsetPtr);
366 LineRange = DebugLineData.getU8(OffsetPtr);
367 OpcodeBase = DebugLineData.getU8(OffsetPtr);
369 StandardOpcodeLengths.reserve(OpcodeBase - 1);
370 for (uint32_t I = 1; I < OpcodeBase; ++I) {
371 uint8_t OpLen = DebugLineData.getU8(OffsetPtr);
372 StandardOpcodeLengths.push_back(OpLen);
375 if (getVersion() >= 5) {
376 if (Error e = parseV5DirFileTables(
377 DebugLineData, OffsetPtr, EndPrologueOffset, FormParams, Ctx, U,
378 ContentTypes, IncludeDirectories, FileNames)) {
379 return joinErrors(
380 createStringError(
381 errc::invalid_argument,
382 "parsing line table prologue at 0x%8.8" PRIx64
383 " found an invalid directory or file table description at"
384 " 0x%8.8" PRIx64,
385 PrologueOffset, *OffsetPtr),
386 std::move(e));
388 } else
389 parseV2DirFileTables(DebugLineData, OffsetPtr, EndPrologueOffset,
390 ContentTypes, IncludeDirectories, FileNames);
392 if (*OffsetPtr != EndPrologueOffset)
393 return createStringError(errc::invalid_argument,
394 "parsing line table prologue at 0x%8.8" PRIx64
395 " should have ended at 0x%8.8" PRIx64
396 " but it ended at 0x%8.8" PRIx64,
397 PrologueOffset, EndPrologueOffset, *OffsetPtr);
398 return Error::success();
401 DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); }
403 void DWARFDebugLine::Row::postAppend() {
404 Discriminator = 0;
405 BasicBlock = false;
406 PrologueEnd = false;
407 EpilogueBegin = false;
410 void DWARFDebugLine::Row::reset(bool DefaultIsStmt) {
411 Address.Address = 0;
412 Address.SectionIndex = object::SectionedAddress::UndefSection;
413 Line = 1;
414 Column = 0;
415 File = 1;
416 Isa = 0;
417 Discriminator = 0;
418 IsStmt = DefaultIsStmt;
419 BasicBlock = false;
420 EndSequence = false;
421 PrologueEnd = false;
422 EpilogueBegin = false;
425 void DWARFDebugLine::Row::dumpTableHeader(raw_ostream &OS) {
426 OS << "Address Line Column File ISA Discriminator Flags\n"
427 << "------------------ ------ ------ ------ --- ------------- "
428 "-------------\n";
431 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
432 OS << format("0x%16.16" PRIx64 " %6u %6u", Address.Address, Line, Column)
433 << format(" %6u %3u %13u ", File, Isa, Discriminator)
434 << (IsStmt ? " is_stmt" : "") << (BasicBlock ? " basic_block" : "")
435 << (PrologueEnd ? " prologue_end" : "")
436 << (EpilogueBegin ? " epilogue_begin" : "")
437 << (EndSequence ? " end_sequence" : "") << '\n';
440 DWARFDebugLine::Sequence::Sequence() { reset(); }
442 void DWARFDebugLine::Sequence::reset() {
443 LowPC = 0;
444 HighPC = 0;
445 SectionIndex = object::SectionedAddress::UndefSection;
446 FirstRowIndex = 0;
447 LastRowIndex = 0;
448 Empty = true;
451 DWARFDebugLine::LineTable::LineTable() { clear(); }
453 void DWARFDebugLine::LineTable::dump(raw_ostream &OS,
454 DIDumpOptions DumpOptions) const {
455 Prologue.dump(OS, DumpOptions);
456 OS << '\n';
458 if (!Rows.empty()) {
459 Row::dumpTableHeader(OS);
460 for (const Row &R : Rows) {
461 R.dump(OS);
466 void DWARFDebugLine::LineTable::clear() {
467 Prologue.clear();
468 Rows.clear();
469 Sequences.clear();
472 DWARFDebugLine::ParsingState::ParsingState(struct LineTable *LT)
473 : LineTable(LT) {
474 resetRowAndSequence();
477 void DWARFDebugLine::ParsingState::resetRowAndSequence() {
478 Row.reset(LineTable->Prologue.DefaultIsStmt);
479 Sequence.reset();
482 void DWARFDebugLine::ParsingState::appendRowToMatrix() {
483 unsigned RowNumber = LineTable->Rows.size();
484 if (Sequence.Empty) {
485 // Record the beginning of instruction sequence.
