Commit | Line | Data |
---|---|---|
5c2c6c95 ILT |
1 | // dwarf_reader.h -- parse dwarf2/3 debug information for gold -*- C++ -*- |
2 | ||
c1027032 | 3 | // Copyright 2007, 2008, 2009, 2010, 2011, 2012 Free Software Foundation, Inc. |
5c2c6c95 ILT |
4 | // Written by Ian Lance Taylor <iant@google.com>. |
5 | ||
6 | // This file is part of gold. | |
7 | ||
8 | // This program is free software; you can redistribute it and/or modify | |
9 | // it under the terms of the GNU General Public License as published by | |
10 | // the Free Software Foundation; either version 3 of the License, or | |
11 | // (at your option) any later version. | |
12 | ||
13 | // This program is distributed in the hope that it will be useful, | |
14 | // but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 | // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
16 | // GNU General Public License for more details. | |
17 | ||
18 | // You should have received a copy of the GNU General Public License | |
19 | // along with this program; if not, write to the Free Software | |
20 | // Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, | |
21 | // MA 02110-1301, USA. | |
22 | ||
23 | #ifndef GOLD_DWARF_READER_H | |
24 | #define GOLD_DWARF_READER_H | |
25 | ||
26 | #include <vector> | |
4c50553d | 27 | #include <map> |
71ff8986 | 28 | #include <limits.h> |
c1027032 | 29 | #include <sys/types.h> |
5c2c6c95 | 30 | |
4c50553d | 31 | #include "elfcpp.h" |
5c2c6c95 ILT |
32 | #include "elfcpp_swap.h" |
33 | #include "dwarf.h" | |
24badc65 | 34 | #include "reloc.h" |
5c2c6c95 ILT |
35 | |
36 | namespace gold | |
37 | { | |
38 | ||
c1027032 | 39 | class Dwarf_info_reader; |
5c2c6c95 ILT |
40 | struct LineStateMachine; |
41 | ||
c1027032 CC |
42 | // This class is used to extract the section index and offset of |
43 | // the target of a relocation for a given offset within the section. | |
44 | ||
45 | class Elf_reloc_mapper | |
46 | { | |
47 | public: | |
48 | Elf_reloc_mapper() | |
49 | { } | |
50 | ||
51 | virtual | |
52 | ~Elf_reloc_mapper() | |
53 | { } | |
54 | ||
55 | // Initialize the relocation tracker for section RELOC_SHNDX. | |
56 | bool | |
57 | initialize(unsigned int reloc_shndx, unsigned int reloc_type) | |
58 | { return this->do_initialize(reloc_shndx, reloc_type); } | |
59 | ||
60 | // Return the next reloc_offset. | |
61 | off_t | |
62 | next_offset() | |
63 | { return this->do_next_offset(); } | |
64 | ||
65 | // Advance to the next relocation past OFFSET. | |
66 | void | |
67 | advance(off_t offset) | |
68 | { this->do_advance(offset); } | |
69 | ||
70 | // Return the section index and offset within the section of the target | |
71 | // of the relocation for RELOC_OFFSET in the referring section. | |
72 | unsigned int | |
73 | get_reloc_target(off_t reloc_offset, off_t* target_offset) | |
74 | { return this->do_get_reloc_target(reloc_offset, target_offset); } | |
75 | ||
76 | // Checkpoint the current position in the reloc section. | |
77 | uint64_t | |
78 | checkpoint() const | |
79 | { return this->do_checkpoint(); } | |
80 | ||
81 | // Reset the current position to the CHECKPOINT. | |
82 | void | |
83 | reset(uint64_t checkpoint) | |
84 | { this->do_reset(checkpoint); } | |
85 | ||
86 | protected: | |
87 | virtual bool | |
88 | do_initialize(unsigned int, unsigned int) = 0; | |
89 | ||
90 | // Return the next reloc_offset. | |
91 | virtual off_t | |
92 | do_next_offset() = 0; | |
93 | ||
94 | // Advance to the next relocation past OFFSET. | |
95 | virtual void | |
96 | do_advance(off_t offset) = 0; | |
97 | ||
98 | virtual unsigned int | |
99 | do_get_reloc_target(off_t reloc_offset, off_t* target_offset) = 0; | |
100 | ||
101 | // Checkpoint the current position in the reloc section. | |
102 | virtual uint64_t | |
103 | do_checkpoint() const = 0; | |
104 | ||
105 | // Reset the current position to the CHECKPOINT. | |
106 | virtual void | |
107 | do_reset(uint64_t checkpoint) = 0; | |
