| 1 | //===- DWARFDebugLine.cpp -------------------------------------------------===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "llvm/DebugInfo/DWARF/DWARFDebugLine.h" |
| 10 | #include "llvm/ADT/SmallString.h" |
| 11 | #include "llvm/ADT/SmallVector.h" |
| 12 | #include "llvm/ADT/StringRef.h" |
| 13 | #include "llvm/BinaryFormat/Dwarf.h" |
| 14 | #include "llvm/DebugInfo/DWARF/DWARFDataExtractor.h" |
| 15 | #include "llvm/DebugInfo/DWARF/DWARFDie.h" |
| 16 | #include "llvm/DebugInfo/DWARF/DWARFFormValue.h" |
| 17 | #include "llvm/Support/Errc.h" |
| 18 | #include "llvm/Support/Format.h" |
| 19 | #include "llvm/Support/FormatVariadic.h" |
| 20 | #include "llvm/Support/raw_ostream.h" |
| 21 | #include <algorithm> |
| 22 | #include <cassert> |
| 23 | #include <cinttypes> |
| 24 | #include <cstdint> |
| 25 | #include <cstdio> |
| 26 | #include <utility> |
| 27 | |
| 28 | using namespace llvm; |
| 29 | using namespace dwarf; |
| 30 | |
| 31 | using FileLineInfoKind = DILineInfoSpecifier::FileLineInfoKind; |
| 32 | |
| 33 | namespace { |
| 34 | |
| 35 | struct ContentDescriptor { |
| 36 | dwarf::LineNumberEntryFormat Type; |
| 37 | dwarf::Form Form; |
| 38 | }; |
| 39 | |
| 40 | using ContentDescriptors = SmallVector<ContentDescriptor, 4>; |
| 41 | |
| 42 | } // end anonymous namespace |
| 43 | |
| 44 | static bool versionIsSupported(uint16_t Version) { |
| 45 | return Version >= 2 && Version <= 5; |
| 46 | } |
| 47 | |
| 48 | void DWARFDebugLine::ContentTypeTracker::trackContentType( |
| 49 | dwarf::LineNumberEntryFormat ContentType) { |
| 50 | switch (ContentType) { |
| 51 | case dwarf::DW_LNCT_timestamp: |
| 52 | HasModTime = true; |
| 53 | break; |
| 54 | case dwarf::DW_LNCT_size: |
| 55 | HasLength = true; |
| 56 | break; |
| 57 | case dwarf::DW_LNCT_MD5: |
| 58 | HasMD5 = true; |
| 59 | break; |
| 60 | case dwarf::DW_LNCT_LLVM_source: |
| 61 | HasSource = true; |
| 62 | break; |
| 63 | default: |
| 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. |
| 66 | break; |
| 67 | } |
| 68 | } |
| 69 | |
| 70 | DWARFDebugLine::Prologue::Prologue() { clear(); } |
| 71 | |
| 72 | bool DWARFDebugLine::Prologue::hasFileAtIndex(uint64_t FileIndex) const { |
| 73 | uint16_t DwarfVersion = getVersion(); |
| 74 | assert(DwarfVersion != 0 && |
| 75 | "line table prologue has no dwarf version information" ); |
| 76 | if (DwarfVersion >= 5) |
| 77 | return FileIndex < FileNames.size(); |
| 78 | return FileIndex != 0 && FileIndex <= FileNames.size(); |
| 79 | } |
| 80 | |
| 81 | std::optional<uint64_t> |
| 82 | DWARFDebugLine::Prologue::getLastValidFileIndex() const { |
| 83 | if (FileNames.empty()) |
| 84 | return std::nullopt; |
| 85 | uint16_t DwarfVersion = getVersion(); |
| 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) |
| 90 | return FileNames.size() - 1; |
| 91 | return FileNames.size(); |
| 92 | } |
| 93 | |
| 94 | const llvm::DWARFDebugLine::FileNameEntry & |
| 95 | DWARFDebugLine::Prologue::getFileNameEntry(uint64_t Index) const { |
| 96 | uint16_t DwarfVersion = getVersion(); |
| 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) |
| 101 | return FileNames[Index]; |
| 102 | return FileNames[Index - 1]; |
| 103 | } |
| 104 | |
| 105 | void DWARFDebugLine::Prologue::clear() { |
| 106 | TotalLength = PrologueLength = 0; |
| 107 | SegSelectorSize = 0; |
| 108 | MinInstLength = MaxOpsPerInst = DefaultIsStmt = LineBase = LineRange = 0; |
| 109 | OpcodeBase = 0; |
| 110 | FormParams = dwarf::FormParams({.Version: 0, .AddrSize: 0, .Format: DWARF32}); |
| 111 | ContentTypes = ContentTypeTracker(); |
| 112 | StandardOpcodeLengths.clear(); |
| 113 | IncludeDirectories.clear(); |
| 114 | FileNames.clear(); |
| 115 | } |
| 116 | |
| 117 | void DWARFDebugLine::Prologue::dump(raw_ostream &OS, |
| 118 | DIDumpOptions DumpOptions) const { |
| 119 | if (!totalLengthIsValid()) |
| 120 | return; |
| 121 | int OffsetDumpWidth = 2 * dwarf::getDwarfOffsetByteSize(Format: FormParams.Format); |
| 122 | OS << "Line table prologue:\n" |
| 123 | << format(Fmt: " total_length: 0x%0*" PRIx64 "\n" , Vals: OffsetDumpWidth, |
| 124 | Vals: TotalLength) |
| 125 | << " format: " << dwarf::FormatString(Format: FormParams.Format) << "\n" |
| 126 | << format(Fmt: " version: %u\n" , Vals: getVersion()); |
| 127 | if (!versionIsSupported(Version: getVersion())) |
| 128 | return; |
| 129 | if (getVersion() >= 5) |
| 130 | OS << format(Fmt: " address_size: %u\n" , Vals: getAddressSize()) |
| 131 | << format(Fmt: " seg_select_size: %u\n" , Vals: SegSelectorSize); |
| 132 | OS << format(Fmt: " prologue_length: 0x%0*" PRIx64 "\n" , Vals: OffsetDumpWidth, |
| 133 | Vals: PrologueLength) |
| 134 | << format(Fmt: " min_inst_length: %u\n" , Vals: MinInstLength) |
| 135 | << format(Fmt: getVersion() >= 4 ? "max_ops_per_inst: %u\n" : "" , Vals: MaxOpsPerInst) |
| 136 | << format(Fmt: " default_is_stmt: %u\n" , Vals: DefaultIsStmt) |
| 137 | << format(Fmt: " line_base: %i\n" , Vals: LineBase) |
| 138 | << format(Fmt: " line_range: %u\n" , Vals: LineRange) |
| 139 | << format(Fmt: " opcode_base: %u\n" , Vals: OpcodeBase); |
| 140 | |
| 141 | for (uint32_t I = 0; I != StandardOpcodeLengths.size(); ++I) |
| 142 | OS << formatv(Fmt: "standard_opcode_lengths[{0}] = {1}\n" , |
| 143 | Vals: static_cast<dwarf::LineNumberOps>(I + 1), |
| 144 | Vals: StandardOpcodeLengths[I]); |
| 145 | |
| 146 | if (!IncludeDirectories.empty()) { |
| 147 | // DWARF v5 starts directory indexes at 0. |
| 148 | uint32_t DirBase = getVersion() >= 5 ? 0 : 1; |
| 149 | for (uint32_t I = 0; I != IncludeDirectories.size(); ++I) { |
| 150 | OS << format(Fmt: "include_directories[%3u] = " , Vals: I + DirBase); |
| 151 | IncludeDirectories[I].dump(OS, DumpOpts: DumpOptions); |
| 152 | OS << '\n'; |
| 153 | } |
| 154 | } |
| 155 | |
| 156 | if (!FileNames.empty()) { |
| 157 | // DWARF v5 starts file indexes at 0. |
| 158 | uint32_t FileBase = getVersion() >= 5 ? 0 : 1; |
| 159 | for (uint32_t I = 0; I != FileNames.size(); ++I) { |
| 160 | const FileNameEntry &FileEntry = FileNames[I]; |
| 161 | OS << format(Fmt: "file_names[%3u]:\n" , Vals: I + FileBase); |
| 162 | OS << " name: " ; |
| 163 | FileEntry.Name.dump(OS, DumpOpts: DumpOptions); |
| 164 | OS << '\n' << format(Fmt: " dir_index: %" PRIu64 "\n" , Vals: FileEntry.DirIdx); |
| 165 | if (ContentTypes.HasMD5) |
| 166 | OS << " md5_checksum: " << FileEntry.Checksum.digest() << '\n'; |
| 167 | if (ContentTypes.HasModTime) |
| 168 | OS << format(Fmt: " mod_time: 0x%8.8" PRIx64 "\n" , Vals: FileEntry.ModTime); |
| 169 | if (ContentTypes.HasLength) |
| 170 | OS << format(Fmt: " length: 0x%8.8" PRIx64 "\n" , Vals: FileEntry.Length); |
| 171 | if (ContentTypes.HasSource) { |
| 172 | auto Source = FileEntry.Source.getAsCString(); |
| 173 | if (!Source) |
| 174 | consumeError(Err: Source.takeError()); |
| 175 | else if ((*Source)[0]) { |
| 176 | OS << " source: " ; |
| 177 | FileEntry.Source.dump(OS, DumpOpts: DumpOptions); |
| 178 | OS << '\n'; |
| 179 | } |
| 180 | } |
| 181 | } |
| 182 | } |
| 183 | } |
| 184 | |
| 185 | // Parse v2-v4 directory and file tables. |
| 186 | static Error |
| 187 | (const DWARFDataExtractor &DebugLineData, |
| 188 | uint64_t *OffsetPtr, |
| 189 | DWARFDebugLine::ContentTypeTracker &ContentTypes, |
| 190 | std::vector<DWARFFormValue> &IncludeDirectories, |
| 191 | std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { |
| 192 | while (true) { |
| 193 | Error Err = Error::success(); |
| 194 | StringRef S = DebugLineData.getCStrRef(OffsetPtr, Err: &Err); |
| 195 | if (Err) { |
| 196 | consumeError(Err: std::move(Err)); |
| 197 | return createStringError(EC: errc::invalid_argument, |
| 198 | S: "include directories table was not null " |
