blob: c2fb111ebcf7ef72ec69c3814ed06d3cf03583b7 [file] [log] [blame]
Benjamin Kramerb848e972011-09-15 02:12:05 +00001//===-- 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"
13#include "llvm/Support/raw_ostream.h"
Benjamin Kramer7393c7f2011-09-15 02:19:33 +000014#include <algorithm>
Benjamin Kramerb848e972011-09-15 02:12:05 +000015using namespace llvm;
16using namespace dwarf;
17
18void 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);
28
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]);
32
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";
37
38 if (!FileNames.empty()) {
39 OS << " Dir Mod Time File Len File Name\n"
40 << " ---- ---------- ---------- -----------"
41 "----------------\n";
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';
47 }
48 }
49}
50
51void DWARFDebugLine::Row::postAppend() {
52 BasicBlock = false;
53 PrologueEnd = false;
54 EpilogueBegin = false;
55}
56
57void DWARFDebugLine::Row::reset(bool default_is_stmt) {
58 Address = 0;
59 Line = 1;
60 Column = 0;
61 File = 1;
62 Isa = 0;
63 IsStmt = default_is_stmt;
64 BasicBlock = false;
65 EndSequence = false;
66 PrologueEnd = false;
67 EpilogueBegin = false;
68}
69
70void 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" : "")
78 << '\n';
79}
80
81void DWARFDebugLine::LineTable::dump(raw_ostream &OS) const {
82 Prologue.dump(OS);
83 OS << '\n';
84
85 if (!Rows.empty()) {
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)
90 pos->dump(OS);
91 }
92}
93
Nick Lewycky6bc4e712011-09-15 03:41:51 +000094DWARFDebugLine::State::~State() {}
95
Benjamin Kramerb848e972011-09-15 02:12:05 +000096void DWARFDebugLine::State::appendRowToMatrix(uint32_t offset) {
97 ++row; // Increase the row number.
98 LineTable::appendRow(*this);
99 Row::postAppend();
100}
101
102void DWARFDebugLine::parse(const DataExtractor debug_line_data) {
103 LineTableMap.clear();
104 uint32_t offset = 0;
105 State state;
106 while (debug_line_data.isValidOffset(offset)) {
107 const uint32_t debug_line_offset = offset;
108
109 if (parseStatementTable(debug_line_data, &offset, state)) {
110 // Make sure we don't don't loop infinitely
111 if (offset <= debug_line_offset)
112 break;
113
114 LineTableMap[debug_line_offset] = state;
115 state.reset();
116 }
117 else
118 ++offset; // Try next byte in line table
119 }
120}
121
Nick Lewycky6bc4e712011-09-15 03:41:51 +0000122DWARFDebugLine::DumpingState::~DumpingState() {}
123
Benjamin Kramerb848e972011-09-15 02:12:05 +0000124void DWARFDebugLine::DumpingState::finalize(uint32_t offset) {
125 LineTable::dump(OS);
126}
127
128void DWARFDebugLine::dump(const DataExtractor debug_line_data, raw_ostream &OS){
129 uint32_t offset = 0;
130 DumpingState state(OS);
131 while (debug_line_data.isValidOffset(offset)) {
132 const uint32_t debug_line_offset = offset;
133
134 if (parseStatementTable(debug_line_data, &offset, state)) {
135 // Make sure we don't don't loop infinitely
136 if (offset <= debug_line_offset)
137 break;
138
139 state.reset();
140 }
141 else
142 ++offset; // Try next byte in line table
143 }
144}
145
146const DWARFDebugLine::LineTable *
147DWARFDebugLine::getLineTable(uint32_t offset) const {
148 LineTableConstIter pos = LineTableMap.find(offset);
149 if (pos != LineTableMap.end())
150 return &pos->second;
151 return 0;
152}
153
154bool
155DWARFDebugLine::parsePrologue(DataExtractor debug_line_data,
156 uint32_t *offset_ptr, Prologue *prologue) {
157 const uint32_t prologue_offset = *offset_ptr;
158
159 prologue->clear();
160 prologue->TotalLength = debug_line_data.getU32(offset_ptr);
161 prologue->Version = debug_line_data.getU16(offset_ptr);
162 if (prologue->Version != 2)
163 return false;
164
165 prologue->PrologueLength = debug_line_data.getU32(offset_ptr);
166 const uint32_t end_prologue_offset = prologue->PrologueLength + *offset_ptr;
167 prologue->MinInstLength = debug_line_data.getU8(offset_ptr);
168 prologue->DefaultIsStmt = debug_line_data.getU8(offset_ptr);
169 prologue->LineBase = debug_line_data.getU8(offset_ptr);
170 prologue->LineRange = debug_line_data.getU8(offset_ptr);
171 prologue->OpcodeBase = debug_line_data.getU8(offset_ptr);
172
173 prologue->StandardOpcodeLengths.reserve(prologue->OpcodeBase-1);
