1 //===-- DWARFDebugLine.cpp ------------------------------------------------===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 #include "DWARFDebugLine.h"
11 #include "llvm/Support/Dwarf.h"
12 #include "llvm/Support/Format.h"
13 #include "llvm/Support/Path.h"
14 #include "llvm/Support/raw_ostream.h"
17 using namespace dwarf;
19 void DWARFDebugLine::Prologue::dump(raw_ostream &OS) const {
20 OS << "Line table prologue:\n"
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("default_is_stmt: %u\n", DefaultIsStmt)
26 << format(" line_base: %i\n", LineBase)
27 << format(" line_range: %u\n", LineRange)
28 << format(" opcode_base: %u\n", OpcodeBase);
30 for (uint32_t i = 0; i < StandardOpcodeLengths.size(); ++i)
31 OS << format("standard_opcode_lengths[%s] = %u\n", LNStandardString(i+1),
32 StandardOpcodeLengths[i]);
34 if (!IncludeDirectories.empty())
35 for (uint32_t i = 0; i < IncludeDirectories.size(); ++i)
36 OS << format("include_directories[%3u] = '", i+1)
37 << IncludeDirectories[i] << "'\n";
39 if (!FileNames.empty()) {
40 OS << " Dir Mod Time File Len File Name\n"
41 << " ---- ---------- ---------- -----------"
43 for (uint32_t i = 0; i < FileNames.size(); ++i) {
44 const FileNameEntry& fileEntry = FileNames[i];
45 OS << format("file_names[%3u] %4" PRIu64 " ", i+1, fileEntry.DirIdx)
46 << format("0x%8.8" PRIx64 " 0x%8.8" PRIx64 " ",
47 fileEntry.ModTime, fileEntry.Length)
48 << fileEntry.Name << '\n';
53 void DWARFDebugLine::Row::postAppend() {
56 EpilogueBegin = false;
59 void DWARFDebugLine::Row::reset(bool default_is_stmt) {
65 IsStmt = default_is_stmt;
69 EpilogueBegin = false;
72 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
73 OS << format("0x%16.16" PRIx64 " %6u %6u", Address, Line, Column)
74 << format(" %6u %3u ", File, Isa)
75 << (IsStmt ? " is_stmt" : "")
76 << (BasicBlock ? " basic_block" : "")
77 << (PrologueEnd ? " prologue_end" : "")
78 << (EpilogueBegin ? " epilogue_begin" : "")
79 << (EndSequence ? " end_sequence" : "")
83 void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const {
88 OS << "Address Line Column File ISA Flags\n"
89 << "------------------ ------ ------ ------ --- -------------\n";
90 for (std::vector<Row>::const_iterator pos = Rows.begin(),
91 end = Rows.end(); pos != end; ++pos)
96 DWARFDebugLine::State::~State() {}
98 void DWARFDebugLine::State::appendRowToMatrix(uint32_t offset) {
99 if (Sequence::Empty) {
100 // Record the beginning of instruction sequence.
101 Sequence::Empty = false;
102 Sequence::LowPC = Address;
103 Sequence::FirstRowIndex = row;
105 ++row; // Increase the row number.
106 LineTable::appendRow(*this);
108 // Record the end of instruction sequence.
109 Sequence::HighPC = Address;
110 Sequence::LastRowIndex = row;
111 if (Sequence::isValid())
112 LineTable::appendSequence(*this);
118 void DWARFDebugLine::State::finalize() {
119 row = DoneParsingLineTable;
120 if (!Sequence::Empty) {
121 fprintf(stderr, "warning: last sequence in debug line table is not"
124 // Sort all sequences so that address lookup will work faster.
125 if (!Sequences.empty()) {
126 std::sort(Sequences.begin(), Sequences.end(), Sequence::orderByLowPC);
127 // Note: actually, instruction address ranges of sequences should not
128 // overlap (in shared objects and executables). If they do, the address
129 // lookup would still work, though, but result would be ambiguous.
130 // We don't report warning in this case. For example,
131 // sometimes .so compiled from multiple object files contains a few
132 // rudimentary sequences for address ranges [0x0, 0xsomething).
136 DWARFDebugLine::DumpingState::~DumpingState() {}
138 void DWARFDebugLine::DumpingState::finalize() {
142 const DWARFDebugLine::LineTable *
143 DWARFDebugLine::getLineTable(uint32_t offset) const {
144 LineTableConstIter pos = LineTableMap.find(offset);
145 if (pos != LineTableMap.end())
150 const DWARFDebugLine::LineTable *
151 DWARFDebugLine::getOrParseLineTable(DataExtractor debug_line_data,
153 std::pair<LineTableIter, bool> pos =
154 LineTableMap.insert(LineTableMapTy::value_type(offset, LineTable()));
156 // Parse and cache the line table for at this offset.