108 | }; | |
109 | ||
110 | template<int size, bool big_endian> | |
111 | class Sized_elf_reloc_mapper : public Elf_reloc_mapper | |
112 | { | |
113 | public: | |
114 | Sized_elf_reloc_mapper(Object* object, const unsigned char* symtab, | |
115 | off_t symtab_size) | |
116 | : object_(object), symtab_(symtab), symtab_size_(symtab_size), | |
117 | reloc_type_(0), track_relocs_() | |
118 | { } | |
119 | ||
120 | protected: | |
121 | bool | |
122 | do_initialize(unsigned int reloc_shndx, unsigned int reloc_type); | |
123 | ||
124 | // Return the next reloc_offset. | |
125 | virtual off_t | |
126 | do_next_offset() | |
127 | { return this->track_relocs_.next_offset(); } | |
128 | ||
129 | // Advance to the next relocation past OFFSET. | |
130 | virtual void | |
131 | do_advance(off_t offset) | |
132 | { this->track_relocs_.advance(offset); } | |
133 | ||
134 | unsigned int | |
135 | do_get_reloc_target(off_t reloc_offset, off_t* target_offset); | |
136 | ||
137 | // Checkpoint the current position in the reloc section. | |
138 | uint64_t | |
139 | do_checkpoint() const | |
140 | { return this->track_relocs_.checkpoint(); } | |
141 | ||
142 | // Reset the current position to the CHECKPOINT. | |
143 | void | |
144 | do_reset(uint64_t checkpoint) | |
145 | { this->track_relocs_.reset(checkpoint); } | |
146 | ||
147 | private: | |
148 | typedef typename elfcpp::Elf_types<size>::Elf_Addr Address; | |
149 | ||
150 | // Return the section index of symbol SYMNDX, and copy its value to *VALUE. | |
151 | // Set *IS_ORDINARY true if the section index is an ordinary section index. | |
152 | unsigned int | |
153 | symbol_section(unsigned int symndx, Address* value, bool* is_ordinary); | |
154 | ||
155 | // The object file. | |
156 | Object* object_; | |
157 | // The ELF symbol table. | |
158 | const unsigned char* symtab_; | |
159 | // The size of the ELF symbol table. | |
160 | off_t symtab_size_; | |
161 | // Type of the relocation section (SHT_REL or SHT_RELA). | |
162 | unsigned int reloc_type_; | |
163 | // Relocations for the referring section. | |
164 | Track_relocs<size, big_endian> track_relocs_; | |
165 | }; | |
166 | ||
167 | // This class is used to read the abbreviations table from the | |
168 | // .debug_abbrev section of the object file. | |
169 | ||
170 | class Dwarf_abbrev_table | |
171 | { | |
172 | public: | |
173 | // An attribute list entry. | |
174 | struct Attribute | |
175 | { | |
176 | Attribute(unsigned int a, unsigned int f) | |
177 | : attr(a), form(f) | |
178 | { } | |
179 | unsigned int attr; | |
180 | unsigned int form; | |
181 | }; | |
182 | ||
183 | // An abbrev code entry. | |
184 | struct Abbrev_code | |
185 | { | |
186 | Abbrev_code(unsigned int t, bool hc) | |
187 | : tag(t), has_children(hc), has_sibling_attribute(false), attributes() | |
188 | { | |
189 | this->attributes.reserve(10); | |
190 | } | |
191 | ||
192 | void | |
193 | add_attribute(unsigned int attr, unsigned int form) | |
194 | { | |
195 | this->attributes.push_back(Attribute(attr, form)); | |
196 | } | |
197 | ||
198 | // The DWARF tag. | |
199 | unsigned int tag; | |
200 | // True if the DIE has children. | |
201 | bool has_children : 1; | |
202 | // True if the DIE has a sibling attribute. | |
203 | bool has_sibling_attribute : 1; | |
204 | // The list of attributes and forms. | |
205 | std::vector<Attribute> attributes; | |
206 | }; | |
207 | ||
208 | Dwarf_abbrev_table() | |
209 | : abbrev_shndx_(0), abbrev_offset_(0), buffer_(NULL), buffer_end_(NULL), | |
210 | owns_buffer_(false), buffer_pos_(NULL), high_abbrev_codes_() | |
211 | { | |
212 | memset(this->low_abbrev_codes_, 0, sizeof(this->low_abbrev_codes_)); | |
213 | } | |
214 | ||
215 | ~Dwarf_abbrev_table() | |
216 | { | |
217 | if (this->owns_buffer_ && this->buffer_ != NULL) | |
218 | delete[] this->buffer_; | |
219 | this->clear_abbrev_codes(); | |
220 | } | |
221 | ||
222 | // Read the abbrev table from an object file. | |
