Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 1 | //===-- DWARFDebugLine.cpp ------------------------------------------------===// |
| 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
| 5 | // This file is distributed under the University of Illinois Open Source |
| 6 | // License. See LICENSE.TXT for details. |
| 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | |
| 10 | #include "DWARFDebugLine.h" |
| 11 | #include "llvm/Support/Dwarf.h" |
| 12 | #include "llvm/Support/Format.h" |
Alexey Samsonov | 38a6381 | 2012-08-30 07:49:50 +0000 | [diff] [blame] | 13 | #include "llvm/Support/Path.h" |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 14 | #include "llvm/Support/raw_ostream.h" |
Benjamin Kramer | 7393c7f | 2011-09-15 02:19:33 +0000 | [diff] [blame] | 15 | #include <algorithm> |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 16 | using namespace llvm; |
| 17 | using namespace dwarf; |
| 18 | |
| 19 | void DWARFDebugLine::Prologue::dump(raw_ostream &OS) const { |
| 20 | OS << "Line table prologue:\n" |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 21 | << format(" total_length: 0x%8.8x\n", TotalLength) |
| 22 | << format(" version: %u\n", Version) |
| 23 | << format(" prologue_length: 0x%8.8x\n", PrologueLength) |
| 24 | << format(" min_inst_length: %u\n", MinInstLength) |
| 25 | << format(Version >= 4 ? "max_ops_per_inst: %u\n" : "", MaxOpsPerInst) |
| 26 | << format(" default_is_stmt: %u\n", DefaultIsStmt) |
| 27 | << format(" line_base: %i\n", LineBase) |
| 28 | << format(" line_range: %u\n", LineRange) |
| 29 | << format(" opcode_base: %u\n", OpcodeBase); |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 30 | |
| 31 | for (uint32_t i = 0; i < StandardOpcodeLengths.size(); ++i) |
| 32 | OS << format("standard_opcode_lengths[%s] = %u\n", LNStandardString(i+1), |
| 33 | StandardOpcodeLengths[i]); |
| 34 | |
| 35 | if (!IncludeDirectories.empty()) |
| 36 | for (uint32_t i = 0; i < IncludeDirectories.size(); ++i) |
| 37 | OS << format("include_directories[%3u] = '", i+1) |
| 38 | << IncludeDirectories[i] << "'\n"; |
| 39 | |
| 40 | if (!FileNames.empty()) { |
| 41 | OS << " Dir Mod Time File Len File Name\n" |
| 42 | << " ---- ---------- ---------- -----------" |
| 43 | "----------------\n"; |
| 44 | for (uint32_t i = 0; i < FileNames.size(); ++i) { |
| 45 | const FileNameEntry& fileEntry = FileNames[i]; |
Benjamin Kramer | 5eccd36 | 2011-11-05 16:01:13 +0000 | [diff] [blame] | 46 | OS << format("file_names[%3u] %4" PRIu64 " ", i+1, fileEntry.DirIdx) |
| 47 | << format("0x%8.8" PRIx64 " 0x%8.8" PRIx64 " ", |
| 48 | fileEntry.ModTime, fileEntry.Length) |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 49 | << fileEntry.Name << '\n'; |
| 50 | } |
| 51 | } |
| 52 | } |
| 53 | |
| 54 | void DWARFDebugLine::Row::postAppend() { |
| 55 | BasicBlock = false; |
| 56 | PrologueEnd = false; |
| 57 | EpilogueBegin = false; |
| 58 | } |
| 59 | |
| 60 | void DWARFDebugLine::Row::reset(bool default_is_stmt) { |
| 61 | Address = 0; |
| 62 | Line = 1; |
| 63 | Column = 0; |
| 64 | File = 1; |
| 65 | Isa = 0; |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 66 | Discriminator = 0; |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 67 | IsStmt = default_is_stmt; |
| 68 | BasicBlock = false; |
| 69 | EndSequence = false; |
| 70 | PrologueEnd = false; |
| 71 | EpilogueBegin = false; |
| 72 | } |
| 73 | |
| 74 | void DWARFDebugLine::Row::dump(raw_ostream &OS) const { |
Benjamin Kramer | 41a9649 | 2011-11-05 08:57:40 +0000 | [diff] [blame] | 75 | OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column) |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 76 | << format(" %6u %3u %13u ", File, Isa, Discriminator) |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 77 | << (IsStmt ? " is_stmt" : "") |
| 78 | << (BasicBlock ? " basic_block" : "") |
| 79 | << (PrologueEnd ? " prologue_end" : "") |
| 80 | << (EpilogueBegin ? " epilogue_begin" : "") |
| 81 | << (EndSequence ? " end_sequence" : "") |