| 199 | "terminated before the end of the prologue" ); |
| 200 | } |
| 201 | if (S.empty()) |
| 202 | break; |
| 203 | DWARFFormValue Dir = |
| 204 | DWARFFormValue::createFromPValue(F: dwarf::DW_FORM_string, V: S.data()); |
| 205 | IncludeDirectories.push_back(x: Dir); |
| 206 | } |
| 207 | |
| 208 | ContentTypes.HasModTime = true; |
| 209 | ContentTypes.HasLength = true; |
| 210 | |
| 211 | while (true) { |
| 212 | Error Err = Error::success(); |
| 213 | StringRef Name = DebugLineData.getCStrRef(OffsetPtr, Err: &Err); |
| 214 | if (!Err && Name.empty()) |
| 215 | break; |
| 216 | |
| 217 | DWARFDebugLine::FileNameEntry FileEntry; |
| 218 | FileEntry.Name = |
| 219 | DWARFFormValue::createFromPValue(F: dwarf::DW_FORM_string, V: Name.data()); |
| 220 | FileEntry.DirIdx = DebugLineData.getULEB128(offset_ptr: OffsetPtr, Err: &Err); |
| 221 | FileEntry.ModTime = DebugLineData.getULEB128(offset_ptr: OffsetPtr, Err: &Err); |
| 222 | FileEntry.Length = DebugLineData.getULEB128(offset_ptr: OffsetPtr, Err: &Err); |
| 223 | |
| 224 | if (Err) { |
| 225 | consumeError(Err: std::move(Err)); |
| 226 | return createStringError( |
| 227 | EC: errc::invalid_argument, |
| 228 | S: "file names table was not null terminated before " |
| 229 | "the end of the prologue" ); |
| 230 | } |
| 231 | FileNames.push_back(x: FileEntry); |
| 232 | } |
| 233 | |
| 234 | return Error::success(); |
| 235 | } |
| 236 | |
| 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. |
| 240 | static llvm::Expected<ContentDescriptors> |
| 241 | (const DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, |
| 242 | DWARFDebugLine::ContentTypeTracker *ContentTypes) { |
| 243 | Error Err = Error::success(); |
| 244 | ContentDescriptors Descriptors; |
| 245 | int FormatCount = DebugLineData.getU8(offset_ptr: OffsetPtr, Err: &Err); |
| 246 | bool HasPath = false; |
| 247 | for (int I = 0; I != FormatCount && !Err; ++I) { |
| 248 | ContentDescriptor Descriptor; |
| 249 | Descriptor.Type = |
| 250 | dwarf::LineNumberEntryFormat(DebugLineData.getULEB128(offset_ptr: OffsetPtr, Err: &Err)); |
| 251 | Descriptor.Form = dwarf::Form(DebugLineData.getULEB128(offset_ptr: OffsetPtr, Err: &Err)); |
| 252 | if (Descriptor.Type == dwarf::DW_LNCT_path) |
| 253 | HasPath = true; |
| 254 | if (ContentTypes) |
| 255 | ContentTypes->trackContentType(ContentType: Descriptor.Type); |
| 256 | Descriptors.push_back(Elt: Descriptor); |
| 257 | } |
| 258 | |
| 259 | if (Err) |
| 260 | return createStringError(EC: errc::invalid_argument, |
| 261 | Fmt: "failed to parse entry content descriptors: %s" , |
| 262 | Vals: toString(E: std::move(Err)).c_str()); |
| 263 | |
| 264 | if (!HasPath) |
| 265 | return createStringError(EC: errc::invalid_argument, |
| 266 | S: "failed to parse entry content descriptions" |
| 267 | " because no path was found" ); |
| 268 | return Descriptors; |
| 269 | } |
| 270 | |
| 271 | static Error |
| 272 | (const DWARFDataExtractor &DebugLineData, |
| 273 | uint64_t *OffsetPtr, const dwarf::FormParams &FormParams, |
| 274 | const DWARFContext &Ctx, const DWARFUnit *U, |
| 275 | DWARFDebugLine::ContentTypeTracker &ContentTypes, |
| 276 | std::vector<DWARFFormValue> &IncludeDirectories, |
| 277 | std::vector<DWARFDebugLine::FileNameEntry> &FileNames) { |
| 278 | // Get the directory entry description. |
| 279 | llvm::Expected<ContentDescriptors> DirDescriptors = |
| 280 | parseV5EntryFormat(DebugLineData, OffsetPtr, ContentTypes: nullptr); |
| 281 | if (!DirDescriptors) |
| 282 | return DirDescriptors.takeError(); |
| 283 | |
| 284 | // Get the directory entries, according to the format described above. |
| 285 | uint64_t DirEntryCount = DebugLineData.getULEB128(offset_ptr: OffsetPtr); |
| 286 | for (uint64_t I = 0; I != DirEntryCount; ++I) { |
| 287 | for (auto Descriptor : *DirDescriptors) { |
| 288 | DWARFFormValue Value(Descriptor.Form); |
| 289 | switch (Descriptor.Type) { |
| 290 | case DW_LNCT_path: |
| 291 | if (!Value.extractValue(Data: DebugLineData, OffsetPtr, FormParams, Context: &Ctx, Unit: U)) |
| 292 | return createStringError(EC: errc::invalid_argument, |
| 293 | S: "failed to parse directory entry because " |
| 294 | "extracting the form value failed" ); |
| 295 | IncludeDirectories.push_back(x: Value); |
| 296 | break; |
| 297 | default: |
| 298 | if (!Value.skipValue(DebugInfoData: DebugLineData, OffsetPtr, Params: FormParams)) |
| 299 | return createStringError(EC: errc::invalid_argument, |
| 300 | S: "failed to parse directory entry because " |
| 301 | "skipping the form value failed" ); |
| 302 | } |
| 303 | } |
| 304 | } |
| 305 | |
| 306 | // Get the file entry description. |
| 307 | llvm::Expected<ContentDescriptors> FileDescriptors = |
| 308 | parseV5EntryFormat(DebugLineData, OffsetPtr, ContentTypes: &ContentTypes); |
| 309 | if (!FileDescriptors) |
| 310 | return FileDescriptors.takeError(); |
| 311 | |
| 312 | // Get the file entries, according to the format described above. |
| 313 | uint64_t FileEntryCount = DebugLineData.getULEB128(offset_ptr: OffsetPtr); |
| 314 | for (uint64_t I = 0; I != FileEntryCount; ++I) { |
| 315 | DWARFDebugLine::FileNameEntry FileEntry; |
| 316 | for (auto Descriptor : *FileDescriptors) { |
| 317 | DWARFFormValue Value(Descriptor.Form); |
| 318 | if (!Value.extractValue(Data: DebugLineData, OffsetPtr, FormParams, Context: &Ctx, Unit: U)) |
| 319 | return createStringError(EC: errc::invalid_argument, |
| 320 | S: "failed to parse file entry because " |
| 321 | "extracting the form value failed" ); |
| 322 | switch (Descriptor.Type) { |
| 323 | case DW_LNCT_path: |
| 324 | FileEntry.Name = Value; |
| 325 | break; |
| 326 | case DW_LNCT_LLVM_source: |
| 327 | FileEntry.Source = Value; |
| 328 | break; |
| 329 | case DW_LNCT_directory_index: |
| 330 | FileEntry.DirIdx = *Value.getAsUnsignedConstant(); |
| 331 | break; |
| 332 | case DW_LNCT_timestamp: |
| 333 | FileEntry.ModTime = *Value.getAsUnsignedConstant(); |
| 334 | break; |
| 335 | case DW_LNCT_size: |
| 336 | FileEntry.Length = *Value.getAsUnsignedConstant(); |
| 337 | break; |
| 338 | case DW_LNCT_MD5: |
| 339 | if (!Value.getAsBlock() || Value.getAsBlock()->size() != 16) |
| 340 | return createStringError( |
| 341 | EC: errc::invalid_argument, |
| 342 | S: "failed to parse file entry because the MD5 hash is invalid" ); |
| 343 | llvm::uninitialized_copy(Src: *Value.getAsBlock(), |
| 344 | Dst: FileEntry.Checksum.begin()); |
| 345 | break; |
| 346 | default: |
| 347 | break; |
| 348 | } |
| 349 | } |
| 350 | FileNames.push_back(x: FileEntry); |
| 351 | } |
| 352 | return Error::success(); |
| 353 | } |
| 354 | |
| 355 | uint64_t DWARFDebugLine::Prologue::getLength() const { |
| 356 | uint64_t Length = PrologueLength + sizeofTotalLength() + |
| 357 | sizeof(getVersion()) + sizeofPrologueLength(); |
| 358 | if (getVersion() >= 5) |
| 359 | Length += 2; // Address + Segment selector sizes. |
| 360 | return Length; |
| 361 | } |
| 362 | |
| 363 | Error DWARFDebugLine::Prologue::( |
| 364 | DWARFDataExtractor DebugLineData, uint64_t *OffsetPtr, |
| 365 | function_ref<void(Error)> RecoverableErrorHandler, const DWARFContext &Ctx, |
| 366 | const DWARFUnit *U) { |
| 367 | const uint64_t PrologueOffset = *OffsetPtr; |
| 368 | |
| 369 | clear(); |
| 370 | DataExtractor::Cursor Cursor(*OffsetPtr); |
| 371 | std::tie(args&: TotalLength, args&: FormParams.Format) = |
| 372 | DebugLineData.getInitialLength(C&: Cursor); |
| 373 | |
| 374 | DebugLineData = |
| 375 | DWARFDataExtractor(DebugLineData, Cursor.tell() + TotalLength); |
| 376 | FormParams.Version = DebugLineData.getU16(C&: Cursor); |
| 377 | if (Cursor && !versionIsSupported(Version: getVersion())) { |