158 if (!parseStatementTable(debug_line_data, &offset, state))
160 pos.first->second = state;
162 return &pos.first->second;
166 DWARFDebugLine::parsePrologue(DataExtractor debug_line_data,
167 uint32_t *offset_ptr, Prologue *prologue) {
168 const uint32_t prologue_offset = *offset_ptr;
171 prologue->TotalLength = debug_line_data.getU32(offset_ptr);
172 prologue->Version = debug_line_data.getU16(offset_ptr);
173 if (prologue->Version != 2)
176 prologue->PrologueLength = debug_line_data.getU32(offset_ptr);
177 const uint32_t end_prologue_offset = prologue->PrologueLength + *offset_ptr;
178 prologue->MinInstLength = debug_line_data.getU8(offset_ptr);
179 prologue->DefaultIsStmt = debug_line_data.getU8(offset_ptr);
180 prologue->LineBase = debug_line_data.getU8(offset_ptr);
181 prologue->LineRange = debug_line_data.getU8(offset_ptr);
182 prologue->OpcodeBase = debug_line_data.getU8(offset_ptr);
184 prologue->StandardOpcodeLengths.reserve(prologue->OpcodeBase-1);
185 for (uint32_t i = 1; i < prologue->OpcodeBase; ++i) {
186 uint8_t op_len = debug_line_data.getU8(offset_ptr);
187 prologue->StandardOpcodeLengths.push_back(op_len);
190 while (*offset_ptr < end_prologue_offset) {
191 const char *s = debug_line_data.getCStr(offset_ptr);
193 prologue->IncludeDirectories.push_back(s);
198 while (*offset_ptr < end_prologue_offset) {
199 const char *name = debug_line_data.getCStr(offset_ptr);
200 if (name && name[0]) {
201 FileNameEntry fileEntry;
202 fileEntry.Name = name;
203 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
204 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
205 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
206 prologue->FileNames.push_back(fileEntry);
212 if (*offset_ptr != end_prologue_offset) {
213 fprintf(stderr, "warning: parsing line table prologue at 0x%8.8x should"
214 " have ended at 0x%8.8x but it ended ad 0x%8.8x\n",
215 prologue_offset, end_prologue_offset, *offset_ptr);
222 DWARFDebugLine::parseStatementTable(DataExtractor debug_line_data,
223 uint32_t *offset_ptr, State &state) {
224 const uint32_t debug_line_offset = *offset_ptr;
226 Prologue *prologue = &state.Prologue;
228 if (!parsePrologue(debug_line_data, offset_ptr, prologue)) {
229 // Restore our offset and return false to indicate failure!
230 *offset_ptr = debug_line_offset;
234 const uint32_t end_offset = debug_line_offset + prologue->TotalLength +
235 sizeof(prologue->TotalLength);
239 while (*offset_ptr < end_offset) {
240 uint8_t opcode = debug_line_data.getU8(offset_ptr);
243 // Extended Opcodes always start with a zero opcode followed by
244 // a uleb128 length so you can skip ones you don't know about
245 uint32_t ext_offset = *offset_ptr;
246 uint64_t len = debug_line_data.getULEB128(offset_ptr);
247 uint32_t arg_size = len - (*offset_ptr - ext_offset);
249 uint8_t sub_opcode = debug_line_data.getU8(offset_ptr);
250 switch (sub_opcode) {
251 case DW_LNE_end_sequence:
252 // Set the end_sequence register of the state machine to true and
253 // append a row to the matrix using the current values of the
254 // state-machine registers. Then reset the registers to the initial
255 // values specified above. Every statement program sequence must end
256 // with a DW_LNE_end_sequence instruction which creates a row whose
257 // address is that of the byte after the last target machine instruction
259 state.EndSequence = true;
260 state.appendRowToMatrix(*offset_ptr);
264 case DW_LNE_set_address:
265 // Takes a single relocatable address as an operand. The size of the
266 // operand is the size appropriate to hold an address on the target
267 // machine. Set the address register to the value given by the
268 // relocatable address. All of the other statement program opcodes
269 // that affect the address register add a delta to it. This instruction
270 // stores a relocatable value into it instead.