| 82 | << '\n'; |
| 83 | } |
| 84 | |
| 85 | void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const { |
| 86 | Prologue.dump(OS); |
| 87 | OS << '\n'; |
| 88 | |
| 89 | if (!Rows.empty()) { |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 90 | OS << "Address Line Column File ISA Discriminator Flags\n" |
| 91 | << "------------------ ------ ------ ------ --- ------------- " |
| 92 | "-------------\n"; |
| 93 | for (const Row &R : Rows) { |
| 94 | R.dump(OS); |
| 95 | } |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 96 | } |
| 97 | } |
| 98 | |
Nick Lewycky | 6bc4e71 | 2011-09-15 03:41:51 +0000 | [diff] [blame] | 99 | DWARFDebugLine::State::~State() {} |
| 100 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 101 | void DWARFDebugLine::State::appendRowToMatrix(uint32_t offset) { |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 102 | if (Sequence::Empty) { |
| 103 | // Record the beginning of instruction sequence. |
| 104 | Sequence::Empty = false; |
| 105 | Sequence::LowPC = Address; |
| 106 | Sequence::FirstRowIndex = row; |
| 107 | } |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 108 | ++row; // Increase the row number. |
| 109 | LineTable::appendRow(*this); |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 110 | if (EndSequence) { |
| 111 | // Record the end of instruction sequence. |
| 112 | Sequence::HighPC = Address; |
| 113 | Sequence::LastRowIndex = row; |
| 114 | if (Sequence::isValid()) |
| 115 | LineTable::appendSequence(*this); |
| 116 | Sequence::reset(); |
| 117 | } |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 118 | Row::postAppend(); |
| 119 | } |
| 120 | |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 121 | void DWARFDebugLine::State::finalize() { |
| 122 | row = DoneParsingLineTable; |
| 123 | if (!Sequence::Empty) { |
| 124 | fprintf(stderr, "warning: last sequence in debug line table is not" |
| 125 | "terminated!\n"); |
| 126 | } |
| 127 | // Sort all sequences so that address lookup will work faster. |
| 128 | if (!Sequences.empty()) { |
| 129 | std::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC); |
| 130 | // Note: actually, instruction address ranges of sequences should not |
| 131 | // overlap (in shared objects and executables). If they do, the address |
| 132 | // lookup would still work, though, but result would be ambiguous. |
| 133 | // We don't report warning in this case. For example, |
| 134 | // sometimes .so compiled from multiple object files contains a few |
| 135 | // rudimentary sequences for address ranges [0x0, 0xsomething). |
| 136 | } |
| 137 | } |
| 138 | |
Nick Lewycky | 6bc4e71 | 2011-09-15 03:41:51 +0000 | [diff] [blame] | 139 | DWARFDebugLine::DumpingState::~DumpingState() {} |
| 140 | |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 141 | void DWARFDebugLine::DumpingState::finalize() { |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 142 | LineTable::dump(OS); |
| 143 | } |
| 144 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 145 | const DWARFDebugLine::LineTable * |
| 146 | DWARFDebugLine::getLineTable(uint32_t offset) const { |
| 147 | LineTableConstIter pos = LineTableMap.find(offset); |
| 148 | if (pos != LineTableMap.end()) |
| 149 | return &pos->second; |
| 150 | return 0; |
| 151 | } |
| 152 | |
Benjamin Kramer | c26ed9b | 2011-09-15 20:43:18 +0000 | [diff] [blame] | 153 | const DWARFDebugLine::LineTable * |
| 154 | DWARFDebugLine::getOrParseLineTable(DataExtractor debug_line_data, |
| 155 | uint32_t offset) { |
Benjamin Kramer | 1d13d9e | 2011-09-21 17:31:42 +0000 | [diff] [blame] | 156 | std::pair<LineTableIter, bool> pos = |
| 157 | LineTableMap.insert(LineTableMapTy::value_type(offset, LineTable())); |
| 158 | if (pos.second) { |
Benjamin Kramer | c26ed9b | 2011-09-15 20:43:18 +0000 | [diff] [blame] | 159 | // Parse and cache the line table for at this offset. |
| 160 | State state; |
Andrew Kaylor | ee7c0d2 | 2013-01-25 22:50:58 +0000 | [diff] [blame] | 161 | if (!parseStatementTable(debug_line_data, RelocMap, &offset, state)) |