| 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(); |
| 382 | return createStringError( |
| 383 | EC: errc::not_supported, |
| 384 | Fmt: "parsing line table prologue at offset 0x%8.8" PRIx64 |
| 385 | ": unsupported version %" PRIu16, |
| 386 | Vals: PrologueOffset, Vals: getVersion()); |
| 387 | } |
| 388 | |
| 389 | if (getVersion() >= 5) { |
| 390 | FormParams.AddrSize = DebugLineData.getU8(C&: Cursor); |
| 391 | const uint8_t DataAddrSize = DebugLineData.getAddressSize(); |
| 392 | const uint8_t PrologueAddrSize = getAddressSize(); |
| 393 | if (Cursor) { |
| 394 | if (DataAddrSize == 0) { |
| 395 | if (PrologueAddrSize != 4 && PrologueAddrSize != 8) { |
| 396 | RecoverableErrorHandler(createStringError( |
| 397 | EC: errc::not_supported, |
| 398 | Fmt: "parsing line table prologue at offset 0x%8.8" PRIx64 |
| 399 | ": invalid address size %" PRIu8, |
| 400 | Vals: PrologueOffset, Vals: PrologueAddrSize)); |
| 401 | } |
| 402 | } else if (DataAddrSize != PrologueAddrSize) { |
| 403 | RecoverableErrorHandler(createStringError( |
| 404 | EC: errc::not_supported, |
| 405 | Fmt: "parsing line table prologue at offset 0x%8.8" PRIx64 ": address " |
| 406 | "size %" PRIu8 " doesn't match architecture address size %" PRIu8, |
| 407 | Vals: PrologueOffset, Vals: PrologueAddrSize, Vals: DataAddrSize)); |
| 408 | } |
| 409 | } |
| 410 | SegSelectorSize = DebugLineData.getU8(C&: Cursor); |
| 411 | } |
| 412 | |
| 413 | PrologueLength = |
| 414 | DebugLineData.getRelocatedValue(C&: Cursor, Size: sizeofPrologueLength()); |
| 415 | const uint64_t EndPrologueOffset = PrologueLength + Cursor.tell(); |
| 416 | DebugLineData = DWARFDataExtractor(DebugLineData, EndPrologueOffset); |
| 417 | MinInstLength = DebugLineData.getU8(C&: Cursor); |
| 418 | if (getVersion() >= 4) |
| 419 | MaxOpsPerInst = DebugLineData.getU8(C&: Cursor); |
| 420 | DefaultIsStmt = DebugLineData.getU8(C&: Cursor); |
| 421 | LineBase = DebugLineData.getU8(C&: Cursor); |
| 422 | LineRange = DebugLineData.getU8(C&: Cursor); |
| 423 | OpcodeBase = DebugLineData.getU8(C&: Cursor); |
| 424 | |
| 425 | if (Cursor && OpcodeBase == 0) { |
| 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. |
| 429 | RecoverableErrorHandler(createStringError( |
| 430 | EC: errc::invalid_argument, |
| 431 | Fmt: "parsing line table prologue at offset 0x%8.8" PRIx64 |
| 432 | " found opcode base of 0. Assuming no standard opcodes" , |
| 433 | Vals: PrologueOffset)); |
| 434 | } else if (Cursor) { |
| 435 | StandardOpcodeLengths.reserve(n: OpcodeBase - 1); |
| 436 | for (uint32_t I = 1; I < OpcodeBase; ++I) { |
| 437 | uint8_t OpLen = DebugLineData.getU8(C&: Cursor); |
| 438 | StandardOpcodeLengths.push_back(x: OpLen); |
| 439 | } |
| 440 | } |
| 441 | |
| 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. |
| 446 | if (!Cursor) |
| 447 | return createStringError( |
| 448 | EC: errc::invalid_argument, |
| 449 | Fmt: "parsing line table prologue at offset 0x%8.8" PRIx64 ": %s" , |
| 450 | Vals: PrologueOffset, Vals: toString(E: Cursor.takeError()).c_str()); |
| 451 | |
| 452 | Error E = |
| 453 | getVersion() >= 5 |
| 454 | ? parseV5DirFileTables(DebugLineData, OffsetPtr, FormParams, Ctx, U, |
| 455 | ContentTypes, IncludeDirectories, FileNames) |
| 456 | : parseV2DirFileTables(DebugLineData, OffsetPtr, ContentTypes, |
| 457 | IncludeDirectories, FileNames); |
| 458 | if (E) { |
| 459 | RecoverableErrorHandler(joinErrors( |
| 460 | E1: createStringError( |
| 461 | EC: errc::invalid_argument, |
| 462 | Fmt: "parsing line table prologue at 0x%8.8" PRIx64 |
| 463 | " found an invalid directory or file table description at" |
| 464 | " 0x%8.8" PRIx64, |
| 465 | Vals: PrologueOffset, Vals: *OffsetPtr), |
| 466 | E2: std::move(E))); |
| 467 | return Error::success(); |
| 468 | } |
| 469 | |
| 470 | assert(*OffsetPtr <= EndPrologueOffset); |
| 471 | if (*OffsetPtr != EndPrologueOffset) { |
| 472 | RecoverableErrorHandler(createStringError( |
| 473 | EC: errc::invalid_argument, |
| 474 | Fmt: "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 | Vals: PrologueOffset, Vals: *OffsetPtr, Vals: EndPrologueOffset)); |
| 478 | } |
| 479 | return Error::success(); |
| 480 | } |
| 481 | |
| 482 | DWARFDebugLine::Row::Row(bool DefaultIsStmt) { reset(DefaultIsStmt); } |
| 483 | |
| 484 | void DWARFDebugLine::Row::postAppend() { |
| 485 | Discriminator = 0; |
| 486 | BasicBlock = false; |
| 487 | PrologueEnd = false; |
| 488 | EpilogueBegin = false; |
| 489 | } |
| 490 | |
| 491 | void DWARFDebugLine::Row::reset(bool DefaultIsStmt) { |
| 492 | Address.Address = 0; |
| 493 | Address.SectionIndex = object::SectionedAddress::UndefSection; |
| 494 | Line = 1; |
| 495 | Column = 0; |
| 496 | File = 1; |
| 497 | Isa = 0; |
| 498 | Discriminator = 0; |
| 499 | IsStmt = DefaultIsStmt; |
| 500 | OpIndex = 0; |
| 501 | BasicBlock = false; |
| 502 | EndSequence = false; |
| 503 | PrologueEnd = false; |
| 504 | EpilogueBegin = false; |
| 505 | } |
| 506 | |
| 507 | void DWARFDebugLine::Row::(raw_ostream &OS, unsigned Indent) { |
| 508 | OS.indent(NumSpaces: Indent) |
| 509 | << "Address Line Column File ISA Discriminator OpIndex " |
| 510 | "Flags\n" ; |
| 511 | OS.indent(NumSpaces: Indent) |
| 512 | << "------------------ ------ ------ ------ --- ------------- ------- " |
| 513 | "-------------\n" ; |
| 514 | } |
| 515 | |
| 516 | void DWARFDebugLine::Row::dump(raw_ostream &OS) const { |
| 517 | OS << format(Fmt: "0x%16.16" PRIx64 " %6u %6u" , Vals: Address.Address, Vals: Line, Vals: Column) |
| 518 | << format(Fmt: " %6u %3u %13u %7u " , Vals: File, Vals: Isa, Vals: Discriminator, Vals: OpIndex) |
| 519 | << (IsStmt ? " is_stmt" : "" ) << (BasicBlock ? " basic_block" : "" ) |
| 520 | << (PrologueEnd ? " prologue_end" : "" ) |
| 521 | << (EpilogueBegin ? " epilogue_begin" : "" ) |
| 522 | << (EndSequence ? " end_sequence" : "" ) << '\n'; |
| 523 | } |
| 524 | |
| 525 | DWARFDebugLine::Sequence::Sequence() { reset(); } |
| 526 | |
| 527 | void DWARFDebugLine::Sequence::reset() { |
| 528 | LowPC = 0; |
| 529 | HighPC = 0; |
| 530 | SectionIndex = object::SectionedAddress::UndefSection; |
| 531 | FirstRowIndex = 0; |
| 532 | LastRowIndex = 0; |
| 533 | Empty = true; |
| 534 | StmtSeqOffset = UINT64_MAX; |
| 535 | } |
| 536 | |
| 537 | DWARFDebugLine::LineTable::LineTable() { clear(); } |
| 538 | |
| 539 | void DWARFDebugLine::LineTable::dump(raw_ostream &OS, |
| 540 | DIDumpOptions DumpOptions) const { |
| 541 | Prologue.dump(OS, DumpOptions); |
| 542 | |
| 543 | if (!Rows.empty()) { |
| 544 | OS << '\n'; |
| 545 | Row::dumpTableHeader(OS, Indent: 0); |
| 546 | for (const Row &R : Rows) { |
| 547 | R.dump(OS); |
| 548 | } |
| 549 | } |
| 550 | |
| 551 | // Terminate the table with a final blank line to clearly delineate it from |
| 552 | // later dumps. |
| 553 | OS << '\n'; |
| 554 | } |
| 555 | |
| 556 | void DWARFDebugLine::LineTable::clear() { |
| 557 | Prologue.clear(); |
| 558 | Rows.clear(); |
| 559 | Sequences.clear(); |
| 560 | } |
| 561 | |
| 562 | DWARFDebugLine::ParsingState::ParsingState( |
| 563 | struct LineTable *LT, uint64_t TableOffset, |
| 564 | function_ref<void(Error)> ErrorHandler) |
| 565 | : LineTable(LT), LineTableOffset(TableOffset), ErrorHandler(ErrorHandler) {} |
| 566 | |
| 567 | void DWARFDebugLine::ParsingState::resetRowAndSequence(uint64_t Offset) { |
| 568 | Row.reset(DefaultIsStmt: LineTable->Prologue.DefaultIsStmt); |
| 569 | Sequence.reset(); |
| 570 | Sequence.StmtSeqOffset = Offset; |
| 571 | } |
| 572 | |
| 573 | void 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; |
| 580 | } |
| 581 | LineTable->appendRow(R: Row); |
| 582 | if (Row.EndSequence) { |
