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/raw_ostream.h"
16 using namespace dwarf;
18 void DWARFDebugLine::Prologue::dump(raw_ostream &OS) const {
19 OS << "Line table prologue:\n"
20 << format(" total_length: 0x%8.8x\n", TotalLength)
21 << format(" version: %u\n", Version)
22 << format("prologue_length: 0x%8.8x\n", PrologueLength)
23 << format("min_inst_length: %u\n", MinInstLength)
24 << format("default_is_stmt: %u\n", DefaultIsStmt)
25 << format(" line_base: %i\n", LineBase)
26 << format(" line_range: %u\n", LineRange)
27 << format(" opcode_base: %u\n", OpcodeBase);
29 for (uint32_t i = 0; i < StandardOpcodeLengths.size(); ++i)
30 OS << format("standard_opcode_lengths[%s] = %u\n", LNStandardString(i+1),
31 StandardOpcodeLengths[i]);
33 if (!IncludeDirectories.empty())
34 for (uint32_t i = 0; i < IncludeDirectories.size(); ++i)
35 OS << format("include_directories[%3u] = '", i+1)
36 << IncludeDirectories[i] << "'\n";
38 if (!FileNames.empty()) {
39 OS << " Dir Mod Time File Len File Name\n"
40 << " ---- ---------- ---------- -----------"
42 for (uint32_t i = 0; i < FileNames.size(); ++i) {
43 const FileNameEntry& fileEntry = FileNames[i];
44 OS << format("file_names[%3u] %4u ", i+1, fileEntry.DirIdx)
45 << format("0x%8.8x 0x%8.8x ", fileEntry.ModTime, fileEntry.Length)
46 << fileEntry.Name << '\n';
51 void DWARFDebugLine::Row::postAppend() {
54 EpilogueBegin = false;
57 void DWARFDebugLine::Row::reset(bool default_is_stmt) {
63 IsStmt = default_is_stmt;
67 EpilogueBegin = false;
70 void DWARFDebugLine::Row::dump(raw_ostream &OS) const {
71 OS << format("0x%16.16llx %6u %6u", Address, Line, Column)
72 << format(" %6u %3u ", File, Isa)
73 << (IsStmt ? " is_stmt" : "")
74 << (BasicBlock ? " basic_block" : "")
75 << (PrologueEnd ? " prologue_end" : "")
76 << (EpilogueBegin ? " epilogue_begin" : "")
77 << (EndSequence ? " end_sequence" : "")
81 void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const {
86 OS << "Address Line Column File ISA Flags\n"
87 << "------------------ ------ ------ ------ --- -------------\n";
88 for (std::vector<Row>::const_iterator pos = Rows.begin(),
89 end = Rows.end(); pos != end; ++pos)
94 DWARFDebugLine::State::~State() {}
96 void DWARFDebugLine::State::appendRowToMatrix(uint32_t offset) {
97 ++row; // Increase the row number.
98 LineTable::appendRow(*this);
102 DWARFDebugLine::DumpingState::~DumpingState() {}
104 void DWARFDebugLine::DumpingState::finalize(uint32_t offset) {
108 const DWARFDebugLine::LineTable *
109 DWARFDebugLine::getLineTable(uint32_t offset) const {
110 LineTableConstIter pos = LineTableMap.find(offset);
111 if (pos != LineTableMap.end())
116 const DWARFDebugLine::LineTable *
117 DWARFDebugLine::getOrParseLineTable(DataExtractor debug_line_data,
119 LineTableIter pos = LineTableMap.find(offset);
120 if (pos == LineTableMap.end()) {
121 // Parse and cache the line table for at this offset.
123 if (!parseStatementTable(debug_line_data, &offset, state))
125 // FIXME: double lookup.