271 state.Address = debug_line_data.getAddress(offset_ptr);
274 case DW_LNE_define_file:
275 // Takes 4 arguments. The first is a null terminated string containing
276 // a source file name. The second is an unsigned LEB128 number
277 // representing the directory index of the directory in which the file
278 // was found. The third is an unsigned LEB128 number representing the
279 // time of last modification of the file. The fourth is an unsigned
280 // LEB128 number representing the length in bytes of the file. The time
281 // and length fields may contain LEB128(0) if the information is not
284 // The directory index represents an entry in the include_directories
285 // section of the statement program prologue. The index is LEB128(0)
286 // if the file was found in the current directory of the compilation,
287 // LEB128(1) if it was found in the first directory in the
288 // include_directories section, and so on. The directory index is
289 // ignored for file names that represent full path names.
291 // The files are numbered, starting at 1, in the order in which they
292 // appear; the names in the prologue come before names defined by
293 // the DW_LNE_define_file instruction. These numbers are used in the
294 // the file register of the state machine.
296 FileNameEntry fileEntry;
297 fileEntry.Name = debug_line_data.getCStr(offset_ptr);
298 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
299 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
300 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
301 prologue->FileNames.push_back(fileEntry);
306 // Length doesn't include the zero opcode byte or the length itself, but
307 // it does include the sub_opcode, so we have to adjust for that below
308 (*offset_ptr) += arg_size;
311 } else if (opcode < prologue->OpcodeBase) {
315 // Takes no arguments. Append a row to the matrix using the
316 // current values of the state-machine registers. Then set
317 // the basic_block register to false.
318 state.appendRowToMatrix(*offset_ptr);
321 case DW_LNS_advance_pc:
322 // Takes a single unsigned LEB128 operand, multiplies it by the
323 // min_inst_length field of the prologue, and adds the
324 // result to the address register of the state machine.
325 state.Address += debug_line_data.getULEB128(offset_ptr) *
326 prologue->MinInstLength;
329 case DW_LNS_advance_line:
330 // Takes a single signed LEB128 operand and adds that value to
331 // the line register of the state machine.
332 state.Line += debug_line_data.getSLEB128(offset_ptr);
335 case DW_LNS_set_file:
336 // Takes a single unsigned LEB128 operand and stores it in the file
337 // register of the state machine.
338 state.File = debug_line_data.getULEB128(offset_ptr);
341 case DW_LNS_set_column:
342 // Takes a single unsigned LEB128 operand and stores it in the
343 // column register of the state machine.
344 state.Column = debug_line_data.getULEB128(offset_ptr);
347 case DW_LNS_negate_stmt:
348 // Takes no arguments. Set the is_stmt register of the state
349 // machine to the logical negation of its current value.
350 state.IsStmt = !state.IsStmt;
353 case DW_LNS_set_basic_block:
354 // Takes no arguments. Set the basic_block register of the
355 // state machine to true
356 state.BasicBlock = true;
359 case DW_LNS_const_add_pc:
360 // Takes no arguments. Add to the address register of the state
361 // machine the address increment value corresponding to special
362 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
363 // when the statement program needs to advance the address by a
364 // small amount, it can use a single special opcode, which occupies
365 // a single byte. When it needs to advance the address by up to
366 // twice the range of the last special opcode, it can use
367 // DW_LNS_const_add_pc followed by a special opcode, for a total
368 // of two bytes. Only if it needs to advance the address by more
369 // than twice that range will it need to use both DW_LNS_advance_pc
370 // and a special opcode, requiring three or more bytes.
372 uint8_t adjust_opcode = 255 - prologue->OpcodeBase;
373 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
374 prologue->MinInstLength;
375 state.Address += addr_offset;
379 case DW_LNS_fixed_advance_pc:
380 // Takes a single uhalf operand. Add to the address register of
381 // the state machine the value of the (unencoded) operand. This
382 // is the only extended opcode that takes an argument that is not
383 // a variable length number. The motivation for DW_LNS_fixed_advance_pc
384 // is this: existing assemblers cannot emit DW_LNS_advance_pc or
385 // special opcodes because they cannot encode LEB128 numbers or
386 // judge when the computation of a special opcode overflows and
387 // requires the use of DW_LNS_advance_pc. Such assemblers, however,
388 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
389 state.Address += debug_line_data.getU16(offset_ptr);
392 case DW_LNS_set_prologue_end:
393 // Takes no arguments. Set the prologue_end register of the
394 // state machine to true
395 state.PrologueEnd = true;
398 case DW_LNS_set_epilogue_begin:
399 // Takes no arguments. Set the basic_block register of the
400 // state machine to true
401 state.EpilogueBegin = true;
405 // Takes a single unsigned LEB128 operand and stores it in the
406 // column register of the state machine.
407 state.Isa = debug_line_data.getULEB128(offset_ptr);
411 // Handle any unknown standard opcodes here. We know the lengths
412 // of such opcodes because they are specified in the prologue
413 // as a multiple of LEB128 operands for each opcode.