| 583 | // Record the end of instruction sequence. |
| 584 | Sequence.HighPC = Row.Address.Address; |
| 585 | Sequence.LastRowIndex = RowNumber + 1; |
| 586 | Sequence.SectionIndex = Row.Address.SectionIndex; |
| 587 | if (Sequence.isValid()) |
| 588 | LineTable->appendSequence(S: Sequence); |
| 589 | Sequence.reset(); |
| 590 | } |
| 591 | Row.postAppend(); |
| 592 | } |
| 593 | |
| 594 | const DWARFDebugLine::LineTable * |
| 595 | DWARFDebugLine::getLineTable(uint64_t Offset) const { |
| 596 | LineTableConstIter Pos = LineTableMap.find(x: Offset); |
| 597 | if (Pos != LineTableMap.end()) |
| 598 | return &Pos->second; |
| 599 | return nullptr; |
| 600 | } |
| 601 | |
| 602 | Expected<const DWARFDebugLine::LineTable *> DWARFDebugLine::( |
| 603 | DWARFDataExtractor &DebugLineData, uint64_t Offset, const DWARFContext &Ctx, |
| 604 | const DWARFUnit *U, function_ref<void(Error)> RecoverableErrorHandler) { |
| 605 | if (!DebugLineData.isValidOffset(offset: Offset)) |
| 606 | return createStringError(EC: errc::invalid_argument, |
| 607 | Fmt: "offset 0x%8.8" PRIx64 |
| 608 | " is not a valid debug line section offset" , |
| 609 | Vals: Offset); |
| 610 | |
| 611 | std::pair<LineTableIter, bool> Pos = |
| 612 | LineTableMap.insert(x: LineTableMapTy::value_type(Offset, LineTable())); |
| 613 | LineTable *LT = &Pos.first->second; |
| 614 | if (Pos.second) { |
| 615 | if (Error Err = |
| 616 | LT->parse(DebugLineData, OffsetPtr: &Offset, Ctx, U, RecoverableErrorHandler)) |
| 617 | return std::move(Err); |
| 618 | return LT; |
| 619 | } |
| 620 | return LT; |
| 621 | } |
| 622 | |
| 623 | void DWARFDebugLine::clearLineTable(uint64_t Offset) { |
| 624 | LineTableMap.erase(x: Offset); |
| 625 | } |
| 626 | |
| 627 | static StringRef getOpcodeName(uint8_t Opcode, uint8_t OpcodeBase) { |
| 628 | assert(Opcode != 0); |
| 629 | if (Opcode < OpcodeBase) |
| 630 | return LNStandardString(Standard: Opcode); |
| 631 | return "special" ; |
| 632 | } |
| 633 | |
| 634 | DWARFDebugLine::ParsingState::AddrOpIndexDelta |
| 635 | DWARFDebugLine::ParsingState::advanceAddrOpIndex(uint64_t OperationAdvance, |
| 636 | uint8_t Opcode, |
| 637 | uint64_t OpcodeOffset) { |
| 638 | StringRef OpcodeName = getOpcodeName(Opcode, OpcodeBase: LineTable->Prologue.OpcodeBase); |
| 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. |
| 642 | if (ReportAdvanceAddrProblem && LineTable->Prologue.getVersion() >= 4 && |
| 643 | LineTable->Prologue.MaxOpsPerInst == 0) |
| 644 | ErrorHandler(createStringError( |
| 645 | EC: errc::invalid_argument, |
| 646 | Fmt: "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 | Vals: LineTableOffset, Vals: OpcodeName.data(), Vals: 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 |
| 655 | // results. |
| 656 | if (ReportAdvanceAddrProblem && LineTable->Prologue.MaxOpsPerInst > 1) |
| 657 | ErrorHandler(createStringError( |
| 658 | EC: errc::not_supported, |
| 659 | Fmt: "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 " |
| 663 | "may be incorrect" , |
| 664 | Vals: LineTableOffset, Vals: OpcodeName.data(), Vals: OpcodeOffset, |
| 665 | Vals: LineTable->Prologue.MaxOpsPerInst)); |
| 666 | if (ReportAdvanceAddrProblem && LineTable->Prologue.MinInstLength == 0) |
| 667 | ErrorHandler( |
| 668 | createStringError(EC: errc::invalid_argument, |
| 669 | Fmt: "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 | Vals: LineTableOffset, Vals: OpcodeName.data(), Vals: OpcodeOffset)); |
| 674 | ReportAdvanceAddrProblem = false; |
| 675 | |
| 676 | // Advances the address and op_index according to DWARFv5, section 6.2.5.1: |
| 677 | // |
| 678 | // new address = address + |
| 679 | // minimum_instruction_length * |
| 680 | // ((op_index + operation advance) / maximum_operations_per_instruction) |
| 681 | // |
| 682 | // new op_index = |
| 683 | // (op_index + operation advance) % maximum_operations_per_instruction |
| 684 | |
| 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 = |
| 688 | std::max(a: LineTable->Prologue.MaxOpsPerInst, b: uint8_t{1}); |
| 689 | |
| 690 | uint64_t AddrOffset = ((Row.OpIndex + OperationAdvance) / MaxOpsPerInst) * |
| 691 | LineTable->Prologue.MinInstLength; |
| 692 | Row.Address.Address += AddrOffset; |
| 693 | |
| 694 | uint8_t PrevOpIndex = Row.OpIndex; |
| 695 | Row.OpIndex = (Row.OpIndex + OperationAdvance) % MaxOpsPerInst; |
| 696 | int16_t OpIndexDelta = static_cast<int16_t>(Row.OpIndex) - PrevOpIndex; |
| 697 | |
| 698 | return {.AddrOffset: AddrOffset, .OpIndexDelta: OpIndexDelta}; |
| 699 | } |
| 700 | |
| 701 | DWARFDebugLine::ParsingState::OpcodeAdvanceResults |
| 702 | DWARFDebugLine::ParsingState::advanceForOpcode(uint8_t Opcode, |
| 703 | uint64_t OpcodeOffset) { |
| 704 | assert(Opcode == DW_LNS_const_add_pc || |
| 705 | Opcode >= LineTable->Prologue.OpcodeBase); |
| 706 | if (ReportBadLineRange && LineTable->Prologue.LineRange == 0) { |
| 707 | StringRef OpcodeName = |
| 708 | getOpcodeName(Opcode, OpcodeBase: LineTable->Prologue.OpcodeBase); |
| 709 | ErrorHandler( |
| 710 | createStringError(EC: errc::not_supported, |
| 711 | Fmt: "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 | Vals: LineTableOffset, Vals: OpcodeName.data(), Vals: OpcodeOffset)); |
| 716 | ReportBadLineRange = false; |
| 717 | } |
| 718 | |
| 719 | uint8_t OpcodeValue = Opcode; |
| 720 | if (Opcode == DW_LNS_const_add_pc) |
| 721 | OpcodeValue = 255; |
| 722 | uint8_t AdjustedOpcode = OpcodeValue - LineTable->Prologue.OpcodeBase; |
| 723 | uint64_t OperationAdvance = |
| 724 | LineTable->Prologue.LineRange != 0 |
| 725 | ? AdjustedOpcode / LineTable->Prologue.LineRange |
| 726 | : 0; |
| 727 | AddrOpIndexDelta Advance = |
| 728 | advanceAddrOpIndex(OperationAdvance, Opcode, OpcodeOffset); |
| 729 | return {.AddrDelta: Advance.AddrOffset, .OpIndexDelta: Advance.OpIndexDelta, .AdjustedOpcode: AdjustedOpcode}; |
| 730 | } |
| 731 | |
| 732 | DWARFDebugLine::ParsingState::SpecialOpcodeDelta |
| 733 | DWARFDebugLine::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: |
| 745 | // |
| 746 | // opcode = (desired line increment - line_base) + |
| 747 | // (line_range * address advance) + opcode_base |
| 748 | // |
| 749 | // If the resulting opcode is greater than 255, a standard opcode |
| 750 | // must be used instead. |
| 751 | // |
| 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: |
| 757 | // |
| 758 | // address increment = (adjusted opcode / line_range) * |
| 759 | // minimum_instruction_length |
| 760 | // |
| 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: |
| 763 | // |
| 764 | // line increment = line_base + (adjusted opcode % line_range) |
| 765 | |
| 766 | DWARFDebugLine::ParsingState::OpcodeAdvanceResults AddrAdvanceResult = |
| 767 | advanceForOpcode(Opcode, OpcodeOffset); |
| 768 | int32_t LineOffset = 0; |
| 769 | if (LineTable->Prologue.LineRange != 0) |
| 770 | LineOffset = |
| 771 | LineTable->Prologue.LineBase + |
| 772 | (AddrAdvanceResult.AdjustedOpcode % LineTable->Prologue.LineRange); |
| 773 | Row.Line += LineOffset; |
| 774 | return {.Address: AddrAdvanceResult.AddrDelta, .Line: LineOffset, |
| 775 | .OpIndex: AddrAdvanceResult.OpIndexDelta}; |
| 776 | } |
| 777 | |
| 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. |
| 780 | template <typename T> |
| 781 | static std::optional<T> (DWARFDataExtractor &Data, |
| 782 | DataExtractor::Cursor &Cursor) { |
| 783 | T Value = Data.getULEB128(C&: Cursor); |
| 784 | if (Cursor) |
| 785 | return Value; |
| 786 | return std::nullopt; |
| 787 | } |
| 788 | |
| 789 | Error DWARFDebugLine::LineTable::( |
| 790 | DWARFDataExtractor &DebugLineData, uint64_t *OffsetPtr, |
| 791 | const DWARFContext &Ctx, const DWARFUnit *U, |