223 | bool | |
224 | read_abbrevs(Relobj* object, | |
225 | unsigned int abbrev_shndx, | |
226 | off_t abbrev_offset) | |
227 | { | |
228 | // If we've already read this abbrev table, return immediately. | |
229 | if (this->abbrev_shndx_ > 0 | |
230 | && this->abbrev_shndx_ == abbrev_shndx | |
231 | && this->abbrev_offset_ == abbrev_offset) | |
232 | return true; | |
233 | return this->do_read_abbrevs(object, abbrev_shndx, abbrev_offset); | |
234 | } | |
235 | ||
236 | // Return the abbrev code entry for CODE. This is a fast path for | |
237 | // abbrev codes that are in the direct lookup table. If not found | |
238 | // there, we call do_get_abbrev() to do the hard work. | |
239 | const Abbrev_code* | |
240 | get_abbrev(unsigned int code) | |
241 | { | |
242 | if (code < this->low_abbrev_code_max_ | |
243 | && this->low_abbrev_codes_[code] != NULL) | |
244 | return this->low_abbrev_codes_[code]; | |
245 | return this->do_get_abbrev(code); | |
246 | } | |
247 | ||
248 | private: | |
249 | // Read the abbrev table from an object file. | |
250 | bool | |
251 | do_read_abbrevs(Relobj* object, | |
252 | unsigned int abbrev_shndx, | |
253 | off_t abbrev_offset); | |
254 | ||
255 | // Lookup the abbrev code entry for CODE. | |
256 | const Abbrev_code* | |
257 | do_get_abbrev(unsigned int code); | |
258 | ||
259 | // Store an abbrev code entry for CODE. | |
260 | void | |
261 | store_abbrev(unsigned int code, const Abbrev_code* entry) | |
262 | { | |
263 | if (code < this->low_abbrev_code_max_) | |
264 | this->low_abbrev_codes_[code] = entry; | |
265 | else | |
266 | this->high_abbrev_codes_[code] = entry; | |
267 | } | |
268 | ||
269 | // Clear the abbrev code table and release the memory it uses. | |
270 | void | |
271 | clear_abbrev_codes(); | |
272 | ||
273 | typedef Unordered_map<unsigned int, const Abbrev_code*> Abbrev_code_table; | |
274 | ||
275 | // The section index of the current abbrev table. | |
276 | unsigned int abbrev_shndx_; | |
277 | // The offset within the section of the current abbrev table. | |
278 | off_t abbrev_offset_; | |
279 | // The buffer containing the .debug_abbrev section. | |
280 | const unsigned char* buffer_; | |
281 | const unsigned char* buffer_end_; | |
282 | // True if this object owns the buffer and needs to delete it. | |
283 | bool owns_buffer_; | |
284 | // Pointer to the current position in the buffer. | |
285 | const unsigned char* buffer_pos_; | |
286 | // The table of abbrev codes. | |
287 | // We use a direct-lookup array for low abbrev codes, | |
288 | // and store the rest in a hash table. | |
289 | static const unsigned int low_abbrev_code_max_ = 256; | |
290 | const Abbrev_code* low_abbrev_codes_[low_abbrev_code_max_]; | |
291 | Abbrev_code_table high_abbrev_codes_; | |
292 | }; | |
293 | ||
294 | // A DWARF range list. The start and end offsets are relative | |
295 | // to the input section SHNDX. Each range must lie entirely | |
296 | // within a single section. | |
297 | ||
298 | class Dwarf_range_list | |
299 | { | |
300 | public: | |
301 | struct Range | |
302 | { | |
303 | Range(unsigned int a_shndx, off_t a_start, off_t a_end) | |
304 | : shndx(a_shndx), start(a_start), end(a_end) | |
305 | { } | |
306 | ||
307 | unsigned int shndx; | |
308 | off_t start; | |
309 | off_t end; | |
310 | }; | |
311 | ||
312 | Dwarf_range_list() | |
313 | : range_list_() | |
314 | { } | |
315 | ||
316 | void | |
317 | add(unsigned int shndx, off_t start, off_t end) | |
318 | { this->range_list_.push_back(Range(shndx, start, end)); } | |
319 | ||
320 | size_t | |
321 | size() const | |
322 | { return this->range_list_.size(); } | |
323 | ||
324 | const Range& | |
325 | operator[](off_t i) const | |
326 | { return this->range_list_[i]; } | |
327 | ||
328 | private: | |
329 | std::vector<Range> range_list_; | |
330 | }; | |
331 | ||
332 | // This class is used to read the ranges table from the | |
333 | // .debug_ranges section of the object file. | |