Benjamin Kramer | c26ed9b | 2011-09-15 20:43:18 +0000 | [diff] [blame] | 162 | return 0; |
Benjamin Kramer | 1d13d9e | 2011-09-21 17:31:42 +0000 | [diff] [blame] | 163 | pos.first->second = state; |
Benjamin Kramer | c26ed9b | 2011-09-15 20:43:18 +0000 | [diff] [blame] | 164 | } |
Benjamin Kramer | 1d13d9e | 2011-09-21 17:31:42 +0000 | [diff] [blame] | 165 | return &pos.first->second; |
Benjamin Kramer | c26ed9b | 2011-09-15 20:43:18 +0000 | [diff] [blame] | 166 | } |
| 167 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 168 | bool |
| 169 | DWARFDebugLine::parsePrologue(DataExtractor debug_line_data, |
| 170 | uint32_t *offset_ptr, Prologue *prologue) { |
| 171 | const uint32_t prologue_offset = *offset_ptr; |
| 172 | |
| 173 | prologue->clear(); |
| 174 | prologue->TotalLength = debug_line_data.getU32(offset_ptr); |
| 175 | prologue->Version = debug_line_data.getU16(offset_ptr); |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 176 | if (prologue->Version < 2) |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 177 | return false; |
| 178 | |
| 179 | prologue->PrologueLength = debug_line_data.getU32(offset_ptr); |
| 180 | const uint32_t end_prologue_offset = prologue->PrologueLength + *offset_ptr; |
| 181 | prologue->MinInstLength = debug_line_data.getU8(offset_ptr); |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 182 | if (prologue->Version >= 4) |
| 183 | prologue->MaxOpsPerInst = debug_line_data.getU8(offset_ptr); |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 184 | prologue->DefaultIsStmt = debug_line_data.getU8(offset_ptr); |
| 185 | prologue->LineBase = debug_line_data.getU8(offset_ptr); |
| 186 | prologue->LineRange = debug_line_data.getU8(offset_ptr); |
| 187 | prologue->OpcodeBase = debug_line_data.getU8(offset_ptr); |
| 188 | |
| 189 | prologue->StandardOpcodeLengths.reserve(prologue->OpcodeBase-1); |
| 190 | for (uint32_t i = 1; i < prologue->OpcodeBase; ++i) { |
| 191 | uint8_t op_len = debug_line_data.getU8(offset_ptr); |
| 192 | prologue->StandardOpcodeLengths.push_back(op_len); |
| 193 | } |
| 194 | |
| 195 | while (*offset_ptr < end_prologue_offset) { |
| 196 | const char *s = debug_line_data.getCStr(offset_ptr); |
| 197 | if (s && s[0]) |
| 198 | prologue->IncludeDirectories.push_back(s); |
| 199 | else |
| 200 | break; |
| 201 | } |
| 202 | |
| 203 | while (*offset_ptr < end_prologue_offset) { |
| 204 | const char *name = debug_line_data.getCStr(offset_ptr); |
| 205 | if (name && name[0]) { |
| 206 | FileNameEntry fileEntry; |
| 207 | fileEntry.Name = name; |
| 208 | fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr); |
| 209 | fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr); |
| 210 | fileEntry.Length = debug_line_data.getULEB128(offset_ptr); |
| 211 | prologue->FileNames.push_back(fileEntry); |
| 212 | } else { |
| 213 | break; |
| 214 | } |
| 215 | } |
| 216 | |
| 217 | if (*offset_ptr != end_prologue_offset) { |
| 218 | fprintf(stderr, "warning: parsing line table prologue at 0x%8.8x should" |
Ed Maste | e1bc6dd | 2013-10-18 13:01:33 +0000 | [diff] [blame] | 219 | " have ended at 0x%8.8x but it ended at 0x%8.8x\n", |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 220 | prologue_offset, end_prologue_offset, *offset_ptr); |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 221 | return false; |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 222 | } |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 223 | return true; |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 224 | } |
| 225 | |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 226 | bool DWARFDebugLine::parseStatementTable(DataExtractor debug_line_data, |
| 227 | const RelocAddrMap *RMap, |
| 228 | uint32_t *offset_ptr, State &state) { |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 229 | const uint32_t debug_line_offset = *offset_ptr; |
| 230 | |
| 231 | Prologue *prologue = &state.Prologue; |