486 Sequence.Empty = false;
487 Sequence.LowPC = Row.Address.Address;
488 Sequence.FirstRowIndex = RowNumber;
490 LineTable->appendRow(Row);
491 if (Row.EndSequence) {
492 // Record the end of instruction sequence.
493 Sequence.HighPC = Row.Address.Address;
494 Sequence.LastRowIndex = RowNumber + 1;
495 Sequence.SectionIndex = Row.Address.SectionIndex;
496 if (Sequence.isValid())
497 LineTable->appendSequence(Sequence);
498 Sequence.reset();
500 Row.postAppend();
503 const DWARFDebugLine::LineTable *
504 DWARFDebugLine::getLineTable(uint64_t Offset) const {
505 LineTableConstIter Pos = LineTableMap.find(Offset);
506 if (Pos != LineTableMap.end())
507 return &Pos->second;
508 return nullptr;
511 Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::getOrParseLineTable(
512 DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx,
513 const DWARFUnit *U, std::function<void(Error)> RecoverableErrorCallback) {
514 if (!DebugLineData.isValidOffset(Offset))
515 return createStringError(errc::invalid_argument, "offset 0x%8.8" PRIx64
516 " is not a valid debug line section offset",
517 Offset);
519 std::pair<LineTableIter, bool> Pos =
520 LineTableMap.insert(LineTableMapTy::value_type(Offset, LineTable()));
521 LineTable *LT = &Pos.first->second;
522 if (Pos.second) {
523 if (Error Err =
524 LT->parse(DebugLineData, &Offset, Ctx, U, RecoverableErrorCallback))
525 return std::move(Err);
526 return LT;
528 return LT;
531 Error DWARFDebugLine::LineTable::parse(
532 DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr,
533 const DWARFContext &Ctx, const DWARFUnit *U,
534 std::function<void(Error)> RecoverableErrorCallback, raw_ostream *OS) {
535 const uint64_t DebugLineOffset = *OffsetPtr;
537 clear();
539 Error PrologueErr = Prologue.parse(DebugLineData, OffsetPtr, Ctx, U);
541 if (OS) {
542 // The presence of OS signals verbose dumping.
543 DIDumpOptions DumpOptions;
544 DumpOptions.Verbose = true;
545 Prologue.dump(*OS, DumpOptions);
548 if (PrologueErr)
549 return PrologueErr;
551 const uint64_t EndOffset =
552 DebugLineOffset + Prologue.TotalLength + Prologue.sizeofTotalLength();
554 // See if we should tell the data extractor the address size.
555 if (DebugLineData.getAddressSize() == 0)
556 DebugLineData.setAddressSize(Prologue.getAddressSize());
557 else
558 assert(Prologue.getAddressSize() == 0 ||
559 Prologue.getAddressSize() == DebugLineData.getAddressSize());
561 ParsingState State(this);
563 while (*OffsetPtr < EndOffset) {
564 if (OS)
565 *OS << format("0x%08.08" PRIx64 ": ", *OffsetPtr);
567 uint8_t Opcode = DebugLineData.getU8(OffsetPtr);
569 if (OS)
570 *OS << format("%02.02" PRIx8 " ", Opcode);
572 if (Opcode == 0) {
573 // Extended Opcodes always start with a zero opcode followed by
574 // a uleb128 length so you can skip ones you don't know about
575 uint64_t Len = DebugLineData.getULEB128(OffsetPtr);
576 uint64_t ExtOffset = *OffsetPtr;
578 // Tolerate zero-length; assume length is correct and soldier on.