126 LineTableMap[offset] = state;
127 return &LineTableMap[offset];
133 DWARFDebugLine::parsePrologue(DataExtractor debug_line_data,
134 uint32_t *offset_ptr, Prologue *prologue) {
135 const uint32_t prologue_offset = *offset_ptr;
138 prologue->TotalLength = debug_line_data.getU32(offset_ptr);
139 prologue->Version = debug_line_data.getU16(offset_ptr);
140 if (prologue->Version != 2)
143 prologue->PrologueLength = debug_line_data.getU32(offset_ptr);
144 const uint32_t end_prologue_offset = prologue->PrologueLength + *offset_ptr;
145 prologue->MinInstLength = debug_line_data.getU8(offset_ptr);
146 prologue->DefaultIsStmt = debug_line_data.getU8(offset_ptr);
147 prologue->LineBase = debug_line_data.getU8(offset_ptr);
148 prologue->LineRange = debug_line_data.getU8(offset_ptr);
149 prologue->OpcodeBase = debug_line_data.getU8(offset_ptr);
151 prologue->StandardOpcodeLengths.reserve(prologue->OpcodeBase-1);
152 for (uint32_t i = 1; i < prologue->OpcodeBase; ++i) {
153 uint8_t op_len = debug_line_data.getU8(offset_ptr);
154 prologue->StandardOpcodeLengths.push_back(op_len);
157 while (*offset_ptr < end_prologue_offset) {
158 const char *s = debug_line_data.getCStr(offset_ptr);
160 prologue->IncludeDirectories.push_back(s);
165 while (*offset_ptr < end_prologue_offset) {
166 const char *name = debug_line_data.getCStr(offset_ptr);
167 if (name && name[0]) {
168 FileNameEntry fileEntry;
169 fileEntry.Name = name;
170 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
171 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
172 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
173 prologue->FileNames.push_back(fileEntry);
179 if (*offset_ptr != end_prologue_offset) {
180 fprintf(stderr, "warning: parsing line table prologue at 0x%8.8x should"
181 " have ended at 0x%8.8x but it ended ad 0x%8.8x\n",
182 prologue_offset, end_prologue_offset, *offset_ptr);
184 return end_prologue_offset;
188 DWARFDebugLine::parseStatementTable(DataExtractor debug_line_data,
189 uint32_t *offset_ptr, State &state) {
190 const uint32_t debug_line_offset = *offset_ptr;
192 Prologue *prologue = &state.Prologue;
194 if (!parsePrologue(debug_line_data, offset_ptr, prologue)) {
195 // Restore our offset and return false to indicate failure!
196 *offset_ptr = debug_line_offset;
200 const uint32_t end_offset = debug_line_offset + prologue->TotalLength +
201 sizeof(prologue->TotalLength);
203 while (*offset_ptr < end_offset) {
204 uint8_t opcode = debug_line_data.getU8(offset_ptr);
207 // Extended Opcodes always start with a zero opcode followed by
208 // a uleb128 length so you can skip ones you don't know about
209 uint32_t ext_offset = *offset_ptr;
210 uint64_t len = debug_line_data.getULEB128(offset_ptr);
211 uint32_t arg_size = len - (*offset_ptr - ext_offset);
213 uint8_t sub_opcode = debug_line_data.getU8(offset_ptr);
214 switch (sub_opcode) {
215 case DW_LNE_end_sequence:
216 // Set the end_sequence register of the state machine to true and
217 // append a row to the matrix using the current values of the
218 // state-machine registers. Then reset the registers to the initial
219 // values specified above. Every statement program sequence must end
220 // with a DW_LNE_end_sequence instruction which creates a row whose
221 // address is that of the byte after the last target machine instruction
223 state.EndSequence = true;
224 state.appendRowToMatrix(*offset_ptr);
228 case DW_LNE_set_address:
229 // Takes a single relocatable address as an operand. The size of the
230 // operand is the size appropriate to hold an address on the target
231 // machine. Set the address register to the value given by the
232 // relocatable address. All of the other statement program opcodes
233 // that affect the address register add a delta to it. This instruction
234 // stores a relocatable value into it instead.