| 232 | |
| 233 | if (!parsePrologue(debug_line_data, offset_ptr, prologue)) { |
| 234 | // Restore our offset and return false to indicate failure! |
| 235 | *offset_ptr = debug_line_offset; |
| 236 | return false; |
| 237 | } |
| 238 | |
| 239 | const uint32_t end_offset = debug_line_offset + prologue->TotalLength + |
| 240 | sizeof(prologue->TotalLength); |
| 241 | |
Benjamin Kramer | 9013db3 | 2011-09-15 21:59:13 +0000 | [diff] [blame] | 242 | state.reset(); |
| 243 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 244 | while (*offset_ptr < end_offset) { |
| 245 | uint8_t opcode = debug_line_data.getU8(offset_ptr); |
| 246 | |
| 247 | if (opcode == 0) { |
| 248 | // Extended Opcodes always start with a zero opcode followed by |
| 249 | // a uleb128 length so you can skip ones you don't know about |
| 250 | uint32_t ext_offset = *offset_ptr; |
| 251 | uint64_t len = debug_line_data.getULEB128(offset_ptr); |
| 252 | uint32_t arg_size = len - (*offset_ptr - ext_offset); |
| 253 | |
| 254 | uint8_t sub_opcode = debug_line_data.getU8(offset_ptr); |
| 255 | switch (sub_opcode) { |
| 256 | case DW_LNE_end_sequence: |
| 257 | // Set the end_sequence register of the state machine to true and |
| 258 | // append a row to the matrix using the current values of the |
| 259 | // state-machine registers. Then reset the registers to the initial |
| 260 | // values specified above. Every statement program sequence must end |
| 261 | // with a DW_LNE_end_sequence instruction which creates a row whose |
| 262 | // address is that of the byte after the last target machine instruction |
| 263 | // of the sequence. |
| 264 | state.EndSequence = true; |
| 265 | state.appendRowToMatrix(*offset_ptr); |
| 266 | state.reset(); |
| 267 | break; |
| 268 | |
| 269 | case DW_LNE_set_address: |
| 270 | // Takes a single relocatable address as an operand. The size of the |
| 271 | // operand is the size appropriate to hold an address on the target |
| 272 | // machine. Set the address register to the value given by the |
| 273 | // relocatable address. All of the other statement program opcodes |
| 274 | // that affect the address register add a delta to it. This instruction |
| 275 | // stores a relocatable value into it instead. |
Andrew Kaylor | ee7c0d2 | 2013-01-25 22:50:58 +0000 | [diff] [blame] | 276 | { |
| 277 | // If this address is in our relocation map, apply the relocation. |
| 278 | RelocAddrMap::const_iterator AI = RMap->find(*offset_ptr); |
| 279 | if (AI != RMap->end()) { |
| 280 | const std::pair<uint8_t, int64_t> &R = AI->second; |
| 281 | state.Address = debug_line_data.getAddress(offset_ptr) + R.second; |
| 282 | } else |
| 283 | state.Address = debug_line_data.getAddress(offset_ptr); |
| 284 | } |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 285 | break; |
| 286 | |
| 287 | case DW_LNE_define_file: |
| 288 | // Takes 4 arguments. The first is a null terminated string containing |
| 289 | // a source file name. The second is an unsigned LEB128 number |
| 290 | // representing the directory index of the directory in which the file |
| 291 | // was found. The third is an unsigned LEB128 number representing the |
| 292 | // time of last modification of the file. The fourth is an unsigned |
| 293 | // LEB128 number representing the length in bytes of the file. The time |
| 294 | // and length fields may contain LEB128(0) if the information is not |
| 295 | // available. |
| 296 | // |
| 297 | // The directory index represents an entry in the include_directories |
| 298 | // section of the statement program prologue. The index is LEB128(0) |
| 299 | // if the file was found in the current directory of the compilation, |
| 300 | // LEB128(1) if it was found in the first directory in the |
| 301 | // include_directories section, and so on. The directory index is |
| 302 | // ignored for file names that represent full path names. |
| 303 | // |
| 304 | // The files are numbered, starting at 1, in the order in which they |