334 | ||
335 | class Dwarf_ranges_table | |
336 | { | |
337 | public: | |
ed5d6712 CC |
338 | Dwarf_ranges_table(Dwarf_info_reader* dwinfo) |
339 | : dwinfo_(dwinfo), ranges_shndx_(0), ranges_buffer_(NULL), | |
340 | ranges_buffer_end_(NULL), owns_ranges_buffer_(false), | |
341 | ranges_reloc_mapper_(NULL), output_section_offset_(0) | |
c1027032 CC |
342 | { } |
343 | ||
344 | ~Dwarf_ranges_table() | |
345 | { | |
346 | if (this->owns_ranges_buffer_ && this->ranges_buffer_ != NULL) | |
347 | delete[] this->ranges_buffer_; | |
348 | if (this->ranges_reloc_mapper_ != NULL) | |
349 | delete this->ranges_reloc_mapper_; | |
350 | } | |
351 | ||
352 | // Read the ranges table from an object file. | |
353 | bool | |
354 | read_ranges_table(Relobj* object, | |
355 | const unsigned char* symtab, | |
356 | off_t symtab_size, | |
357 | unsigned int ranges_shndx); | |
358 | ||
359 | // Read the range table from an object file. | |
360 | Dwarf_range_list* | |
361 | read_range_list(Relobj* object, | |
362 | const unsigned char* symtab, | |
363 | off_t symtab_size, | |
364 | unsigned int address_size, | |
365 | unsigned int ranges_shndx, | |
366 | off_t ranges_offset); | |
367 | ||
368 | private: | |
ed5d6712 CC |
369 | // The Dwarf_info_reader, for reading data. |
370 | Dwarf_info_reader* dwinfo_; | |
c1027032 CC |
371 | // The section index of the ranges table. |
372 | unsigned int ranges_shndx_; | |
373 | // The buffer containing the .debug_ranges section. | |
374 | const unsigned char* ranges_buffer_; | |
375 | const unsigned char* ranges_buffer_end_; | |
376 | // True if this object owns the buffer and needs to delete it. | |
377 | bool owns_ranges_buffer_; | |
378 | // Relocation mapper for the .debug_ranges section. | |
379 | Elf_reloc_mapper* ranges_reloc_mapper_; | |
380 | // For incremental update links, this will hold the offset of the | |
381 | // input section within the output section. Offsets read from | |
382 | // relocated data will be relative to the output section, and need | |
383 | // to be corrected before reading data from the input section. | |
384 | uint64_t output_section_offset_; | |
385 | }; | |
386 | ||
387 | // This class is used to read the pubnames and pubtypes tables from the | |
388 | // .debug_pubnames and .debug_pubtypes sections of the object file. | |
389 | ||
390 | class Dwarf_pubnames_table | |
391 | { | |
392 | public: | |
ed5d6712 CC |
393 | Dwarf_pubnames_table(Dwarf_info_reader* dwinfo, bool is_pubtypes) |
394 | : dwinfo_(dwinfo), buffer_(NULL), buffer_end_(NULL), owns_buffer_(false), | |
c1027032 CC |
395 | offset_size_(0), pinfo_(NULL), is_pubtypes_(is_pubtypes), |
396 | output_section_offset_(0) | |
397 | { } | |
398 | ||
399 | ~Dwarf_pubnames_table() | |
400 | { | |
401 | if (this->owns_buffer_ && this->buffer_ != NULL) | |
402 | delete[] this->buffer_; | |
403 | } | |
404 | ||
405 | // Read the pubnames section SHNDX from the object file. | |
406 | bool | |
407 | read_section(Relobj* object, unsigned int shndx); | |
408 | ||
409 | // Read the header for the set at OFFSET. | |
410 | bool | |
411 | read_header(off_t offset); | |
412 | ||
413 | // Read the next name from the set. | |
414 | const char* | |
415 | next_name(); | |
416 | ||
417 | private: | |
ed5d6712 CC |
418 | // The Dwarf_info_reader, for reading data. |
419 | Dwarf_info_reader* dwinfo_; | |
c1027032 CC |
420 | // The buffer containing the .debug_ranges section. |
421 | const unsigned char* buffer_; | |
422 | const unsigned char* buffer_end_; | |
423 | // True if this object owns the buffer and needs to delete it. | |
424 | bool owns_buffer_; | |
425 | // The size of a DWARF offset for the current set. | |
426 | unsigned int offset_size_; | |
427 | // The current position within the buffer. | |
428 | const unsigned char* pinfo_; | |
429 | // TRUE if this is a .debug_pubtypes section. | |
430 | bool is_pubtypes_; | |
431 | // For incremental update links, this will hold the offset of the | |