174 for (uint32_t i = 1; i < prologue->OpcodeBase; ++i) {
175 uint8_t op_len = debug_line_data.getU8(offset_ptr);
176 prologue->StandardOpcodeLengths.push_back(op_len);
177 }
178
179 while (*offset_ptr < end_prologue_offset) {
180 const char *s = debug_line_data.getCStr(offset_ptr);
181 if (s && s[0])
182 prologue->IncludeDirectories.push_back(s);
183 else
184 break;
185 }
186
187 while (*offset_ptr < end_prologue_offset) {
188 const char *name = debug_line_data.getCStr(offset_ptr);
189 if (name && name[0]) {
190 FileNameEntry fileEntry;
191 fileEntry.Name = name;
192 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
193 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
194 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
195 prologue->FileNames.push_back(fileEntry);
196 } else {
197 break;
198 }
199 }
200
201 if (*offset_ptr != end_prologue_offset) {
202 fprintf(stderr, "warning: parsing line table prologue at 0x%8.8x should"
203 " have ended at 0x%8.8x but it ended ad 0x%8.8x\n",
204 prologue_offset, end_prologue_offset, *offset_ptr);
205 }
206 return end_prologue_offset;
207}
208
209bool
210DWARFDebugLine::parseStatementTable(DataExtractor debug_line_data,
211 uint32_t *offset_ptr, State &state) {
212 const uint32_t debug_line_offset = *offset_ptr;
213
214 Prologue *prologue = &state.Prologue;
215
216 if (!parsePrologue(debug_line_data, offset_ptr, prologue)) {
217 // Restore our offset and return false to indicate failure!
218 *offset_ptr = debug_line_offset;
219 return false;
220 }
221
222 const uint32_t end_offset = debug_line_offset + prologue->TotalLength +
223 sizeof(prologue->TotalLength);
224
225 while (*offset_ptr < end_offset) {
226 uint8_t opcode = debug_line_data.getU8(offset_ptr);
227
228 if (opcode == 0) {
229 // Extended Opcodes always start with a zero opcode followed by
230 // a uleb128 length so you can skip ones you don't know about
231 uint32_t ext_offset = *offset_ptr;
232 uint64_t len = debug_line_data.getULEB128(offset_ptr);
233 uint32_t arg_size = len - (*offset_ptr - ext_offset);
234
235 uint8_t sub_opcode = debug_line_data.getU8(offset_ptr);
236 switch (sub_opcode) {
237 case DW_LNE_end_sequence:
238 // Set the end_sequence register of the state machine to true and
239 // append a row to the matrix using the current values of the
240 // state-machine registers. Then reset the registers to the initial
241 // values specified above. Every statement program sequence must end
242 // with a DW_LNE_end_sequence instruction which creates a row whose
243 // address is that of the byte after the last target machine instruction
244 // of the sequence.
245 state.EndSequence = true;
246 state.appendRowToMatrix(*offset_ptr);
247 state.reset();
248 break;
249
250 case DW_LNE_set_address:
251 // Takes a single relocatable address as an operand. The size of the
252 // operand is the size appropriate to hold an address on the target
253 // machine. Set the address register to the value given by the
254 // relocatable address. All of the other statement program opcodes
255 // that affect the address register add a delta to it. This instruction
256 // stores a relocatable value into it instead.
257 state.Address = debug_line_data.getAddress(offset_ptr);
258 break;
259
260 case DW_LNE_define_file:
261 // Takes 4 arguments. The first is a null terminated string containing
262 // a source file name. The second is an unsigned LEB128 number
263 // representing the directory index of the directory in which the file
264 // was found. The third is an unsigned LEB128 number representing the
265 // time of last modification of the file. The fourth is an unsigned
266 // LEB128 number representing the length in bytes of the file. The time
267 // and length fields may contain LEB128(0) if the information is not
268 // available.
269 //
270 // The directory index represents an entry in the include_directories
271 // section of the statement program prologue. The index is LEB128(0)
272 // if the file was found in the current directory of the compilation,
273 // LEB128(1) if it was found in the first directory in the
274 // include_directories section, and so on. The directory index is
275 // ignored for file names that represent full path names.