235 state.Address = debug_line_data.getAddress(offset_ptr);
238 case DW_LNE_define_file:
239 // Takes 4 arguments. The first is a null terminated string containing
240 // a source file name. The second is an unsigned LEB128 number
241 // representing the directory index of the directory in which the file
242 // was found. The third is an unsigned LEB128 number representing the
243 // time of last modification of the file. The fourth is an unsigned
244 // LEB128 number representing the length in bytes of the file. The time
245 // and length fields may contain LEB128(0) if the information is not
248 // The directory index represents an entry in the include_directories
249 // section of the statement program prologue. The index is LEB128(0)
250 // if the file was found in the current directory of the compilation,
251 // LEB128(1) if it was found in the first directory in the
252 // include_directories section, and so on. The directory index is
253 // ignored for file names that represent full path names.
255 // The files are numbered, starting at 1, in the order in which they
256 // appear; the names in the prologue come before names defined by
257 // the DW_LNE_define_file instruction. These numbers are used in the
258 // the file register of the state machine.
260 FileNameEntry fileEntry;
261 fileEntry.Name = debug_line_data.getCStr(offset_ptr);
262 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
263 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
264 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
265 prologue->FileNames.push_back(fileEntry);
270 // Length doesn't include the zero opcode byte or the length itself, but
271 // it does include the sub_opcode, so we have to adjust for that below
272 (*offset_ptr) += arg_size;
275 } else if (opcode < prologue->OpcodeBase) {
279 // Takes no arguments. Append a row to the matrix using the
280 // current values of the state-machine registers. Then set
281 // the basic_block register to false.
282 state.appendRowToMatrix(*offset_ptr);
285 case DW_LNS_advance_pc:
286 // Takes a single unsigned LEB128 operand, multiplies it by the
287 // min_inst_length field of the prologue, and adds the
288 // result to the address register of the state machine.
289 state.Address += debug_line_data.getULEB128(offset_ptr) *
290 prologue->MinInstLength;
293 case DW_LNS_advance_line:
294 // Takes a single signed LEB128 operand and adds that value to
295 // the line register of the state machine.
296 state.Line += debug_line_data.getSLEB128(offset_ptr);
299 case DW_LNS_set_file:
300 // Takes a single unsigned LEB128 operand and stores it in the file
301 // register of the state machine.
302 state.File = debug_line_data.getULEB128(offset_ptr);
305 case DW_LNS_set_column:
306 // Takes a single unsigned LEB128 operand and stores it in the
307 // column register of the state machine.
308 state.Column = debug_line_data.getULEB128(offset_ptr);
311 case DW_LNS_negate_stmt:
312 // Takes no arguments. Set the is_stmt register of the state
313 // machine to the logical negation of its current value.
314 state.IsStmt = !state.IsStmt;
317 case DW_LNS_set_basic_block:
318 // Takes no arguments. Set the basic_block register of the
319 // state machine to true
320 state.BasicBlock = true;
323 case DW_LNS_const_add_pc:
324 // Takes no arguments. Add to the address register of the state
325 // machine the address increment value corresponding to special
326 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
327 // when the statement program needs to advance the address by a
328 // small amount, it can use a single special opcode, which occupies
329 // a single byte. When it needs to advance the address by up to
330 // twice the range of the last special opcode, it can use
331 // DW_LNS_const_add_pc followed by a special opcode, for a total
332 // of two bytes. Only if it needs to advance the address by more
333 // than twice that range will it need to use both DW_LNS_advance_pc
334 // and a special opcode, requiring three or more bytes.
336 uint8_t adjust_opcode = 255 - prologue->OpcodeBase;
337 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
338 prologue->MinInstLength;
339 state.Address += addr_offset;
343 case DW_LNS_fixed_advance_pc:
344 // Takes a single uhalf operand. Add to the address register of
345 // the state machine the value of the (unencoded) operand. This
346 // is the only extended opcode that takes an argument that is not
347 // a variable length number. The motivation for DW_LNS_fixed_advance_pc
348 // is this: existing assemblers cannot emit DW_LNS_advance_pc or
349 // special opcodes because they cannot encode LEB128 numbers or
350 // judge when the computation of a special opcode overflows and
351 // requires the use of DW_LNS_advance_pc. Such assemblers, however,
352 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
353 state.Address += debug_line_data.getU16(offset_ptr);
356 case DW_LNS_set_prologue_end:
357 // Takes no arguments. Set the prologue_end register of the
358 // state machine to true
359 state.PrologueEnd = true;
362 case DW_LNS_set_epilogue_begin:
363 // Takes no arguments. Set the basic_block register of the
364 // state machine to true
365 state.EpilogueBegin = true;
369 // Takes a single unsigned LEB128 operand and stores it in the
370 // column register of the state machine.