579 if (Len == 0) {
580 if (OS)
581 *OS << "Badly formed extended line op (length 0)\n";
582 continue;
585 uint8_t SubOpcode = DebugLineData.getU8(OffsetPtr);
586 if (OS)
587 *OS << LNExtendedString(SubOpcode);
588 switch (SubOpcode) {
589 case DW_LNE_end_sequence:
590 // Set the end_sequence register of the state machine to true and
591 // append a row to the matrix using the current values of the
592 // state-machine registers. Then reset the registers to the initial
593 // values specified above. Every statement program sequence must end
594 // with a DW_LNE_end_sequence instruction which creates a row whose
595 // address is that of the byte after the last target machine instruction
596 // of the sequence.
597 State.Row.EndSequence = true;
598 State.appendRowToMatrix();
599 if (OS) {
600 *OS << "\n";
601 OS->indent(12);
602 State.Row.dump(*OS);
604 State.resetRowAndSequence();
605 break;
607 case DW_LNE_set_address:
608 // Takes a single relocatable address as an operand. The size of the
609 // operand is the size appropriate to hold an address on the target
610 // machine. Set the address register to the value given by the
611 // relocatable address. All of the other statement program opcodes
612 // that affect the address register add a delta to it. This instruction
613 // stores a relocatable value into it instead.
615 // Make sure the extractor knows the address size. If not, infer it
616 // from the size of the operand.
617 if (DebugLineData.getAddressSize() == 0)
618 DebugLineData.setAddressSize(Len - 1);
619 else if (DebugLineData.getAddressSize() != Len - 1) {
620 return createStringError(errc::invalid_argument,
621 "mismatching address size at offset 0x%8.8" PRIx64
622 " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64,
623 ExtOffset, DebugLineData.getAddressSize(),
624 Len - 1);
626 State.Row.Address.Address = DebugLineData.getRelocatedAddress(
627 OffsetPtr, &State.Row.Address.SectionIndex);
628 if (OS)
629 *OS << format(" (0x%16.16" PRIx64 ")", State.Row.Address.Address);
630 break;
632 case DW_LNE_define_file:
633 // Takes 4 arguments. The first is a null terminated string containing
634 // a source file name. The second is an unsigned LEB128 number
635 // representing the directory index of the directory in which the file
636 // was found. The third is an unsigned LEB128 number representing the
637 // time of last modification of the file. The fourth is an unsigned
638 // LEB128 number representing the length in bytes of the file. The time
639 // and length fields may contain LEB128(0) if the information is not
640 // available.
642 // The directory index represents an entry in the include_directories
643 // section of the statement program prologue. The index is LEB128(0)
644 // if the file was found in the current directory of the compilation,
645 // LEB128(1) if it was found in the first directory in the
646 // include_directories section, and so on. The directory index is
647 // ignored for file names that represent full path names.
649 // The files are numbered, starting at 1, in the order in which they
650 // appear; the names in the prologue come before names defined by
651 // the DW_LNE_define_file instruction. These numbers are used in the
652 // the file register of the state machine.
654 FileNameEntry FileEntry;
655 const char *Name = DebugLineData.getCStr(OffsetPtr);
656 FileEntry.Name =
657 DWARFFormValue::createFromPValue(dwarf::DW_FORM_string, Name);
658 FileEntry.DirIdx = DebugLineData.getULEB128(OffsetPtr);
659 FileEntry.ModTime = DebugLineData.getULEB128(OffsetPtr);
660 FileEntry.Length = DebugLineData.getULEB128(OffsetPtr);
661 Prologue.FileNames.push_back(FileEntry);
662 if (OS)
663 *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time="
664 << format("(0x%16.16" PRIx64 ")", FileEntry.ModTime)
665 << ", length=" << FileEntry.Length << ")";
667 break;
669 case DW_LNE_set_discriminator:
670 State.Row.Discriminator = DebugLineData.getULEB128(OffsetPtr);
671 if (OS)
672 *OS << " (" << State.Row.Discriminator << ")";
673 break;
675 default:
676 if (OS)
677 *OS << format("Unrecognized extended op 0x%02.02" PRIx8, SubOpcode)
678 << format(" length %" PRIx64, Len);
679 // Len doesn't include the zero opcode byte or the length itself, but
680 // it does include the sub_opcode, so we have to adjust for that.