432 | // input section within the output section. Offsets read from | |
433 | // relocated data will be relative to the output section, and need | |
434 | // to be corrected before reading data from the input section. | |
435 | uint64_t output_section_offset_; | |
436 | }; | |
437 | ||
438 | // This class represents a DWARF Debug Info Entry (DIE). | |
439 | ||
440 | class Dwarf_die | |
441 | { | |
442 | public: | |
443 | // An attribute value. | |
444 | struct Attribute_value | |
445 | { | |
446 | unsigned int attr; | |
447 | unsigned int form; | |
448 | union | |
449 | { | |
450 | int64_t intval; | |
451 | uint64_t uintval; | |
452 | const char* stringval; | |
453 | const unsigned char* blockval; | |
454 | off_t refval; | |
455 | } val; | |
456 | union | |
457 | { | |
458 | // Section index for reference forms. | |
459 | unsigned int shndx; | |
460 | // Block length for block forms. | |
461 | unsigned int blocklen; | |
462 | // Attribute offset for DW_FORM_strp. | |
463 | unsigned int attr_off; | |
464 | } aux; | |
465 | }; | |
466 | ||
467 | // A list of attribute values. | |
468 | typedef std::vector<Attribute_value> Attributes; | |
469 | ||
470 | Dwarf_die(Dwarf_info_reader* dwinfo, | |
471 | off_t die_offset, | |
472 | Dwarf_die* parent); | |
473 | ||
474 | // Return the DWARF tag for this DIE. | |
475 | unsigned int | |
476 | tag() const | |
477 | { | |
478 | if (this->abbrev_code_ == NULL) | |
479 | return 0; | |
480 | return this->abbrev_code_->tag; | |
481 | } | |
482 | ||
483 | // Return true if this DIE has children. | |
484 | bool | |
485 | has_children() const | |
486 | { | |
487 | gold_assert(this->abbrev_code_ != NULL); | |
488 | return this->abbrev_code_->has_children; | |
489 | } | |
490 | ||
491 | // Return true if this DIE has a sibling attribute. | |
492 | bool | |
493 | has_sibling_attribute() const | |
494 | { | |
495 | gold_assert(this->abbrev_code_ != NULL); | |
496 | return this->abbrev_code_->has_sibling_attribute; | |
497 | } | |
498 | ||
499 | // Return the value of attribute ATTR. | |
500 | const Attribute_value* | |
501 | attribute(unsigned int attr); | |
502 | ||
503 | // Return the value of the DW_AT_name attribute. | |
504 | const char* | |
505 | name() | |
506 | { | |
507 | if (this->name_ == NULL) | |
508 | this->set_name(); | |
509 | return this->name_; | |
510 | } | |
511 | ||
512 | // Return the value of the DW_AT_linkage_name | |
513 | // or DW_AT_MIPS_linkage_name attribute. | |
514 | const char* | |
515 | linkage_name() | |
516 | { | |
517 | if (this->linkage_name_ == NULL) | |
518 | this->set_linkage_name(); | |
519 | return this->linkage_name_; | |
520 | } | |
521 | ||
522 | // Return the value of the DW_AT_specification attribute. | |
523 | off_t | |
524 | specification() | |
525 | { | |
526 | if (!this->attributes_read_) | |
527 | this->read_attributes(); | |
528 | return this->specification_; | |
529 | } | |
530 | ||
531 | // Return the value of the DW_AT_abstract_origin attribute. | |
532 | off_t | |
533 | abstract_origin() | |
534 | { | |
535 | if (!this->attributes_read_) | |
536 | this->read_attributes(); | |
537 | return this->abstract_origin_; | |
538 | } | |
539 | ||
540 | // Return the value of attribute ATTR as a string. | |
541 | const char* | |
542 | string_attribute(unsigned int attr); | |
543 | ||
544 | // Return the value of attribute ATTR as an integer. | |
545 | int64_t | |
546 | int_attribute(unsigned int attr); | |
547 | ||
548 | // Return the value of attribute ATTR as an unsigned integer. | |
549 | uint64_t | |
550 | uint_attribute(unsigned int attr); | |
551 | ||
552 | // Return the value of attribute ATTR as a reference. | |
553 | off_t | |
57923f48 MW |
554 | ref_attribute(unsigned int attr, unsigned int* shndx); |
555 | ||
556 | // Return the value of attribute ATTR as a address. | |
557 | off_t | |
558 | address_attribute(unsigned int attr, unsigned int* shndx); | |
c1027032 CC |
559 | |