| 305 | // appear; the names in the prologue come before names defined by |
| 306 | // the DW_LNE_define_file instruction. These numbers are used in the |
| 307 | // the file register of the state machine. |
| 308 | { |
| 309 | FileNameEntry fileEntry; |
| 310 | fileEntry.Name = debug_line_data.getCStr(offset_ptr); |
| 311 | fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr); |
| 312 | fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr); |
| 313 | fileEntry.Length = debug_line_data.getULEB128(offset_ptr); |
| 314 | prologue->FileNames.push_back(fileEntry); |
| 315 | } |
| 316 | break; |
| 317 | |
Stephen Hines | 36b5688 | 2014-04-23 16:57:46 -0700 | [diff] [blame^] | 318 | case DW_LNE_set_discriminator: |
| 319 | state.Discriminator = debug_line_data.getULEB128(offset_ptr); |
| 320 | break; |
| 321 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 322 | default: |
| 323 | // Length doesn't include the zero opcode byte or the length itself, but |
| 324 | // it does include the sub_opcode, so we have to adjust for that below |
| 325 | (*offset_ptr) += arg_size; |
| 326 | break; |
| 327 | } |
| 328 | } else if (opcode < prologue->OpcodeBase) { |
| 329 | switch (opcode) { |
| 330 | // Standard Opcodes |
| 331 | case DW_LNS_copy: |
| 332 | // Takes no arguments. Append a row to the matrix using the |
| 333 | // current values of the state-machine registers. Then set |
| 334 | // the basic_block register to false. |
| 335 | state.appendRowToMatrix(*offset_ptr); |
| 336 | break; |
| 337 | |
| 338 | case DW_LNS_advance_pc: |
| 339 | // Takes a single unsigned LEB128 operand, multiplies it by the |
| 340 | // min_inst_length field of the prologue, and adds the |
| 341 | // result to the address register of the state machine. |
| 342 | state.Address += debug_line_data.getULEB128(offset_ptr) * |
| 343 | prologue->MinInstLength; |
| 344 | break; |
| 345 | |
| 346 | case DW_LNS_advance_line: |
| 347 | // Takes a single signed LEB128 operand and adds that value to |
| 348 | // the line register of the state machine. |
| 349 | state.Line += debug_line_data.getSLEB128(offset_ptr); |
| 350 | break; |
| 351 | |
| 352 | case DW_LNS_set_file: |
| 353 | // Takes a single unsigned LEB128 operand and stores it in the file |
| 354 | // register of the state machine. |
| 355 | state.File = debug_line_data.getULEB128(offset_ptr); |
| 356 | break; |
| 357 | |
| 358 | case DW_LNS_set_column: |
| 359 | // Takes a single unsigned LEB128 operand and stores it in the |
| 360 | // column register of the state machine. |
| 361 | state.Column = debug_line_data.getULEB128(offset_ptr); |
| 362 | break; |
| 363 | |
| 364 | case DW_LNS_negate_stmt: |
| 365 | // Takes no arguments. Set the is_stmt register of the state |
| 366 | // machine to the logical negation of its current value. |
| 367 | state.IsStmt = !state.IsStmt; |
| 368 | break; |
| 369 | |
| 370 | case DW_LNS_set_basic_block: |
| 371 | // Takes no arguments. Set the basic_block register of the |
| 372 | // state machine to true |
| 373 | state.BasicBlock = true; |
| 374 | break; |
| 375 | |
| 376 | case DW_LNS_const_add_pc: |
| 377 | // Takes no arguments. Add to the address register of the state |
| 378 | // machine the address increment value corresponding to special |
| 379 | // opcode 255. The motivation for DW_LNS_const_add_pc is this: |
| 380 | // when the statement program needs to advance the address by a |
| 381 | // small amount, it can use a single special opcode, which occupies |
| 382 | // a single byte. When it needs to advance the address by up to |
| 383 | // twice the range of the last special opcode, it can use |
| 384 | // DW_LNS_const_add_pc followed by a special opcode, for a total |
| 385 | // of two bytes. Only if it needs to advance the address by more |
| 386 | // than twice that range will it need to use both DW_LNS_advance_pc |
| 387 | // and a special opcode, requiring three or more bytes. |
| 388 | { |
| 389 | uint8_t adjust_opcode = 255 - prologue->OpcodeBase; |