681 (*OffsetPtr) += Len - 1;
682 break;
684 // Make sure the stated and parsed lengths are the same.
685 // Otherwise we have an unparseable line-number program.
686 if (*OffsetPtr - ExtOffset != Len)
687 return createStringError(errc::illegal_byte_sequence,
688 "unexpected line op length at offset 0x%8.8" PRIx64
689 " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64,
690 ExtOffset, Len, *OffsetPtr - ExtOffset);
691 } else if (Opcode < Prologue.OpcodeBase) {
692 if (OS)
693 *OS << LNStandardString(Opcode);
694 switch (Opcode) {
695 // Standard Opcodes
696 case DW_LNS_copy:
697 // Takes no arguments. Append a row to the matrix using the
698 // current values of the state-machine registers.
699 if (OS) {
700 *OS << "\n";
701 OS->indent(12);
702 State.Row.dump(*OS);
703 *OS << "\n";
705 State.appendRowToMatrix();
706 break;
708 case DW_LNS_advance_pc:
709 // Takes a single unsigned LEB128 operand, multiplies it by the
710 // min_inst_length field of the prologue, and adds the
711 // result to the address register of the state machine.
713 uint64_t AddrOffset =
714 DebugLineData.getULEB128(OffsetPtr) * Prologue.MinInstLength;
715 State.Row.Address.Address += AddrOffset;
716 if (OS)
717 *OS << " (" << AddrOffset << ")";
719 break;
721 case DW_LNS_advance_line:
722 // Takes a single signed LEB128 operand and adds that value to
723 // the line register of the state machine.
724 State.Row.Line += DebugLineData.getSLEB128(OffsetPtr);
725 if (OS)
726 *OS << " (" << State.Row.Line << ")";
727 break;
729 case DW_LNS_set_file:
730 // Takes a single unsigned LEB128 operand and stores it in the file
731 // register of the state machine.
732 State.Row.File = DebugLineData.getULEB128(OffsetPtr);
733 if (OS)
734 *OS << " (" << State.Row.File << ")";
735 break;
737 case DW_LNS_set_column:
738 // Takes a single unsigned LEB128 operand and stores it in the
739 // column register of the state machine.
740 State.Row.Column = DebugLineData.getULEB128(OffsetPtr);
741 if (OS)
742 *OS << " (" << State.Row.Column << ")";
743 break;
745 case DW_LNS_negate_stmt:
746 // Takes no arguments. Set the is_stmt register of the state
747 // machine to the logical negation of its current value.
748 State.Row.IsStmt = !State.Row.IsStmt;
749 break;
751 case DW_LNS_set_basic_block:
752 // Takes no arguments. Set the basic_block register of the
753 // state machine to true
754 State.Row.BasicBlock = true;
755 break;
757 case DW_LNS_const_add_pc:
758 // Takes no arguments. Add to the address register of the state
759 // machine the address increment value corresponding to special
760 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
761 // when the statement program needs to advance the address by a
762 // small amount, it can use a single special opcode, which occupies
763 // a single byte. When it needs to advance the address by up to
764 // twice the range of the last special opcode, it can use
765 // DW_LNS_const_add_pc followed by a special opcode, for a total
766 // of two bytes. Only if it needs to advance the address by more
767 // than twice that range will it need to use both DW_LNS_advance_pc
768 // and a special opcode, requiring three or more bytes.