| 792 | function_ref<void(Error)> RecoverableErrorHandler, raw_ostream *OS, |
| 793 | bool Verbose) { |
| 794 | assert((OS || !Verbose) && "cannot have verbose output without stream" ); |
| 795 | const uint64_t DebugLineOffset = *OffsetPtr; |
| 796 | |
| 797 | clear(); |
| 798 | |
| 799 | Error PrologueErr = |
| 800 | Prologue.parse(DebugLineData, OffsetPtr, RecoverableErrorHandler, Ctx, U); |
| 801 | |
| 802 | if (OS) { |
| 803 | DIDumpOptions DumpOptions; |
| 804 | DumpOptions.Verbose = Verbose; |
| 805 | Prologue.dump(OS&: *OS, DumpOptions); |
| 806 | } |
| 807 | |
| 808 | if (PrologueErr) { |
| 809 | // Ensure there is a blank line after the prologue to clearly delineate it |
| 810 | // from later dumps. |
| 811 | if (OS) |
| 812 | *OS << "\n" ; |
| 813 | return PrologueErr; |
| 814 | } |
| 815 | |
| 816 | uint64_t ProgramLength = Prologue.TotalLength + Prologue.sizeofTotalLength(); |
| 817 | if (!DebugLineData.isValidOffsetForDataOfSize(offset: DebugLineOffset, |
| 818 | length: ProgramLength)) { |
| 819 | assert(DebugLineData.size() > DebugLineOffset && |
| 820 | "prologue parsing should handle invalid offset" ); |
| 821 | uint64_t BytesRemaining = DebugLineData.size() - DebugLineOffset; |
| 822 | RecoverableErrorHandler( |
| 823 | createStringError(EC: errc::invalid_argument, |
| 824 | Fmt: "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 | Vals: DebugLineOffset, Vals: ProgramLength, Vals: BytesRemaining)); |
| 828 | // Continue by capping the length at the number of remaining bytes. |
| 829 | ProgramLength = BytesRemaining; |
| 830 | } |
| 831 | |
| 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; |
| 835 | DWARFDataExtractor TableData(DebugLineData, EndOffset); |
| 836 | |
| 837 | // See if we should tell the data extractor the address size. |
| 838 | if (TableData.getAddressSize() == 0) |
| 839 | TableData.setAddressSize(Prologue.getAddressSize()); |
| 840 | else |
| 841 | assert(Prologue.getAddressSize() == 0 || |
| 842 | Prologue.getAddressSize() == TableData.getAddressSize()); |
| 843 | |
| 844 | ParsingState State(this, DebugLineOffset, RecoverableErrorHandler); |
| 845 | |
| 846 | *OffsetPtr = DebugLineOffset + Prologue.getLength(); |
| 847 | if (OS && *OffsetPtr < EndOffset) { |
| 848 | *OS << '\n'; |
| 849 | Row::dumpTableHeader(OS&: *OS, /*Indent=*/Verbose ? 12 : 0); |
| 850 | } |
| 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(Offset: *OffsetPtr); |
| 854 | |
| 855 | bool TombstonedAddress = false; |
| 856 | auto EmitRow = [&] { |
| 857 | if (!TombstonedAddress) { |
| 858 | if (Verbose) { |
| 859 | *OS << "\n" ; |
| 860 | OS->indent(NumSpaces: 12); |
| 861 | } |
| 862 | if (OS) |
| 863 | State.Row.dump(OS&: *OS); |
| 864 | State.appendRowToMatrix(); |
| 865 | } |
| 866 | }; |
| 867 | while (*OffsetPtr < EndOffset) { |
| 868 | DataExtractor::Cursor Cursor(*OffsetPtr); |
| 869 | |
| 870 | if (Verbose) |
| 871 | *OS << format(Fmt: "0x%08.08" PRIx64 ": " , Vals: *OffsetPtr); |
| 872 | |
| 873 | uint64_t OpcodeOffset = *OffsetPtr; |
| 874 | uint8_t Opcode = TableData.getU8(C&: Cursor); |
| 875 | size_t RowCount = Rows.size(); |
| 876 | |
| 877 | if (Cursor && Verbose) |
| 878 | *OS << format(Fmt: "%02.02" PRIx8 " " , Vals: Opcode); |
| 879 | |
| 880 | if (Opcode == 0) { |
| 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 |
| 883 | uint64_t Len = TableData.getULEB128(C&: Cursor); |
| 884 | uint64_t ExtOffset = Cursor.tell(); |
| 885 | |
| 886 | // Tolerate zero-length; assume length is correct and soldier on. |
| 887 | if (Len == 0) { |
| 888 | if (Cursor && Verbose) |
| 889 | *OS << "Badly formed extended line op (length 0)\n" ; |
| 890 | if (!Cursor) { |
| 891 | if (Verbose) |
| 892 | *OS << "\n" ; |
| 893 | RecoverableErrorHandler(Cursor.takeError()); |
| 894 | } |
| 895 | *OffsetPtr = Cursor.tell(); |
| 896 | continue; |
| 897 | } |
| 898 | |
| 899 | uint8_t SubOpcode = TableData.getU8(C&: Cursor); |
| 900 | // OperandOffset will be the same as ExtOffset, if it was not possible to |
| 901 | // read the SubOpcode. |
| 902 | uint64_t OperandOffset = Cursor.tell(); |
| 903 | if (Verbose) |
| 904 | *OS << LNExtendedString(Encoding: SubOpcode); |
| 905 | switch (SubOpcode) { |
| 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 |
| 913 | // of the sequence. |
| 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 |
| 917 | // followed. |
| 918 | EmitRow(); |
| 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(Offset: Cursor.tell()); |
| 922 | break; |
| 923 | |
| 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 |
| 931 | // it instead. |
| 932 | // |
| 933 | // Make sure the extractor knows the address size. If not, infer it |
| 934 | // from the size of the operand. |
| 935 | { |
| 936 | uint8_t = TableData.getAddressSize(); |
| 937 | uint64_t OpcodeAddressSize = Len - 1; |
| 938 | if (ExtractorAddressSize != OpcodeAddressSize && |
| 939 | ExtractorAddressSize != 0) |
| 940 | RecoverableErrorHandler(createStringError( |
| 941 | EC: errc::invalid_argument, |
| 942 | Fmt: "mismatching address size at offset 0x%8.8" PRIx64 |
| 943 | " expected 0x%2.2" PRIx8 " found 0x%2.2" PRIx64, |
| 944 | Vals: ExtOffset, Vals: ExtractorAddressSize, Vals: Len - 1)); |
| 945 | |
| 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) { |
| 951 | RecoverableErrorHandler(createStringError( |
| 952 | EC: errc::invalid_argument, |
| 953 | Fmt: "address size 0x%2.2" PRIx64 |
| 954 | " of DW_LNE_set_address opcode at offset 0x%8.8" PRIx64 |
| 955 | " is unsupported" , |
| 956 | Vals: OpcodeAddressSize, Vals: ExtOffset)); |
| 957 | TableData.skip(C&: Cursor, Length: OpcodeAddressSize); |
| 958 | } else { |
| 959 | TableData.setAddressSize(OpcodeAddressSize); |
| 960 | State.Row.Address.Address = TableData.getRelocatedAddress( |
| 961 | C&: Cursor, SecIx: &State.Row.Address.SectionIndex); |
| 962 | State.Row.OpIndex = 0; |
| 963 | |
| 964 | uint64_t Tombstone = |
| 965 | dwarf::computeTombstoneAddress(AddressByteSize: OpcodeAddressSize); |
| 966 | TombstonedAddress = State.Row.Address.Address == Tombstone; |
| 967 | |
| 968 | // Restore the address size if the extractor already had it. |
| 969 | if (ExtractorAddressSize != 0) |
| 970 | TableData.setAddressSize(ExtractorAddressSize); |
| 971 | } |
| 972 | |
| 973 | if (Cursor && Verbose) { |
| 974 | *OS << " (" ; |
| 975 | DWARFFormValue::dumpAddress(OS&: *OS, AddressSize: OpcodeAddressSize, |
| 976 | Address: State.Row.Address.Address); |
| 977 | *OS << ')'; |
| 978 | } |
| 979 | } |
| 980 | break; |
| 981 | |
| 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 |
| 990 | // available. |
| 991 | // |
| 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. |
| 998 | // |
| 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. |
| 1003 | { |
| 1004 | FileNameEntry FileEntry; |
| 1005 | const char *Name = TableData.getCStr(C&: Cursor); |
| 1006 | FileEntry.Name = |
| 1007 | DWARFFormValue::createFromPValue(F: dwarf::DW_FORM_string, V: Name); |
| 1008 | FileEntry.DirIdx = TableData.getULEB128(C&: Cursor); |
| 1009 | FileEntry.ModTime = TableData.getULEB128(C&: Cursor); |
| 1010 | FileEntry.Length = TableData.getULEB128(C&: Cursor); |
| 1011 | Prologue.FileNames.push_back(x: FileEntry); |
| 1012 | if (Cursor && Verbose) |
| 1013 | *OS << " (" << Name << ", dir=" << FileEntry.DirIdx << ", mod_time=" |
| 1014 | << format(Fmt: "(0x%16.16" PRIx64 ")" , Vals: FileEntry.ModTime) |
| 1015 | << ", length=" << FileEntry.Length << ")" ; |
| 1016 | } |
| 1017 | break; |
| 1018 | |
| 1019 | case DW_LNE_set_discriminator: |
| 1020 | State.Row.Discriminator = TableData.getULEB128(C&: Cursor); |
| 1021 | if (Cursor && Verbose) |
| 1022 | *OS << " (" << State.Row.Discriminator << ")" ; |