560 | // Return the value of attribute ATTR as a flag. | |
561 | bool | |
562 | flag_attribute(unsigned int attr) | |
563 | { return this->int_attribute(attr) != 0; } | |
564 | ||
565 | // Return true if this DIE is a declaration. | |
566 | bool | |
567 | is_declaration() | |
568 | { return this->flag_attribute(elfcpp::DW_AT_declaration); } | |
569 | ||
570 | // Return the parent of this DIE. | |
571 | Dwarf_die* | |
572 | parent() const | |
573 | { return this->parent_; } | |
574 | ||
575 | // Return the offset of this DIE. | |
576 | off_t | |
577 | offset() const | |
578 | { return this->die_offset_; } | |
579 | ||
580 | // Return the offset of this DIE's first child. | |
581 | off_t | |
582 | child_offset(); | |
583 | ||
584 | // Set the offset of this DIE's next sibling. | |
585 | void | |
586 | set_sibling_offset(off_t sibling_offset) | |
587 | { this->sibling_offset_ = sibling_offset; } | |
588 | ||
589 | // Return the offset of this DIE's next sibling. | |
590 | off_t | |
591 | sibling_offset(); | |
592 | ||
593 | private: | |
594 | typedef Dwarf_abbrev_table::Abbrev_code Abbrev_code; | |
595 | ||
596 | // Read all the attributes of the DIE. | |
597 | bool | |
598 | read_attributes(); | |
599 | ||
600 | // Set the name of the DIE if present. | |
601 | void | |
602 | set_name(); | |
603 | ||
604 | // Set the linkage name if present. | |
605 | void | |
606 | set_linkage_name(); | |
607 | ||
608 | // Skip all the attributes of the DIE and return the offset | |
609 | // of the next DIE. | |
610 | off_t | |
611 | skip_attributes(); | |
612 | ||
613 | // The Dwarf_info_reader, for reading attributes. | |
614 | Dwarf_info_reader* dwinfo_; | |
615 | // The parent of this DIE. | |
616 | Dwarf_die* parent_; | |
617 | // Offset of this DIE within its compilation unit. | |
618 | off_t die_offset_; | |
619 | // Offset of the first attribute, relative to the beginning of the DIE. | |
620 | off_t attr_offset_; | |
621 | // Offset of the first child, relative to the compilation unit. | |
622 | off_t child_offset_; | |
623 | // Offset of the next sibling, relative to the compilation unit. | |
624 | off_t sibling_offset_; | |
625 | // The abbreviation table entry. | |
626 | const Abbrev_code* abbrev_code_; | |
627 | // The list of attributes. | |
628 | Attributes attributes_; | |
629 | // True if the attributes have been read. | |
630 | bool attributes_read_; | |
631 | // The following fields hold common attributes to avoid a linear | |
632 | // search through the attribute list. | |
633 | // The DIE name (DW_AT_name). | |
634 | const char* name_; | |
635 | // Offset of the name in the string table (for DW_FORM_strp). | |
636 | off_t name_off_; | |
637 | // The linkage name (DW_AT_linkage_name or DW_AT_MIPS_linkage_name). | |
638 | const char* linkage_name_; | |
639 | // Offset of the linkage name in the string table (for DW_FORM_strp). | |
640 | off_t linkage_name_off_; | |
641 | // Section index of the string table (for DW_FORM_strp). | |
642 | unsigned int string_shndx_; | |
643 | // The value of a DW_AT_specification attribute. | |
644 | off_t specification_; | |
645 | // The value of a DW_AT_abstract_origin attribute. | |
646 | off_t abstract_origin_; | |
647 | }; | |
648 | ||
649 | // This class is used to read the debug info from the .debug_info | |
650 | // or .debug_types sections. This is a base class that implements | |
651 | // the generic parsing of the compilation unit header and DIE | |
652 | // structure. The parse() method parses the entire section, and | |
653 | // calls the various visit_xxx() methods for each header. Clients | |
654 | // should derive a new class from this one and implement the | |
655 | // visit_compilation_unit() and visit_type_unit() functions. | |
656 | ||
657 | class Dwarf_info_reader | |
658 | { | |
659 | public: | |
660 | Dwarf_info_reader(bool is_type_unit, | |
661 | Relobj* object, | |
662 | const unsigned char* symtab, | |
663 | off_t symtab_size, | |