770 uint8_t AdjustOpcode = 255 - Prologue.OpcodeBase;
771 uint64_t AddrOffset =
772 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
773 State.Row.Address.Address += AddrOffset;
774 if (OS)
776 << format(" (0x%16.16" PRIx64 ")", AddrOffset);
778 break;
780 case DW_LNS_fixed_advance_pc:
781 // Takes a single uhalf operand. Add to the address register of
782 // the state machine the value of the (unencoded) operand. This
783 // is the only extended opcode that takes an argument that is not
784 // a variable length number. The motivation for DW_LNS_fixed_advance_pc
785 // is this: existing assemblers cannot emit DW_LNS_advance_pc or
786 // special opcodes because they cannot encode LEB128 numbers or
787 // judge when the computation of a special opcode overflows and
788 // requires the use of DW_LNS_advance_pc. Such assemblers, however,
789 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
791 uint16_t PCOffset = DebugLineData.getRelocatedValue(2, OffsetPtr);
792 State.Row.Address.Address += PCOffset;
793 if (OS)
795 << format(" (0x%4.4" PRIx16 ")", PCOffset);
797 break;
799 case DW_LNS_set_prologue_end:
800 // Takes no arguments. Set the prologue_end register of the
801 // state machine to true
802 State.Row.PrologueEnd = true;
803 break;
805 case DW_LNS_set_epilogue_begin:
806 // Takes no arguments. Set the basic_block register of the
807 // state machine to true
808 State.Row.EpilogueBegin = true;
809 break;
811 case DW_LNS_set_isa:
812 // Takes a single unsigned LEB128 operand and stores it in the
813 // column register of the state machine.
814 State.Row.Isa = DebugLineData.getULEB128(OffsetPtr);
815 if (OS)
816 *OS << " (" << State.Row.Isa << ")";
817 break;
819 default:
820 // Handle any unknown standard opcodes here. We know the lengths
821 // of such opcodes because they are specified in the prologue
822 // as a multiple of LEB128 operands for each opcode.
824 assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size());
825 uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1];
826 for (uint8_t I = 0; I < OpcodeLength; ++I) {
827 uint64_t Value = DebugLineData.getULEB128(OffsetPtr);
828 if (OS)
829 *OS << format("Skipping ULEB128 value: 0x%16.16" PRIx64 ")\n",
830 Value);
833 break;
835 } else {
836 // Special Opcodes
838 // A special opcode value is chosen based on the amount that needs
839 // to be added to the line and address registers. The maximum line
840 // increment for a special opcode is the value of the line_base
841 // field in the header, plus the value of the line_range field,
842 // minus 1 (line base + line range - 1). If the desired line
843 // increment is greater than the maximum line increment, a standard
844 // opcode must be used instead of a special opcode. The "address
845 // advance" is calculated by dividing the desired address increment
846 // by the minimum_instruction_length field from the header. The
847 // special opcode is then calculated using the following formula:
849 // opcode = (desired line increment - line_base) +
850 // (line_range * address advance) + opcode_base
852 // If the resulting opcode is greater than 255, a standard opcode
853 // must be used instead.
855 // To decode a special opcode, subtract the opcode_base from the
856 // opcode itself to give the adjusted opcode. The amount to
857 // increment the address register is the result of the adjusted
858 // opcode divided by the line_range multiplied by the
859 // minimum_instruction_length field from the header. That is:
861 // address increment = (adjusted opcode / line_range) *
862 // minimum_instruction_length
864 // The amount to increment the line register is the line_base plus
865 // the result of the adjusted opcode modulo the line_range. That is:
867 // line increment = line_base + (adjusted opcode % line_range)
869 uint8_t AdjustOpcode = Opcode - Prologue.OpcodeBase;
870 uint64_t AddrOffset =
871 (AdjustOpcode / Prologue.LineRange) * Prologue.MinInstLength;
872 int32_t LineOffset =
873 Prologue.LineBase + (AdjustOpcode % Prologue.LineRange);
874 State.Row.Line += LineOffset;
875 State.Row.Address.Address += AddrOffset;
877 if (OS) {
878 *OS << "address += " << AddrOffset << ", line += " << LineOffset
879 << "\n";
880 OS->indent(12);
881 State.Row.dump(*OS);
884 State.appendRowToMatrix();
886 if(OS)
887 *OS << "\n";
890 if (!State.Sequence.Empty)
891 RecoverableErrorCallback(
892 createStringError(errc::illegal_byte_sequence,
893 "last sequence in debug line table is not terminated!"));
895 // Sort all sequences so that address lookup will work faster.