276 //
277 // The files are numbered, starting at 1, in the order in which they
278 // appear; the names in the prologue come before names defined by
279 // the DW_LNE_define_file instruction. These numbers are used in the
280 // the file register of the state machine.
281 {
282 FileNameEntry fileEntry;
283 fileEntry.Name = debug_line_data.getCStr(offset_ptr);
284 fileEntry.DirIdx = debug_line_data.getULEB128(offset_ptr);
285 fileEntry.ModTime = debug_line_data.getULEB128(offset_ptr);
286 fileEntry.Length = debug_line_data.getULEB128(offset_ptr);
287 prologue->FileNames.push_back(fileEntry);
288 }
289 break;
290
291 default:
292 // Length doesn't include the zero opcode byte or the length itself, but
293 // it does include the sub_opcode, so we have to adjust for that below
294 (*offset_ptr) += arg_size;
295 break;
296 }
297 } else if (opcode < prologue->OpcodeBase) {
298 switch (opcode) {
299 // Standard Opcodes
300 case DW_LNS_copy:
301 // Takes no arguments. Append a row to the matrix using the
302 // current values of the state-machine registers. Then set
303 // the basic_block register to false.
304 state.appendRowToMatrix(*offset_ptr);
305 break;
306
307 case DW_LNS_advance_pc:
308 // Takes a single unsigned LEB128 operand, multiplies it by the
309 // min_inst_length field of the prologue, and adds the
310 // result to the address register of the state machine.
311 state.Address += debug_line_data.getULEB128(offset_ptr) *
312 prologue->MinInstLength;
313 break;
314
315 case DW_LNS_advance_line:
316 // Takes a single signed LEB128 operand and adds that value to
317 // the line register of the state machine.
318 state.Line += debug_line_data.getSLEB128(offset_ptr);
319 break;
320
321 case DW_LNS_set_file:
322 // Takes a single unsigned LEB128 operand and stores it in the file
323 // register of the state machine.
324 state.File = debug_line_data.getULEB128(offset_ptr);
325 break;
326
327 case DW_LNS_set_column:
328 // Takes a single unsigned LEB128 operand and stores it in the
329 // column register of the state machine.
330 state.Column = debug_line_data.getULEB128(offset_ptr);
331 break;
332
333 case DW_LNS_negate_stmt:
334 // Takes no arguments. Set the is_stmt register of the state
335 // machine to the logical negation of its current value.
336 state.IsStmt = !state.IsStmt;
337 break;
338
339 case DW_LNS_set_basic_block:
340 // Takes no arguments. Set the basic_block register of the
341 // state machine to true
342 state.BasicBlock = true;
343 break;
344
345 case DW_LNS_const_add_pc:
346 // Takes no arguments. Add to the address register of the state
347 // machine the address increment value corresponding to special
348 // opcode 255. The motivation for DW_LNS_const_add_pc is this:
349 // when the statement program needs to advance the address by a
350 // small amount, it can use a single special opcode, which occupies
351 // a single byte. When it needs to advance the address by up to
352 // twice the range of the last special opcode, it can use
353 // DW_LNS_const_add_pc followed by a special opcode, for a total
354 // of two bytes. Only if it needs to advance the address by more
355 // than twice that range will it need to use both DW_LNS_advance_pc
356 // and a special opcode, requiring three or more bytes.
357 {
358 uint8_t adjust_opcode = 255 - prologue->OpcodeBase;
359 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
360 prologue->MinInstLength;
361 state.Address += addr_offset;
362 }
363 break;
364
365 case DW_LNS_fixed_advance_pc:
366 // Takes a single uhalf operand. Add to the address register of
367 // the state machine the value of the (unencoded) operand. This
368 // is the only extended opcode that takes an argument that is not
369 // a variable length number. The motivation for DW_LNS_fixed_advance_pc
370 // is this: existing assemblers cannot emit DW_LNS_advance_pc or
371 // special opcodes because they cannot encode LEB128 numbers or
372 // judge when the computation of a special opcode overflows and
373 // requires the use of DW_LNS_advance_pc. Such assemblers, however,
374 // can use DW_LNS_fixed_advance_pc instead, sacrificing compression.