664 | unsigned int shndx, | |
665 | unsigned int reloc_shndx, | |
666 | unsigned int reloc_type) | |
667 | : is_type_unit_(is_type_unit), object_(object), symtab_(symtab), | |
668 | symtab_size_(symtab_size), shndx_(shndx), reloc_shndx_(reloc_shndx), | |
8787852d CC |
669 | reloc_type_(reloc_type), abbrev_shndx_(0), string_shndx_(0), |
670 | buffer_(NULL), buffer_end_(NULL), cu_offset_(0), cu_length_(0), | |
671 | offset_size_(0), address_size_(0), cu_version_(0), type_signature_(0), | |
ed5d6712 CC |
672 | type_offset_(0), abbrev_table_(), ranges_table_(this), |
673 | reloc_mapper_(NULL), string_buffer_(NULL), string_buffer_end_(NULL), | |
8787852d | 674 | owns_string_buffer_(false), string_output_section_offset_(0) |
c1027032 CC |
675 | { } |
676 | ||
677 | virtual | |
678 | ~Dwarf_info_reader() | |
679 | { | |
680 | if (this->reloc_mapper_ != NULL) | |
681 | delete this->reloc_mapper_; | |
682 | if (this->owns_string_buffer_ && this->string_buffer_ != NULL) | |
683 | delete[] this->string_buffer_; | |
684 | } | |
685 | ||
686 | // Begin parsing the debug info. This calls visit_compilation_unit() | |
687 | // or visit_type_unit() for each compilation or type unit found in the | |
688 | // section, and visit_die() for each top-level DIE. | |
689 | void | |
690 | parse(); | |
691 | ||
692 | // Return the abbrev code entry for a CODE. | |
693 | const Dwarf_abbrev_table::Abbrev_code* | |
694 | get_abbrev(unsigned int code) | |
695 | { return this->abbrev_table_.get_abbrev(code); } | |
696 | ||
697 | // Return a pointer to the DWARF info buffer at OFFSET. | |
698 | const unsigned char* | |
699 | buffer_at_offset(off_t offset) const | |
700 | { | |
701 | const unsigned char* p = this->buffer_ + this->cu_offset_ + offset; | |
702 | if (this->check_buffer(p + 1)) | |
703 | return p; | |
704 | return NULL; | |
705 | } | |
706 | ||
ed5d6712 CC |
707 | // Read a possibly unaligned integer of SIZE. |
708 | template <int valsize> | |
709 | inline typename elfcpp::Valtype_base<valsize>::Valtype | |
710 | read_from_pointer(const unsigned char* source); | |
711 | ||
712 | // Read a possibly unaligned integer of SIZE. Update SOURCE after read. | |
713 | template <int valsize> | |
714 | inline typename elfcpp::Valtype_base<valsize>::Valtype | |
715 | read_from_pointer(const unsigned char** source); | |
716 | ||
c1027032 CC |
717 | // Look for a relocation at offset ATTR_OFF in the dwarf info, |
718 | // and return the section index and offset of the target. | |
719 | unsigned int | |
720 | lookup_reloc(off_t attr_off, off_t* target_off); | |
721 | ||
722 | // Return a string from the DWARF string table. | |
723 | const char* | |
724 | get_string(off_t str_off, unsigned int string_shndx); | |
725 | ||
726 | // Return the size of a DWARF offset. | |
727 | unsigned int | |
728 | offset_size() const | |
729 | { return this->offset_size_; } | |
730 | ||
731 | // Return the size of an address. | |
732 | unsigned int | |
733 | address_size() const | |
734 | { return this->address_size_; } | |
735 | ||
8787852d CC |
736 | // Set the section index of the .debug_abbrev section. |
737 | // We use this if there are no relocations for the .debug_info section. | |
738 | // If not set, the code parse() routine will search for the section by name. | |
739 | void | |
740 | set_abbrev_shndx(unsigned int abbrev_shndx) | |
741 | { this->abbrev_shndx_ = abbrev_shndx; } | |
742 | ||
c1027032 CC |
743 | protected: |
744 | // Begin parsing the debug info. This calls visit_compilation_unit() | |
745 | // or visit_type_unit() for each compilation or type unit found in the | |
746 | // section, and visit_die() for each top-level DIE. | |
747 | template<bool big_endian> | |
748 | void | |
749 | do_parse(); | |
750 | ||
751 | // The following methods are hooks that are meant to be implemented | |
752 | // by a derived class. A default, do-nothing, implementation of | |
753 | // each is provided for this base class. | |