| 390 | uint64_t addr_offset = (adjust_opcode / prologue->LineRange) * |
| 391 | prologue->MinInstLength; |
| 392 | state.Address += addr_offset; |
| 393 | } |
| 394 | break; |
| 395 | |
| 396 | case DW_LNS_fixed_advance_pc: |
| 397 | // Takes a single uhalf operand. Add to the address register of |
| 398 | // the state machine the value of the (unencoded) operand. This |
| 399 | // is the only extended opcode that takes an argument that is not |
| 400 | // a variable length number. The motivation for DW_LNS_fixed_advance_pc |
| 401 | // is this: existing assemblers cannot emit DW_LNS_advance_pc or |
| 402 | // special opcodes because they cannot encode LEB128 numbers or |
| 403 | // judge when the computation of a special opcode overflows and |
| 404 | // requires the use of DW_LNS_advance_pc. Such assemblers, however, |
| 405 | // can use DW_LNS_fixed_advance_pc instead, sacrificing compression. |
| 406 | state.Address += debug_line_data.getU16(offset_ptr); |
| 407 | break; |
| 408 | |
| 409 | case DW_LNS_set_prologue_end: |
| 410 | // Takes no arguments. Set the prologue_end register of the |
| 411 | // state machine to true |
| 412 | state.PrologueEnd = true; |
| 413 | break; |
| 414 | |
| 415 | case DW_LNS_set_epilogue_begin: |
| 416 | // Takes no arguments. Set the basic_block register of the |
| 417 | // state machine to true |
| 418 | state.EpilogueBegin = true; |
| 419 | break; |
| 420 | |
| 421 | case DW_LNS_set_isa: |
| 422 | // Takes a single unsigned LEB128 operand and stores it in the |
| 423 | // column register of the state machine. |
| 424 | state.Isa = debug_line_data.getULEB128(offset_ptr); |
| 425 | break; |
| 426 | |
| 427 | default: |
| 428 | // Handle any unknown standard opcodes here. We know the lengths |
| 429 | // of such opcodes because they are specified in the prologue |
| 430 | // as a multiple of LEB128 operands for each opcode. |
| 431 | { |
Benjamin Kramer | 068d9a5 | 2011-09-15 03:20:04 +0000 | [diff] [blame] | 432 | assert(opcode - 1U < prologue->StandardOpcodeLengths.size()); |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 433 | uint8_t opcode_length = prologue->StandardOpcodeLengths[opcode - 1]; |
| 434 | for (uint8_t i=0; i<opcode_length; ++i) |
| 435 | debug_line_data.getULEB128(offset_ptr); |
| 436 | } |
| 437 | break; |
| 438 | } |
| 439 | } else { |
| 440 | // Special Opcodes |
| 441 | |
| 442 | // A special opcode value is chosen based on the amount that needs |
| 443 | // to be added to the line and address registers. The maximum line |
| 444 | // increment for a special opcode is the value of the line_base |
| 445 | // field in the header, plus the value of the line_range field, |
| 446 | // minus 1 (line base + line range - 1). If the desired line |
| 447 | // increment is greater than the maximum line increment, a standard |
NAKAMURA Takumi | 017449d | 2011-10-08 11:22:47 +0000 | [diff] [blame] | 448 | // opcode must be used instead of a special opcode. The "address |
| 449 | // advance" is calculated by dividing the desired address increment |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 450 | // by the minimum_instruction_length field from the header. The |
| 451 | // special opcode is then calculated using the following formula: |
| 452 | // |
| 453 | // opcode = (desired line increment - line_base) + |
| 454 | // (line_range * address advance) + opcode_base |
| 455 | // |
| 456 | // If the resulting opcode is greater than 255, a standard opcode |
| 457 | // must be used instead. |
| 458 | // |
| 459 | // To decode a special opcode, subtract the opcode_base from the |
| 460 | // opcode itself to give the adjusted opcode. The amount to |
| 461 | // increment the address register is the result of the adjusted |
| 462 | // opcode divided by the line_range multiplied by the |
| 463 | // minimum_instruction_length field from the header. That is: |
| 464 | // |
| 465 | // address increment = (adjusted opcode / line_range) * |
| 466 | // minimum_instruction_length |