371 state.Isa = debug_line_data.getULEB128(offset_ptr);
375 // Handle any unknown standard opcodes here. We know the lengths
376 // of such opcodes because they are specified in the prologue
377 // as a multiple of LEB128 operands for each opcode.
379 assert(opcode - 1U < prologue->StandardOpcodeLengths.size());
380 uint8_t opcode_length = prologue->StandardOpcodeLengths[opcode - 1];
381 for (uint8_t i=0; i<opcode_length; ++i)
382 debug_line_data.getULEB128(offset_ptr);
389 // A special opcode value is chosen based on the amount that needs
390 // to be added to the line and address registers. The maximum line
391 // increment for a special opcode is the value of the line_base
392 // field in the header, plus the value of the line_range field,
393 // minus 1 (line base + line range - 1). If the desired line
394 // increment is greater than the maximum line increment, a standard
395 // opcode must be used instead of a special opcode. The “address
396 // advance” is calculated by dividing the desired address increment
397 // by the minimum_instruction_length field from the header. The
398 // special opcode is then calculated using the following formula:
400 // opcode = (desired line increment - line_base) +
401 // (line_range * address advance) + opcode_base
403 // If the resulting opcode is greater than 255, a standard opcode
404 // must be used instead.
406 // To decode a special opcode, subtract the opcode_base from the
407 // opcode itself to give the adjusted opcode. The amount to
408 // increment the address register is the result of the adjusted
409 // opcode divided by the line_range multiplied by the
410 // minimum_instruction_length field from the header. That is:
412 // address increment = (adjusted opcode / line_range) *
413 // minimum_instruction_length
415 // The amount to increment the line register is the line_base plus
416 // the result of the adjusted opcode modulo the line_range. That is:
418 // line increment = line_base + (adjusted opcode % line_range)
420 uint8_t adjust_opcode = opcode - prologue->OpcodeBase;
421 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
422 prologue->MinInstLength;
423 int32_t line_offset = prologue->LineBase +
424 (adjust_opcode % prologue->LineRange);
425 state.Line += line_offset;
426 state.Address += addr_offset;
427 state.appendRowToMatrix(*offset_ptr);
431 state.finalize(*offset_ptr);
436 static bool findMatchingAddress(const DWARFDebugLine::Row& row1,
437 const DWARFDebugLine::Row& row2) {
438 return row1.Address < row2.Address;
442 DWARFDebugLine::LineTable::lookupAddress(uint64_t address,
443 uint64_t cu_high_pc) const {
444 uint32_t index = UINT32_MAX;
446 // Use the lower_bound algorithm to perform a binary search since we know
447 // that our line table data is ordered by address.
448 DWARFDebugLine::Row row;
449 row.Address = address;
450 typedef std::vector<Row>::const_iterator iterator;
451 iterator begin_pos = Rows.begin();
452 iterator end_pos = Rows.end();
453 iterator pos = std::lower_bound(begin_pos, end_pos, row,
454 findMatchingAddress);
455 if (pos == end_pos) {
456 if (address < cu_high_pc)
457 return Rows.size()-1;
459 // Rely on fact that we are using a std::vector and we can do
460 // pointer arithmetic to find the row index (which will be one less
461 // that what we found since it will find the first position after
462 // the current address) since std::vector iterators are just
463 // pointers to the container type.
464 index = pos - begin_pos;
465 if (pos->Address > address) {
473 return index; // Failed to find address.