896 if (!Sequences.empty()) {
897 llvm::sort(Sequences, Sequence::orderByHighPC);
898 // Note: actually, instruction address ranges of sequences should not
899 // overlap (in shared objects and executables). If they do, the address
900 // lookup would still work, though, but result would be ambiguous.
901 // We don't report warning in this case. For example,
902 // sometimes .so compiled from multiple object files contains a few
903 // rudimentary sequences for address ranges [0x0, 0xsomething).
906 return Error::success();
909 uint32_t DWARFDebugLine::LineTable::findRowInSeq(
910 const DWARFDebugLine::Sequence &Seq,
911 object::SectionedAddress Address) const {
912 if (!Seq.containsPC(Address))
913 return UnknownRowIndex;
914 assert(Seq.SectionIndex == Address.SectionIndex);
915 // In some cases, e.g. first instruction in a function, the compiler generates
916 // two entries, both with the same address. We want the last one.
918 // In general we want a non-empty range: the last row whose address is less
919 // than or equal to Address. This can be computed as upper_bound - 1.
920 DWARFDebugLine::Row Row;
921 Row.Address = Address;
922 RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex;
923 RowIter LastRow = Rows.begin() + Seq.LastRowIndex;
924 assert(FirstRow->Address.Address <= Row.Address.Address &&
925 Row.Address.Address < LastRow[-1].Address.Address);
926 RowIter RowPos = std::upper_bound(FirstRow + 1, LastRow - 1, Row,
927 DWARFDebugLine::Row::orderByAddress) -
929 assert(Seq.SectionIndex == RowPos->Address.SectionIndex);
930 return RowPos - Rows.begin();
933 uint32_t DWARFDebugLine::LineTable::lookupAddress(
934 object::SectionedAddress Address) const {
936 // Search for relocatable addresses
937 uint32_t Result = lookupAddressImpl(Address);
939 if (Result != UnknownRowIndex ||
940 Address.SectionIndex == object::SectionedAddress::UndefSection)
941 return Result;
943 // Search for absolute addresses
944 Address.SectionIndex = object::SectionedAddress::UndefSection;
945 return lookupAddressImpl(Address);
948 uint32_t DWARFDebugLine::LineTable::lookupAddressImpl(
949 object::SectionedAddress Address) const {
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 It = llvm::upper_bound(Sequences, Sequence,
955 DWARFDebugLine::Sequence::orderByHighPC);
956 if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex)
957 return UnknownRowIndex;
958 return findRowInSeq(*It, Address);
961 bool DWARFDebugLine::LineTable::lookupAddressRange(
962 object::SectionedAddress Address, uint64_t Size,
963 std::vector<uint32_t> &Result) const {
965 // Search for relocatable addresses
966 if (lookupAddressRangeImpl(Address, Size, Result))
967 return true;
969 if (Address.SectionIndex == object::SectionedAddress::UndefSection)
970 return false;
972 // Search for absolute addresses
973 Address.SectionIndex = object::SectionedAddress::UndefSection;
974 return lookupAddressRangeImpl(Address, Size, Result);
977 bool DWARFDebugLine::LineTable::lookupAddressRangeImpl(
978 object::SectionedAddress Address, uint64_t Size,
979 std::vector<uint32_t> &Result) const {
980 if (Sequences.empty())
981 return false;
982 uint64_t EndAddr = Address.Address + Size;
983 // First, find an instruction sequence containing the given address.
984 DWARFDebugLine::Sequence Sequence;
985 Sequence.SectionIndex = Address.SectionIndex;
986 Sequence.HighPC = Address.Address;
987 SequenceIter LastSeq = Sequences.end();
988 SequenceIter SeqPos = llvm::upper_bound(
989 Sequences, Sequence, DWARFDebugLine::Sequence::orderByHighPC);
990 if (SeqPos == LastSeq || !SeqPos->containsPC(Address))
991 return false;
993 SequenceIter StartPos = SeqPos;
995 // Add the rows from the first sequence to the vector, starting with the
996 // index we just calculated
998 while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) {
999 const DWARFDebugLine::Sequence &CurSeq = *SeqPos;
1000 // For the first sequence, we need to find which row in the sequence is the
1001 // first in our range.