| 1023 | break; |
| 1024 | |
| 1025 | default: |
| 1026 | if (Cursor && Verbose) |
| 1027 | *OS << format(Fmt: "Unrecognized extended op 0x%02.02" PRIx8, Vals: SubOpcode) |
| 1028 | << format(Fmt: " length %" PRIx64, Vals: 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(C&: Cursor, Length: Len - 1); |
| 1032 | break; |
| 1033 | } |
| 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 |
| 1037 | // a parsing error. |
| 1038 | uint64_t End = ExtOffset + Len; |
| 1039 | if (Cursor && Cursor.tell() != End) |
| 1040 | RecoverableErrorHandler(createStringError( |
| 1041 | EC: errc::illegal_byte_sequence, |
| 1042 | Fmt: "unexpected line op length at offset 0x%8.8" PRIx64 |
| 1043 | " expected 0x%2.2" PRIx64 " found 0x%2.2" PRIx64, |
| 1044 | Vals: ExtOffset, Vals: Len, Vals: Cursor.tell() - ExtOffset)); |
| 1045 | if (!Cursor && Verbose) { |
| 1046 | DWARFDataExtractor::Cursor ByteCursor(OperandOffset); |
| 1047 | uint8_t Byte = TableData.getU8(C&: ByteCursor); |
| 1048 | if (ByteCursor) { |
| 1049 | *OS << " (<parsing error>" ; |
| 1050 | do { |
| 1051 | *OS << format(Fmt: " %2.2" PRIx8, Vals: Byte); |
| 1052 | Byte = TableData.getU8(C&: ByteCursor); |
| 1053 | } while (ByteCursor); |
| 1054 | *OS << ")" ; |
| 1055 | } |
| 1056 | |
| 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 |
| 1059 | // be sufficient. |
| 1060 | consumeError(Err: ByteCursor.takeError()); |
| 1061 | } |
| 1062 | *OffsetPtr = End; |
| 1063 | } else if (Opcode < Prologue.OpcodeBase) { |
| 1064 | if (Verbose) |
| 1065 | *OS << LNStandardString(Standard: Opcode); |
| 1066 | switch (Opcode) { |
| 1067 | // Standard Opcodes |
| 1068 | case DW_LNS_copy: |
| 1069 | // Takes no arguments. Append a row to the matrix using the |
| 1070 | // current values of the state-machine registers. |
| 1071 | EmitRow(); |
| 1072 | break; |
| 1073 | |
| 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 = |
| 1079 | parseULEB128<uint64_t>(Data&: TableData, Cursor)) { |
| 1080 | ParsingState::AddrOpIndexDelta Advance = |
| 1081 | State.advanceAddrOpIndex(OperationAdvance: *Operand, Opcode, OpcodeOffset); |
| 1082 | if (Verbose) |
| 1083 | *OS << " (addr += " << Advance.AddrOffset |
| 1084 | << ", op-index += " << Advance.OpIndexDelta << ")" ; |
| 1085 | } |
| 1086 | break; |
| 1087 | |
| 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. |
| 1091 | { |
| 1092 | int64_t LineDelta = TableData.getSLEB128(C&: Cursor); |
| 1093 | if (Cursor) { |
| 1094 | State.Row.Line += LineDelta; |
| 1095 | if (Verbose) |
| 1096 | *OS << " (" << State.Row.Line << ")" ; |
| 1097 | } |
| 1098 | } |
| 1099 | break; |
| 1100 | |
| 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 = |
| 1105 | parseULEB128<uint16_t>(Data&: TableData, Cursor)) { |
| 1106 | State.Row.File = *File; |
| 1107 | if (Verbose) |
| 1108 | *OS << " (" << State.Row.File << ")" ; |
| 1109 | } |
| 1110 | break; |
| 1111 | |
| 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 = |
| 1116 | parseULEB128<uint16_t>(Data&: TableData, Cursor)) { |
| 1117 | State.Row.Column = *Column; |
| 1118 | if (Verbose) |
| 1119 | *OS << " (" << State.Row.Column << ")" ; |
| 1120 | } |
| 1121 | break; |
| 1122 | |
| 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; |
| 1127 | break; |
| 1128 | |
| 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; |
| 1133 | break; |
| 1134 | |
| 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. |
| 1147 | { |
| 1148 | ParsingState::OpcodeAdvanceResults Advance = |
| 1149 | State.advanceForOpcode(Opcode, OpcodeOffset); |
| 1150 | if (Verbose) |
| 1151 | *OS << format(Fmt: " (addr += 0x%16.16" PRIx64 ", op-index += %" PRIu8 |
| 1152 | ")" , |
| 1153 | Vals: Advance.AddrDelta, Vals: Advance.OpIndexDelta); |
| 1154 | } |
| 1155 | break; |
| 1156 | |
| 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. |
| 1168 | { |
| 1169 | uint16_t PCOffset = TableData.getRelocatedValue(C&: Cursor, Size: 2); |
| 1170 | if (Cursor) { |
| 1171 | State.Row.Address.Address += PCOffset; |
| 1172 | State.Row.OpIndex = 0; |
| 1173 | if (Verbose) |
| 1174 | *OS << format(Fmt: " (addr += 0x%4.4" PRIx16 ", op-index = 0)" , |
| 1175 | Vals: PCOffset); |
| 1176 | } |
| 1177 | } |
| 1178 | break; |
| 1179 | |
| 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; |
| 1184 | break; |
| 1185 | |
| 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; |
| 1190 | break; |
| 1191 | |
| 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 = |
| 1196 | parseULEB128<uint8_t>(Data&: TableData, Cursor)) { |
| 1197 | State.Row.Isa = *Isa; |
| 1198 | if (Verbose) |
| 1199 | *OS << " (" << (uint64_t)State.Row.Isa << ")" ; |
| 1200 | } |
| 1201 | break; |
| 1202 | |
| 1203 | default: |
| 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. |
| 1207 | { |
| 1208 | assert(Opcode - 1U < Prologue.StandardOpcodeLengths.size()); |
| 1209 | if (Verbose) |
| 1210 | *OS << "Unrecognized standard opcode" ; |
| 1211 | uint8_t OpcodeLength = Prologue.StandardOpcodeLengths[Opcode - 1]; |
| 1212 | std::vector<uint64_t> Operands; |
| 1213 | for (uint8_t I = 0; I < OpcodeLength; ++I) { |
| 1214 | if (std::optional<uint64_t> Value = |
| 1215 | parseULEB128<uint64_t>(Data&: TableData, Cursor)) |
| 1216 | Operands.push_back(x: *Value); |
| 1217 | else |
| 1218 | break; |
| 1219 | } |
| 1220 | if (Verbose && !Operands.empty()) { |
| 1221 | *OS << " (operands: " ; |
| 1222 | bool First = true; |
| 1223 | for (uint64_t Value : Operands) { |
| 1224 | if (!First) |
| 1225 | *OS << ", " ; |
| 1226 | First = false; |
| 1227 | *OS << format(Fmt: "0x%16.16" PRIx64, Vals: Value); |
| 1228 | } |
| 1229 | if (Verbose) |
| 1230 | *OS << ')'; |
| 1231 | } |
| 1232 | } |
| 1233 | break; |
| 1234 | } |
| 1235 | |
| 1236 | *OffsetPtr = Cursor.tell(); |
| 1237 | } else { |
| 1238 | // Special Opcodes. |
| 1239 | ParsingState::SpecialOpcodeDelta Delta = |
| 1240 | State.handleSpecialOpcode(Opcode, OpcodeOffset); |
| 1241 | |
| 1242 | if (Verbose) |
| 1243 | *OS << "address += " << Delta.Address << ", line += " << Delta.Line |
| 1244 | << ", op-index += " << Delta.OpIndex; |
| 1245 | EmitRow(); |
| 1246 | *OffsetPtr = Cursor.tell(); |
| 1247 | } |
| 1248 | |
| 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. |
| 1251 | if (Verbose && Rows.size() == RowCount) |
| 1252 | *OS << "\n" ; |
| 1253 | |
| 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) { |
| 1259 | if (Verbose) |
| 1260 | *OS << "\n" ; |
| 1261 | return Cursor.takeError(); |
| 1262 | } |
| 1263 | |
| 1264 | if (!Cursor) |
| 1265 | RecoverableErrorHandler(Cursor.takeError()); |
| 1266 | } |
| 1267 | |
| 1268 | if (!State.Sequence.Empty) |
| 1269 | RecoverableErrorHandler(createStringError( |
| 1270 | EC: errc::illegal_byte_sequence, |
| 1271 | Fmt: "last sequence in debug line table at offset 0x%8.8" PRIx64 |
| 1272 | " is not terminated" , |
| 1273 | Vals: DebugLineOffset)); |
| 1274 | |
| 1275 | // Sort all sequences so that address lookup will work faster. |
| 1276 | if (!Sequences.empty()) { |
| 1277 | llvm::stable_sort(Range&: Sequences, C: Sequence::orderByHighPC); |
| 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. |
| 1285 | } |
| 1286 | |
| 1287 | // Terminate the table with a final blank line to clearly delineate it from |
| 1288 | // later dumps. |
| 1289 | if (OS) |
| 1290 | *OS << "\n" ; |
| 1291 | |
| 1292 | return Error::success(); |
| 1293 | } |
| 1294 | |
| 1295 | uint32_t DWARFDebugLine::LineTable::findRowInSeq( |
| 1296 | const DWARFDebugLine::Sequence &Seq, |
| 1297 | object::SectionedAddress Address) const { |
| 1298 | if (!Seq.containsPC(PC: Address)) |
| 1299 | return UnknownRowIndex; |
| 1300 | assert(Seq.SectionIndex == Address.SectionIndex); |