415 assert(opcode - 1U < prologue->StandardOpcodeLengths.size());
416 uint8_t opcode_length = prologue->StandardOpcodeLengths[opcode - 1];
417 for (uint8_t i=0; i<opcode_length; ++i)
418 debug_line_data.getULEB128(offset_ptr);
425 // A special opcode value is chosen based on the amount that needs
426 // to be added to the line and address registers. The maximum line
427 // increment for a special opcode is the value of the line_base
428 // field in the header, plus the value of the line_range field,
429 // minus 1 (line base + line range - 1). If the desired line
430 // increment is greater than the maximum line increment, a standard
431 // opcode must be used instead of a special opcode. The "address
432 // advance" is calculated by dividing the desired address increment
433 // by the minimum_instruction_length field from the header. The
434 // special opcode is then calculated using the following formula:
436 // opcode = (desired line increment - line_base) +
437 // (line_range * address advance) + opcode_base
439 // If the resulting opcode is greater than 255, a standard opcode
440 // must be used instead.
442 // To decode a special opcode, subtract the opcode_base from the
443 // opcode itself to give the adjusted opcode. The amount to
444 // increment the address register is the result of the adjusted
445 // opcode divided by the line_range multiplied by the
446 // minimum_instruction_length field from the header. That is:
448 // address increment = (adjusted opcode / line_range) *
449 // minimum_instruction_length
451 // The amount to increment the line register is the line_base plus
452 // the result of the adjusted opcode modulo the line_range. That is:
454 // line increment = line_base + (adjusted opcode % line_range)
456 uint8_t adjust_opcode = opcode - prologue->OpcodeBase;
457 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
458 prologue->MinInstLength;
459 int32_t line_offset = prologue->LineBase +
460 (adjust_opcode % prologue->LineRange);
461 state.Line += line_offset;
462 state.Address += addr_offset;
463 state.appendRowToMatrix(*offset_ptr);
473 DWARFDebugLine::LineTable::lookupAddress(uint64_t address) const {
474 uint32_t unknown_index = UINT32_MAX;
475 if (Sequences.empty())
476 return unknown_index;
477 // First, find an instruction sequence containing the given address.
478 DWARFDebugLine::Sequence sequence;
479 sequence.LowPC = address;
480 SequenceIter first_seq = Sequences.begin();
481 SequenceIter last_seq = Sequences.end();
482 SequenceIter seq_pos = std::lower_bound(first_seq, last_seq, sequence,
483 DWARFDebugLine::Sequence::orderByLowPC);
484 DWARFDebugLine::Sequence found_seq;
485 if (seq_pos == last_seq) {
486 found_seq = Sequences.back();
487 } else if (seq_pos->LowPC == address) {
488 found_seq = *seq_pos;
490 if (seq_pos == first_seq)
491 return unknown_index;
492 found_seq = *(seq_pos - 1);
494 if (!found_seq.containsPC(address))
495 return unknown_index;
496 // Search for instruction address in the rows describing the sequence.
497 // Rows are stored in a vector, so we may use arithmetical operations with
499 DWARFDebugLine::Row row;
500 row.Address = address;
501 RowIter first_row = Rows.begin() + found_seq.FirstRowIndex;
502 RowIter last_row = Rows.begin() + found_seq.LastRowIndex;
503 RowIter row_pos = std::lower_bound(first_row, last_row, row,
504 DWARFDebugLine::Row::orderByAddress);
505 if (row_pos == last_row) {
506 return found_seq.LastRowIndex - 1;
508 uint32_t index = found_seq.FirstRowIndex + (row_pos - first_row);
509 if (row_pos->Address > address) {
510 if (row_pos == first_row)
511 return unknown_index;
519 DWARFDebugLine::LineTable::getFileNameByIndex(uint64_t FileIndex,
520 bool NeedsAbsoluteFilePath,
521 std::string &Result) const {
522 if (FileIndex == 0 || FileIndex > Prologue.FileNames.size())
524 const FileNameEntry &Entry = Prologue.FileNames[FileIndex - 1];
525 const char *FileName = Entry.Name;
526 if (!NeedsAbsoluteFilePath ||
527 sys::path::is_absolute(FileName)) {
531 SmallString<16> FilePath;
532 uint64_t IncludeDirIndex = Entry.DirIdx;
533 // Be defensive about the contents of Entry.
534 if (IncludeDirIndex > 0 &&
535 IncludeDirIndex <= Prologue.IncludeDirectories.size()) {
536 const char *IncludeDir = Prologue.IncludeDirectories[IncludeDirIndex - 1];
537 sys::path::append(FilePath, IncludeDir);
539 sys::path::append(FilePath, FileName);
540 Result = FilePath.str();