1002 uint32_t FirstRowIndex = CurSeq.FirstRowIndex;
1003 if (SeqPos == StartPos)
1004 FirstRowIndex = findRowInSeq(CurSeq, Address);
1006 // Figure out the last row in the range.
1007 uint32_t LastRowIndex =
1008 findRowInSeq(CurSeq, {EndAddr - 1, Address.SectionIndex});
1009 if (LastRowIndex == UnknownRowIndex)
1010 LastRowIndex = CurSeq.LastRowIndex - 1;
1012 assert(FirstRowIndex != UnknownRowIndex);
1013 assert(LastRowIndex != UnknownRowIndex);
1015 for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) {
1016 Result.push_back(I);
1019 ++SeqPos;
1022 return true;
1025 Optional<StringRef> DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex,
1026 FileLineInfoKind Kind) const {
1027 if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex))
1028 return None;
1029 const FileNameEntry &Entry = Prologue.getFileNameEntry(FileIndex);
1030 if (Optional<const char *> source = Entry.Source.getAsCString())
1031 return StringRef(*source);
1032 return None;
1035 static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) {
1036 // Debug info can contain paths from any OS, not necessarily
1037 // an OS we're currently running on. Moreover different compilation units can
1038 // be compiled on different operating systems and linked together later.
1039 return sys::path::is_absolute(Path, sys::path::Style::posix) ||
1040 sys::path::is_absolute(Path, sys::path::Style::windows);
1043 bool DWARFDebugLine::Prologue::getFileNameByIndex(
1044 uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind,
1045 std::string &Result, sys::path::Style Style) const {
1046 if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex))
1047 return false;
1048 const FileNameEntry &Entry = getFileNameEntry(FileIndex);
1049 StringRef FileName = Entry.Name.getAsCString().getValue();
1050 if (Kind != FileLineInfoKind::AbsoluteFilePath ||
1051 isPathAbsoluteOnWindowsOrPosix(FileName)) {
1052 Result = FileName;
1053 return true;
1056 SmallString<16> FilePath;
1057 StringRef IncludeDir;
1058 // Be defensive about the contents of Entry.
1059 if (getVersion() >= 5) {
1060 if (Entry.DirIdx < IncludeDirectories.size())
1061 IncludeDir = IncludeDirectories[Entry.DirIdx].getAsCString().getValue();
1062 } else {
1063 if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size())
1064 IncludeDir =
1065 IncludeDirectories[Entry.DirIdx - 1].getAsCString().getValue();
1067 // We may still need to append compilation directory of compile unit.
1068 // We know that FileName is not absolute, the only way to have an
1069 // absolute path at this point would be if IncludeDir is absolute.
1070 if (!CompDir.empty() && !isPathAbsoluteOnWindowsOrPosix(IncludeDir))
1071 sys::path::append(FilePath, Style, CompDir);
1074 // sys::path::append skips empty strings.
1075 sys::path::append(FilePath, Style, IncludeDir, FileName);
1076 Result = FilePath.str();
1077 return true;
1080 bool DWARFDebugLine::LineTable::getFileLineInfoForAddress(
1081 object::SectionedAddress Address, const char *CompDir,
1082 FileLineInfoKind Kind, DILineInfo &Result) const {
1083 // Get the index of row we're looking for in the line table.
1084 uint32_t RowIndex = lookupAddress(Address);
1085 if (RowIndex == -1U)
1086 return false;
1087 // Take file number and line/column from the row.