| 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. |
| 1303 | // |
| 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. |
| 1306 | // |
| 1307 | // TODO: This function, and its users, needs to be update to return multiple |
| 1308 | // rows for bundles with multiple op-indexes. |
| 1309 | DWARFDebugLine::Row Row; |
| 1310 | Row.Address = Address; |
| 1311 | RowIter FirstRow = Rows.begin() + Seq.FirstRowIndex; |
| 1312 | RowIter LastRow = Rows.begin() + Seq.LastRowIndex; |
| 1313 | assert(FirstRow->Address.Address <= Row.Address.Address && |
| 1314 | Row.Address.Address < LastRow[-1].Address.Address); |
| 1315 | RowIter RowPos = std::upper_bound(first: FirstRow + 1, last: LastRow - 1, val: Row, |
| 1316 | comp: DWARFDebugLine::Row::orderByAddress) - |
| 1317 | 1; |
| 1318 | assert(Seq.SectionIndex == RowPos->Address.SectionIndex); |
| 1319 | return RowPos - Rows.begin(); |
| 1320 | } |
| 1321 | |
| 1322 | uint32_t |
| 1323 | DWARFDebugLine::LineTable::lookupAddress(object::SectionedAddress Address, |
| 1324 | bool *IsApproximateLine) const { |
| 1325 | |
| 1326 | // Search for relocatable addresses |
| 1327 | uint32_t Result = lookupAddressImpl(Address, IsApproximateLine); |
| 1328 | |
| 1329 | if (Result != UnknownRowIndex || |
| 1330 | Address.SectionIndex == object::SectionedAddress::UndefSection) |
| 1331 | return Result; |
| 1332 | |
| 1333 | // Search for absolute addresses |
| 1334 | Address.SectionIndex = object::SectionedAddress::UndefSection; |
| 1335 | return lookupAddressImpl(Address, IsApproximateLine); |
| 1336 | } |
| 1337 | |
| 1338 | uint32_t |
| 1339 | DWARFDebugLine::LineTable::lookupAddressImpl(object::SectionedAddress Address, |
| 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. |
| 1345 | DWARFDebugLine::Sequence Sequence; |
| 1346 | Sequence.SectionIndex = Address.SectionIndex; |
| 1347 | Sequence.HighPC = Address.Address; |
| 1348 | SequenceIter It = llvm::upper_bound(Range: Sequences, Value&: Sequence, |
| 1349 | C: DWARFDebugLine::Sequence::orderByHighPC); |
| 1350 | if (It == Sequences.end() || It->SectionIndex != Address.SectionIndex) |
| 1351 | return UnknownRowIndex; |
| 1352 | |
| 1353 | uint32_t RowIndex = findRowInSeq(Seq: *It, Address); |
| 1354 | if (RowIndex == UnknownRowIndex || !IsApproximateLine) |
| 1355 | return RowIndex; |
| 1356 | |
| 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; |
| 1364 | } |
| 1365 | // Approximation Loop fails to find the valid ApproxRowIndex |
| 1366 | if (ApproxRowIndex < It->FirstRowIndex) |
| 1367 | *IsApproximateLine = false; |
| 1368 | |
| 1369 | return RowIndex; |
| 1370 | } |
| 1371 | |
| 1372 | bool DWARFDebugLine::LineTable::lookupAddressRange( |
| 1373 | object::SectionedAddress Address, uint64_t Size, |
| 1374 | std::vector<uint32_t> &Result, |
| 1375 | std::optional<uint64_t> StmtSequenceOffset) const { |
| 1376 | |
| 1377 | // Search for relocatable addresses |
| 1378 | if (lookupAddressRangeImpl(Address, Size, Result, StmtSequenceOffset)) |
| 1379 | return true; |
| 1380 | |
| 1381 | if (Address.SectionIndex == object::SectionedAddress::UndefSection) |
| 1382 | return false; |
| 1383 | |
| 1384 | // Search for absolute addresses |
| 1385 | Address.SectionIndex = object::SectionedAddress::UndefSection; |
| 1386 | return lookupAddressRangeImpl(Address, Size, Result, StmtSequenceOffset); |
| 1387 | } |
| 1388 | |
| 1389 | bool DWARFDebugLine::LineTable::lookupAddressRangeImpl( |
| 1390 | object::SectionedAddress Address, uint64_t Size, |
| 1391 | std::vector<uint32_t> &Result, |
| 1392 | std::optional<uint64_t> StmtSequenceOffset) const { |
| 1393 | if (Sequences.empty()) |
| 1394 | return false; |
| 1395 | uint64_t EndAddr = Address.Address + Size; |
| 1396 | // First, find an instruction sequence containing the given address. |
| 1397 | DWARFDebugLine::Sequence Sequence; |
| 1398 | Sequence.SectionIndex = Address.SectionIndex; |
| 1399 | Sequence.HighPC = Address.Address; |
| 1400 | SequenceIter LastSeq = Sequences.end(); |
| 1401 | SequenceIter SeqPos; |
| 1402 | |
| 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(first: Sequences.begin(), last: LastSeq, |
| 1407 | pred: [&](const DWARFDebugLine::Sequence &S) { |
| 1408 | return S.StmtSeqOffset == *StmtSequenceOffset; |
| 1409 | }); |
| 1410 | |
| 1411 | // If sequence not found, return false |
| 1412 | if (SeqPos == LastSeq) |
| 1413 | return false; |
| 1414 | |
| 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; |
| 1418 | } else { |
| 1419 | // No specific sequence requested, find first sequence containing address |
| 1420 | SeqPos = std::upper_bound(first: Sequences.begin(), last: LastSeq, val: Sequence, |
| 1421 | comp: DWARFDebugLine::Sequence::orderByHighPC); |
| 1422 | if (SeqPos == LastSeq) |
| 1423 | return false; |
| 1424 | } |
| 1425 | |
| 1426 | // If the start sequence doesn't contain the address, nothing to do |
| 1427 | if (!SeqPos->containsPC(PC: Address)) |
| 1428 | return false; |
| 1429 | |
| 1430 | SequenceIter StartPos = SeqPos; |
| 1431 | |
| 1432 | // Process sequences that overlap with the desired range |
| 1433 | while (SeqPos != LastSeq && SeqPos->LowPC < EndAddr) { |
| 1434 | const DWARFDebugLine::Sequence &CurSeq = *SeqPos; |
| 1435 | // For the first sequence, we need to find which row in the sequence is the |
| 1436 | // first in our range. |
| 1437 | uint32_t FirstRowIndex = CurSeq.FirstRowIndex; |
| 1438 | if (SeqPos == StartPos) |
| 1439 | FirstRowIndex = findRowInSeq(Seq: CurSeq, Address); |
| 1440 | |
| 1441 | // Figure out the last row in the range. |
| 1442 | uint32_t LastRowIndex = |
| 1443 | findRowInSeq(Seq: CurSeq, Address: {.Address: EndAddr - 1, .SectionIndex: Address.SectionIndex}); |
| 1444 | if (LastRowIndex == UnknownRowIndex) |
| 1445 | LastRowIndex = CurSeq.LastRowIndex - 1; |
| 1446 | |
| 1447 | assert(FirstRowIndex != UnknownRowIndex); |
| 1448 | assert(LastRowIndex != UnknownRowIndex); |
| 1449 | |
| 1450 | for (uint32_t I = FirstRowIndex; I <= LastRowIndex; ++I) { |
| 1451 | Result.push_back(x: I); |
| 1452 | } |
| 1453 | |
| 1454 | ++SeqPos; |
| 1455 | } |
| 1456 | |
| 1457 | return true; |
| 1458 | } |
| 1459 | |
| 1460 | std::optional<StringRef> |
| 1461 | DWARFDebugLine::LineTable::getSourceByIndex(uint64_t FileIndex, |
| 1462 | FileLineInfoKind Kind) const { |
| 1463 | if (Kind == FileLineInfoKind::None || !Prologue.hasFileAtIndex(FileIndex)) |
| 1464 | return std::nullopt; |
| 1465 | const FileNameEntry &Entry = Prologue.getFileNameEntry(Index: FileIndex); |
| 1466 | if (auto E = dwarf::toString(V: Entry.Source)) |
| 1467 | return StringRef(*E); |
| 1468 | return std::nullopt; |
| 1469 | } |
| 1470 | |
| 1471 | static bool isPathAbsoluteOnWindowsOrPosix(const Twine &Path) { |
| 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. |
| 1475 | return sys::path::is_absolute(path: Path, style: sys::path::Style::posix) || |
| 1476 | sys::path::is_absolute(path: Path, style: sys::path::Style::windows); |
| 1477 | } |
| 1478 | |
| 1479 | bool DWARFDebugLine::Prologue::getFileNameByIndex( |
| 1480 | uint64_t FileIndex, StringRef CompDir, FileLineInfoKind Kind, |
| 1481 | std::string &Result, sys::path::Style Style) const { |
| 1482 | if (Kind == FileLineInfoKind::None || !hasFileAtIndex(FileIndex)) |
| 1483 | return false; |
| 1484 | const FileNameEntry &Entry = getFileNameEntry(Index: FileIndex); |
| 1485 | auto E = dwarf::toString(V: Entry.Name); |
| 1486 | if (!E) |
| 1487 | return false; |
| 1488 | StringRef FileName = *E; |
| 1489 | if (Kind == FileLineInfoKind::RawValue || |
| 1490 | isPathAbsoluteOnWindowsOrPosix(Path: FileName)) { |
| 1491 | Result = std::string(FileName); |
| 1492 | return true; |
| 1493 | } |