| 467 | // |
| 468 | // The amount to increment the line register is the line_base plus |
| 469 | // the result of the adjusted opcode modulo the line_range. That is: |
| 470 | // |
| 471 | // line increment = line_base + (adjusted opcode % line_range) |
| 472 | |
| 473 | uint8_t adjust_opcode = opcode - prologue->OpcodeBase; |
| 474 | uint64_t addr_offset = (adjust_opcode / prologue->LineRange) * |
| 475 | prologue->MinInstLength; |
| 476 | int32_t line_offset = prologue->LineBase + |
| 477 | (adjust_opcode % prologue->LineRange); |
| 478 | state.Line += line_offset; |
| 479 | state.Address += addr_offset; |
| 480 | state.appendRowToMatrix(*offset_ptr); |
| 481 | } |
| 482 | } |
| 483 | |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 484 | state.finalize(); |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 485 | |
| 486 | return end_offset; |
| 487 | } |
| 488 | |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 489 | uint32_t |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 490 | DWARFDebugLine::LineTable::lookupAddress(uint64_t address) const { |
| 491 | uint32_t unknown_index = UINT32_MAX; |
| 492 | if (Sequences.empty()) |
| 493 | return unknown_index; |
| 494 | // First, find an instruction sequence containing the given address. |
| 495 | DWARFDebugLine::Sequence sequence; |
| 496 | sequence.LowPC = address; |
| 497 | SequenceIter first_seq = Sequences.begin(); |
| 498 | SequenceIter last_seq = Sequences.end(); |
| 499 | SequenceIter seq_pos = std::lower_bound(first_seq, last_seq, sequence, |
| 500 | DWARFDebugLine::Sequence::orderByLowPC); |
| 501 | DWARFDebugLine::Sequence found_seq; |
| 502 | if (seq_pos == last_seq) { |
| 503 | found_seq = Sequences.back(); |
| 504 | } else if (seq_pos->LowPC == address) { |
| 505 | found_seq = *seq_pos; |
| 506 | } else { |
| 507 | if (seq_pos == first_seq) |
| 508 | return unknown_index; |
| 509 | found_seq = *(seq_pos - 1); |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 510 | } |
Alexey Samsonov | 351f83b | 2012-08-07 11:46:57 +0000 | [diff] [blame] | 511 | if (!found_seq.containsPC(address)) |
| 512 | return unknown_index; |
| 513 | // Search for instruction address in the rows describing the sequence. |
| 514 | // Rows are stored in a vector, so we may use arithmetical operations with |
| 515 | // iterators. |
| 516 | DWARFDebugLine::Row row; |
| 517 | row.Address = address; |
| 518 | RowIter first_row = Rows.begin() + found_seq.FirstRowIndex; |
| 519 | RowIter last_row = Rows.begin() + found_seq.LastRowIndex; |
| 520 | RowIter row_pos = std::lower_bound(first_row, last_row, row, |
| 521 | DWARFDebugLine::Row::orderByAddress); |
| 522 | if (row_pos == last_row) { |
| 523 | return found_seq.LastRowIndex - 1; |
| 524 | } |
| 525 | uint32_t index = found_seq.FirstRowIndex + (row_pos - first_row); |
| 526 | if (row_pos->Address > address) { |
| 527 | if (row_pos == first_row) |
| 528 | return unknown_index; |
| 529 | else |
| 530 | index--; |
| 531 | } |
| 532 | return index; |
Benjamin Kramer | b848e97 | 2011-09-15 02:12:05 +0000 | [diff] [blame] | 533 | } |
Alexey Samsonov | 38a6381 | 2012-08-30 07:49:50 +0000 | [diff] [blame] | 534 | |
| 535 | bool |
Andrew Kaylor | e27a787 | 2013-01-26 00:28:05 +0000 | [diff] [blame] | 536 | DWARFDebugLine::LineTable::lookupAddressRange(uint64_t address, |
| 537 | uint64_t size, |
| 538 | std::vector<uint32_t>& result) const { |
| 539 | if (Sequences.empty()) |
| 540 | return false; |
| 541 | uint64_t end_addr = address + size; |
| 542 | // First, find an instruction sequence containing the given address. |
| 543 | DWARFDebugLine::Sequence sequence; |
| 544 | sequence.LowPC = address; |
| 545 | SequenceIter first_seq = Sequences.begin(); |
| 546 | SequenceIter last_seq = Sequences.end(); |
| 547 | SequenceIter seq_pos = std::lower_bound(first_seq, last_seq, sequence, |
| 548 | DWARFDebugLine::Sequence::orderByLowPC); |
| 549 | if (seq_pos == last_seq || seq_pos->LowPC != address) { |