1088 const auto &Row = Rows[RowIndex];
1089 if (!getFileNameByIndex(Row.File, CompDir, Kind, Result.FileName))
1090 return false;
1091 Result.Line = Row.Line;
1092 Result.Column = Row.Column;
1093 Result.Discriminator = Row.Discriminator;
1094 Result.Source = getSourceByIndex(Row.File, Kind);
1095 return true;
1098 // We want to supply the Unit associated with a .debug_line[.dwo] table when
1099 // we dump it, if possible, but still dump the table even if there isn't a Unit.
1100 // Therefore, collect up handles on all the Units that point into the
1101 // line-table section.
1102 static DWARFDebugLine::SectionParser::LineToUnitMap
1103 buildLineToUnitMap(DWARFDebugLine::SectionParser::cu_range CUs,
1104 DWARFDebugLine::SectionParser::tu_range TUs) {
1105 DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit;
1106 for (const auto &CU : CUs)
1107 if (auto CUDIE = CU->getUnitDIE())
1108 if (auto StmtOffset = toSectionOffset(CUDIE.find(DW_AT_stmt_list)))
1109 LineToUnit.insert(std::make_pair(*StmtOffset, &*CU));
1110 for (const auto &TU : TUs)
1111 if (auto TUDIE = TU->getUnitDIE())
1112 if (auto StmtOffset = toSectionOffset(TUDIE.find(DW_AT_stmt_list)))
1113 LineToUnit.insert(std::make_pair(*StmtOffset, &*TU));
1114 return LineToUnit;
1117 DWARFDebugLine::SectionParser::SectionParser(DWARFDataExtractor &Data,
1118 const DWARFContext &C,
1119 cu_range CUs, tu_range TUs)
1120 : DebugLineData(Data), Context(C) {
1121 LineToUnit = buildLineToUnitMap(CUs, TUs);
1122 if (!DebugLineData.isValidOffset(Offset))
1123 Done = true;
1126 bool DWARFDebugLine::Prologue::totalLengthIsValid() const {
1127 return TotalLength == dwarf::DW_LENGTH_DWARF64 ||
1128 TotalLength < dwarf::DW_LENGTH_lo_reserved;
1131 DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext(
1132 function_ref<void(Error)> RecoverableErrorCallback,
1133 function_ref<void(Error)> UnrecoverableErrorCallback, raw_ostream *OS) {
1134 assert(DebugLineData.isValidOffset(Offset) &&
1135 "parsing should have terminated");
1136 DWARFUnit *U = prepareToParse(Offset);
1137 uint64_t OldOffset = Offset;
1138 LineTable LT;
1139 if (Error Err = LT.parse(DebugLineData, &Offset, Context, U,
1140 RecoverableErrorCallback, OS))
1141 UnrecoverableErrorCallback(std::move(Err));
1142 moveToNextTable(OldOffset, LT.Prologue);
1143 return LT;
1146 void DWARFDebugLine::SectionParser::skip(
1147 function_ref<void(Error)> ErrorCallback) {
1148 assert(DebugLineData.isValidOffset(Offset) &&
1149 "parsing should have terminated");
1150 DWARFUnit *U = prepareToParse(Offset);
1151 uint64_t OldOffset = Offset;
1152 LineTable LT;
1153 if (Error Err = LT.Prologue.parse(DebugLineData, &Offset, Context, U))
1154 ErrorCallback(std::move(Err));
1155 moveToNextTable(OldOffset, LT.Prologue);
1158 DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) {
1159 DWARFUnit *U = nullptr;
1160 auto It = LineToUnit.find(Offset);
1161 if (It != LineToUnit.end())
1162 U = It->second;
1163 DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0);
1164 return U;
1167 void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset,
1168 const Prologue &P) {
1169 // If the length field is not valid, we don't know where the next table is, so
1170 // cannot continue to parse. Mark the parser as done, and leave the Offset
1171 // value as it currently is. This will be the end of the bad length field.
1172 if (!P.totalLengthIsValid()) {
1173 Done = true;
1174 return;
1177 Offset = OldOffset + P.TotalLength + P.sizeofTotalLength();
1178 if (!DebugLineData.isValidOffset(Offset)) {
1179 Done = true;