| 1494 | if (Kind == FileLineInfoKind::BaseNameOnly) { |
| 1495 | Result = std::string(llvm::sys::path::filename(path: FileName)); |
| 1496 | return true; |
| 1497 | } |
| 1498 | |
| 1499 | SmallString<16> FilePath; |
| 1500 | StringRef IncludeDir; |
| 1501 | // Be defensive about the contents of Entry. |
| 1502 | if (getVersion() >= 5) { |
| 1503 | // DirIdx 0 is the compilation directory, so don't include it for |
| 1504 | // relative names. |
| 1505 | if ((Entry.DirIdx != 0 || Kind != FileLineInfoKind::RelativeFilePath) && |
| 1506 | Entry.DirIdx < IncludeDirectories.size()) |
| 1507 | IncludeDir = dwarf::toStringRef(V: IncludeDirectories[Entry.DirIdx]); |
| 1508 | } else { |
| 1509 | if (0 < Entry.DirIdx && Entry.DirIdx <= IncludeDirectories.size()) |
| 1510 | IncludeDir = dwarf::toStringRef(V: IncludeDirectories[Entry.DirIdx - 1]); |
| 1511 | } |
| 1512 | |
| 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 && |
| 1518 | (getVersion() < 5 || Entry.DirIdx != 0) && !CompDir.empty() && |
| 1519 | !isPathAbsoluteOnWindowsOrPosix(Path: IncludeDir)) |
| 1520 | sys::path::append(path&: FilePath, style: Style, a: CompDir); |
| 1521 | |
| 1522 | assert((Kind == FileLineInfoKind::AbsoluteFilePath || |
| 1523 | Kind == FileLineInfoKind::RelativeFilePath) && |
| 1524 | "invalid FileLineInfo Kind" ); |
| 1525 | |
| 1526 | // sys::path::append skips empty strings. |
| 1527 | sys::path::append(path&: FilePath, style: Style, a: IncludeDir, b: FileName); |
| 1528 | Result = std::string(FilePath); |
| 1529 | return true; |
| 1530 | } |
| 1531 | |
| 1532 | bool DWARFDebugLine::LineTable::getFileLineInfoForAddress( |
| 1533 | object::SectionedAddress Address, bool Approximate, const char *CompDir, |
| 1534 | FileLineInfoKind Kind, DILineInfo &Result) const { |
| 1535 | // Get the index of row we're looking for in the line table. |
| 1536 | uint32_t RowIndex = |
| 1537 | lookupAddress(Address, IsApproximateLine: Approximate ? &Result.IsApproximateLine : nullptr); |
| 1538 | if (RowIndex == -1U) |
| 1539 | return false; |
| 1540 | // Take file number and line/column from the row. |
| 1541 | const auto &Row = Rows[RowIndex]; |
| 1542 | if (!getFileNameByIndex(FileIndex: Row.File, CompDir, Kind, Result&: Result.FileName)) |
| 1543 | return false; |
| 1544 | Result.Line = Row.Line; |
| 1545 | Result.Column = Row.Column; |
| 1546 | Result.Discriminator = Row.Discriminator; |
| 1547 | Result.Source = getSourceByIndex(FileIndex: Row.File, Kind); |
| 1548 | return true; |
| 1549 | } |
| 1550 | |
| 1551 | bool DWARFDebugLine::LineTable::getDirectoryForEntry( |
| 1552 | const FileNameEntry &Entry, std::string &Directory) const { |
| 1553 | if (Prologue.getVersion() >= 5) { |
| 1554 | if (Entry.DirIdx < Prologue.IncludeDirectories.size()) { |
| 1555 | Directory = |
| 1556 | dwarf::toString(V: Prologue.IncludeDirectories[Entry.DirIdx], Default: "" ); |
| 1557 | return true; |
| 1558 | } |
| 1559 | return false; |
| 1560 | } |
| 1561 | if (0 < Entry.DirIdx && Entry.DirIdx <= Prologue.IncludeDirectories.size()) { |
| 1562 | Directory = |
| 1563 | dwarf::toString(V: Prologue.IncludeDirectories[Entry.DirIdx - 1], Default: "" ); |
| 1564 | return true; |
| 1565 | } |
| 1566 | return false; |
| 1567 | } |
| 1568 | |
| 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. |
| 1573 | static DWARFDebugLine::SectionParser::LineToUnitMap |
| 1574 | buildLineToUnitMap(DWARFUnitVector::iterator_range Units) { |
| 1575 | DWARFDebugLine::SectionParser::LineToUnitMap LineToUnit; |
| 1576 | for (const auto &U : Units) |
| 1577 | if (auto CUDIE = U->getUnitDIE()) |
| 1578 | if (auto StmtOffset = toSectionOffset(V: CUDIE.find(Attr: DW_AT_stmt_list))) |
| 1579 | LineToUnit.insert(x: std::make_pair(x&: *StmtOffset, y: &*U)); |
| 1580 | return LineToUnit; |
| 1581 | } |
| 1582 | |
| 1583 | DWARFDebugLine::SectionParser::( |
| 1584 | DWARFDataExtractor &Data, const DWARFContext &C, |
| 1585 | DWARFUnitVector::iterator_range Units) |
| 1586 | : DebugLineData(Data), Context(C) { |
| 1587 | LineToUnit = buildLineToUnitMap(Units); |
| 1588 | if (!DebugLineData.isValidOffset(offset: Offset)) |
| 1589 | Done = true; |
| 1590 | } |
| 1591 | |
| 1592 | bool DWARFDebugLine::Prologue::totalLengthIsValid() const { |
| 1593 | return TotalLength != 0u; |
| 1594 | } |
| 1595 | |
| 1596 | DWARFDebugLine::LineTable DWARFDebugLine::SectionParser::parseNext( |
| 1597 | function_ref<void(Error)> RecoverableErrorHandler, |
| 1598 | function_ref<void(Error)> UnrecoverableErrorHandler, raw_ostream *OS, |
| 1599 | bool Verbose) { |
| 1600 | assert(DebugLineData.isValidOffset(Offset) && |
| 1601 | "parsing should have terminated" ); |
| 1602 | DWARFUnit *U = prepareToParse(Offset); |
| 1603 | uint64_t OldOffset = Offset; |
| 1604 | LineTable LT; |
| 1605 | if (Error Err = LT.parse(DebugLineData, OffsetPtr: &Offset, Ctx: Context, U, |
| 1606 | RecoverableErrorHandler, OS, Verbose)) |
| 1607 | UnrecoverableErrorHandler(std::move(Err)); |
| 1608 | moveToNextTable(OldOffset, P: LT.Prologue); |
| 1609 | return LT; |
| 1610 | } |
| 1611 | |
| 1612 | void DWARFDebugLine::SectionParser::skip( |
| 1613 | function_ref<void(Error)> RecoverableErrorHandler, |
| 1614 | function_ref<void(Error)> UnrecoverableErrorHandler) { |
| 1615 | assert(DebugLineData.isValidOffset(Offset) && |
| 1616 | "parsing should have terminated" ); |
| 1617 | DWARFUnit *U = prepareToParse(Offset); |
| 1618 | uint64_t OldOffset = Offset; |
| 1619 | LineTable LT; |
| 1620 | if (Error Err = LT.Prologue.parse(DebugLineData, OffsetPtr: &Offset, |
| 1621 | RecoverableErrorHandler, Ctx: Context, U)) |
| 1622 | UnrecoverableErrorHandler(std::move(Err)); |
| 1623 | moveToNextTable(OldOffset, P: LT.Prologue); |
| 1624 | } |
| 1625 | |
| 1626 | DWARFUnit *DWARFDebugLine::SectionParser::prepareToParse(uint64_t Offset) { |
| 1627 | DWARFUnit *U = nullptr; |
| 1628 | auto It = LineToUnit.find(x: Offset); |
| 1629 | if (It != LineToUnit.end()) |
| 1630 | U = It->second; |
| 1631 | DebugLineData.setAddressSize(U ? U->getAddressByteSize() : 0); |
| 1632 | return U; |
| 1633 | } |
| 1634 | |
| 1635 | bool DWARFDebugLine::SectionParser::hasValidVersion(uint64_t Offset) { |
| 1636 | DataExtractor::Cursor Cursor(Offset); |
| 1637 | auto [TotalLength, _] = DebugLineData.getInitialLength(C&: Cursor); |
| 1638 | DWARFDataExtractor (DebugLineData, Cursor.tell() + TotalLength); |
| 1639 | uint16_t Version = HeaderData.getU16(C&: Cursor); |
| 1640 | if (!Cursor) { |
| 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). |
| 1644 | consumeError(Err: Cursor.takeError()); |
| 1645 | return false; |
| 1646 | } |
| 1647 | return versionIsSupported(Version); |
| 1648 | } |
| 1649 | |
| 1650 | void DWARFDebugLine::SectionParser::moveToNextTable(uint64_t OldOffset, |
| 1651 | const Prologue &P) { |
| 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()) { |
| 1656 | Done = true; |
| 1657 | return; |
| 1658 | } |
| 1659 | |
| 1660 | Offset = OldOffset + P.TotalLength + P.sizeofTotalLength(); |
| 1661 | if (!DebugLineData.isValidOffset(offset: Offset)) { |
| 1662 | Done = true; |
| 1663 | return; |
| 1664 | } |
| 1665 | |
| 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)) |
| 1669 | return; |
| 1670 | |
| 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}) { |
| 1675 | uint64_t AlignedOffset = alignTo(Value: Offset, Align); |
| 1676 | if (!DebugLineData.isValidOffset(offset: AlignedOffset)) { |
| 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). |
| 1680 | Done = true; |
| 1681 | return; |
| 1682 | } |
| 1683 | if (hasValidVersion(Offset: AlignedOffset)) { |
| 1684 | Offset = AlignedOffset; |
| 1685 | break; |
| 1686 | } |
| 1687 | } |
| 1688 | } |
| 1689 | |