| 550 | if (seq_pos == first_seq) |
| 551 | return false; |
| 552 | seq_pos--; |
| 553 | } |
| 554 | if (!seq_pos->containsPC(address)) |
| 555 | return false; |
| 556 | |
| 557 | SequenceIter start_pos = seq_pos; |
| 558 | |
| 559 | // Add the rows from the first sequence to the vector, starting with the |
| 560 | // index we just calculated |
| 561 | |
| 562 | while (seq_pos != last_seq && seq_pos->LowPC < end_addr) { |
| 563 | DWARFDebugLine::Sequence cur_seq = *seq_pos; |
| 564 | uint32_t first_row_index; |
| 565 | uint32_t last_row_index; |
| 566 | if (seq_pos == start_pos) { |
| 567 | // For the first sequence, we need to find which row in the sequence is the |
| 568 | // first in our range. Rows are stored in a vector, so we may use |
| 569 | // arithmetical operations with iterators. |
| 570 | DWARFDebugLine::Row row; |
| 571 | row.Address = address; |
| 572 | RowIter first_row = Rows.begin() + cur_seq.FirstRowIndex; |
| 573 | RowIter last_row = Rows.begin() + cur_seq.LastRowIndex; |
| 574 | RowIter row_pos = std::upper_bound(first_row, last_row, row, |
| 575 | DWARFDebugLine::Row::orderByAddress); |
| 576 | // The 'row_pos' iterator references the first row that is greater than |
| 577 | // our start address. Unless that's the first row, we want to start at |
| 578 | // the row before that. |
| 579 | first_row_index = cur_seq.FirstRowIndex + (row_pos - first_row); |
| 580 | if (row_pos != first_row) |
| 581 | --first_row_index; |
| 582 | } else |
| 583 | first_row_index = cur_seq.FirstRowIndex; |
| 584 | |
| 585 | // For the last sequence in our range, we need to figure out the last row in |
| 586 | // range. For all other sequences we can go to the end of the sequence. |
| 587 | if (cur_seq.HighPC > end_addr) { |
| 588 | DWARFDebugLine::Row row; |
| 589 | row.Address = end_addr; |
| 590 | RowIter first_row = Rows.begin() + cur_seq.FirstRowIndex; |
| 591 | RowIter last_row = Rows.begin() + cur_seq.LastRowIndex; |
| 592 | RowIter row_pos = std::upper_bound(first_row, last_row, row, |
| 593 | DWARFDebugLine::Row::orderByAddress); |
| 594 | // The 'row_pos' iterator references the first row that is greater than |
| 595 | // our end address. The row before that is the last row we want. |
| 596 | last_row_index = cur_seq.FirstRowIndex + (row_pos - first_row) - 1; |
| 597 | } else |
| 598 | // Contrary to what you might expect, DWARFDebugLine::SequenceLastRowIndex |
| 599 | // isn't a valid index within the current sequence. It's that plus one. |
| 600 | last_row_index = cur_seq.LastRowIndex - 1; |
| 601 | |
| 602 | for (uint32_t i = first_row_index; i <= last_row_index; ++i) { |
| 603 | result.push_back(i); |
| 604 | } |
| 605 | |
| 606 | ++seq_pos; |
| 607 | } |
NAKAMURA Takumi | 8090a24 | 2013-01-26 01:45:06 +0000 | [diff] [blame] | 608 | |
| 609 | return true; |
Andrew Kaylor | e27a787 | 2013-01-26 00:28:05 +0000 | [diff] [blame] | 610 | } |
| 611 | |
| 612 | bool |
Alexey Samsonov | 38a6381 | 2012-08-30 07:49:50 +0000 | [diff] [blame] | 613 | DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex, |
| 614 | bool NeedsAbsoluteFilePath, |
| 615 | std::string &Result) const { |
| 616 | if (FileIndex == 0 || FileIndex > Prologue.FileNames.size()) |
| 617 | return false; |
| 618 | const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1]; |
| 619 | const char *FileName = Entry.Name; |
| 620 | if (!NeedsAbsoluteFilePath || |
| 621 | sys::path::is_absolute(FileName)) { |
| 622 | Result = FileName; |
| 623 | return true; |
| 624 | } |
| 625 | SmallString<16> FilePath; |
| 626 | uint64_t IncludeDirIndex = Entry.DirIdx; |
| 627 | // Be defensive about the contents of Entry. |
| 628 | if (IncludeDirIndex > 0 && |
| 629 | IncludeDirIndex <= Prologue.IncludeDirectories.size()) { |
| 630 | const char *IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1]; |
| 631 | sys::path::append(FilePath, IncludeDir); |
| 632 | } |
| 633 | sys::path::append(FilePath, FileName); |
| 634 | Result = FilePath.str(); |
| 635 | return true; |
| 636 | } |