375 state.Address += debug_line_data.getU16(offset_ptr);
376 break;
377
378 case DW_LNS_set_prologue_end:
379 // Takes no arguments. Set the prologue_end register of the
380 // state machine to true
381 state.PrologueEnd = true;
382 break;
383
384 case DW_LNS_set_epilogue_begin:
385 // Takes no arguments. Set the basic_block register of the
386 // state machine to true
387 state.EpilogueBegin = true;
388 break;
389
390 case DW_LNS_set_isa:
391 // Takes a single unsigned LEB128 operand and stores it in the
392 // column register of the state machine.
393 state.Isa = debug_line_data.getULEB128(offset_ptr);
394 break;
395
396 default:
397 // Handle any unknown standard opcodes here. We know the lengths
398 // of such opcodes because they are specified in the prologue
399 // as a multiple of LEB128 operands for each opcode.
400 {
Benjamin Kramer068d9a52011-09-15 03:20:04 +0000401 assert(opcode - 1U < prologue->StandardOpcodeLengths.size());
Benjamin Kramerb848e972011-09-15 02:12:05 +0000402 uint8_t opcode_length = prologue->StandardOpcodeLengths[opcode - 1];
403 for (uint8_t i=0; i<opcode_length; ++i)
404 debug_line_data.getULEB128(offset_ptr);
405 }
406 break;
407 }
408 } else {
409 // Special Opcodes
410
411 // A special opcode value is chosen based on the amount that needs
412 // to be added to the line and address registers. The maximum line
413 // increment for a special opcode is the value of the line_base
414 // field in the header, plus the value of the line_range field,
415 // minus 1 (line base + line range - 1). If the desired line
416 // increment is greater than the maximum line increment, a standard
417 // opcode must be used instead of a special opcode. The “address
418 // advance” is calculated by dividing the desired address increment
419 // by the minimum_instruction_length field from the header. The
420 // special opcode is then calculated using the following formula:
421 //
422 // opcode = (desired line increment - line_base) +
423 // (line_range * address advance) + opcode_base
424 //
425 // If the resulting opcode is greater than 255, a standard opcode
426 // must be used instead.
427 //
428 // To decode a special opcode, subtract the opcode_base from the
429 // opcode itself to give the adjusted opcode. The amount to
430 // increment the address register is the result of the adjusted
431 // opcode divided by the line_range multiplied by the
432 // minimum_instruction_length field from the header. That is:
433 //
434 // address increment = (adjusted opcode / line_range) *
435 // minimum_instruction_length
436 //
437 // The amount to increment the line register is the line_base plus
438 // the result of the adjusted opcode modulo the line_range. That is:
439 //
440 // line increment = line_base + (adjusted opcode % line_range)
441
442 uint8_t adjust_opcode = opcode - prologue->OpcodeBase;
443 uint64_t addr_offset = (adjust_opcode / prologue->LineRange) *
444 prologue->MinInstLength;
445 int32_t line_offset = prologue->LineBase +
446 (adjust_opcode % prologue->LineRange);
447 state.Line += line_offset;
448 state.Address += addr_offset;
449 state.appendRowToMatrix(*offset_ptr);
450 }
451 }
452
453 state.finalize(*offset_ptr);
454
455 return end_offset;
456}
457
458static bool findMatchingAddress(const DWARFDebugLine::Row& row1,
459 const DWARFDebugLine::Row& row2) {
460 return row1.Address < row2.Address;
461}
462
463uint32_t
464DWARFDebugLine::LineTable::lookupAddress(uint64_t address,
465 uint64_t cu_high_pc) const {
466 uint32_t index = UINT32_MAX;
467 if (!Rows.empty()) {
468 // Use the lower_bound algorithm to perform a binary search since we know
469 // that our line table data is ordered by address.
470 DWARFDebugLine::Row row;
471 row.Address = address;
472 typedef std::vector<Row>::const_iterator iterator;
473 iterator begin_pos = Rows.begin();
474 iterator end_pos = Rows.end();
475 iterator pos = std::lower_bound(begin_pos, end_pos, row,
476 findMatchingAddress);
477 if (pos == end_pos) {
478 if (address < cu_high_pc)
479 return Rows.size()-1;
480 } else {
481 // Rely on fact that we are using a std::vector and we can do
482 // pointer arithmetic to find the row index (which will be one less
483 // that what we found since it will find the first position after
484 // the current address) since std::vector iterators are just
485 // pointers to the container type.
486 index = pos - begin_pos;
487 if (pos->Address > address) {
488 if (index > 0)
489 --index;
490 else
491 index = UINT32_MAX;
492 }
493 }
494 }
495 return index; // Failed to find address.
496}