754 | ||
755 | // Visit a compilation unit. | |
756 | virtual void | |
757 | visit_compilation_unit(off_t cu_offset, off_t cu_length, Dwarf_die* root_die); | |
758 | ||
759 | // Visit a type unit. | |
760 | virtual void | |
d5834edb CC |
761 | visit_type_unit(off_t tu_offset, off_t tu_length, off_t type_offset, |
762 | uint64_t signature, Dwarf_die* root_die); | |
c1027032 CC |
763 | |
764 | // Read the range table. | |
765 | Dwarf_range_list* | |
766 | read_range_list(unsigned int ranges_shndx, off_t ranges_offset) | |
767 | { | |
768 | return this->ranges_table_.read_range_list(this->object_, | |
769 | this->symtab_, | |
770 | this->symtab_size_, | |
771 | this->address_size_, | |
772 | ranges_shndx, | |
773 | ranges_offset); | |
774 | } | |
775 | ||
776 | // Return the object. | |
777 | Relobj* | |
778 | object() const | |
779 | { return this->object_; } | |
780 | ||
781 | // Return a pointer to the object file's ELF symbol table. | |
782 | const unsigned char* | |
783 | symtab() const | |
784 | { return this->symtab_; } | |
785 | ||
786 | // Return the size of the object file's ELF symbol table. | |
787 | off_t | |
788 | symtab_size() const | |
789 | { return this->symtab_size_; } | |
790 | ||
791 | // Checkpoint the relocation tracker. | |
792 | uint64_t | |
793 | get_reloc_checkpoint() const | |
794 | { return this->reloc_mapper_->checkpoint(); } | |
795 | ||
796 | // Reset the relocation tracker to the CHECKPOINT. | |
797 | void | |
798 | reset_relocs(uint64_t checkpoint) | |
799 | { this->reloc_mapper_->reset(checkpoint); } | |
800 | ||
801 | private: | |
a68a081d CC |
802 | // Print a warning about a corrupt debug section. |
803 | void | |
804 | warn_corrupt_debug_section() const; | |
805 | ||
c1027032 CC |
806 | // Check that P is within the bounds of the current section. |
807 | bool | |
a68a081d CC |
808 | check_buffer(const unsigned char* p) const |
809 | { | |
810 | if (p > this->buffer_ + this->cu_offset_ + this->cu_length_) | |
811 | { | |
812 | this->warn_corrupt_debug_section(); | |
813 | return false; | |
814 | } | |
815 | return true; | |
816 | } | |
c1027032 CC |
817 | |
818 | // Read the DWARF string table. | |
819 | bool | |
820 | read_string_table(unsigned int string_shndx) | |
821 | { | |
822 | // If we've already read this string table, return immediately. | |
823 | if (this->string_shndx_ > 0 && this->string_shndx_ == string_shndx) | |
824 | return true; | |
825 | if (string_shndx == 0 && this->string_shndx_ > 0) | |
826 | return true; | |
827 | return this->do_read_string_table(string_shndx); | |
828 | } | |
829 | ||
830 | bool | |
831 | do_read_string_table(unsigned int string_shndx); | |
832 | ||
833 | // True if this is a type unit; false for a compilation unit. | |
834 | bool is_type_unit_; | |
835 | // The object containing the .debug_info or .debug_types input section. | |
836 | Relobj* object_; | |
837 | // The ELF symbol table. | |
838 | const unsigned char* symtab_; | |
839 | // The size of the ELF symbol table. | |
840 | off_t symtab_size_; | |
841 | // Index of the .debug_info or .debug_types section. | |
842 | unsigned int shndx_; | |
843 | // Index of the relocation section. | |
844 | unsigned int reloc_shndx_; | |
845 | // Type of the relocation section (SHT_REL or SHT_RELA). | |
846 | unsigned int reloc_type_; | |
8787852d CC |
847 | // Index of the .debug_abbrev section (0 if not known). |
848 | unsigned int abbrev_shndx_; | |
c1027032 CC |
849 | // Index of the .debug_str section. |
850 | unsigned int string_shndx_; | |
851 | // The buffer for the debug info. | |
852 | const unsigned char* buffer_; | |
853 | const unsigned char* buffer_end_; | |
854 | // Offset of the current compilation unit. | |
855 | off_t cu_offset_; | |
856 | // Length of the current compilation unit. | |
857 | off_t cu_length_; | |
858 | // Size of a DWARF offset for the current compilation unit. | |
859 | unsigned int offset_size_; | |
860 | // Size of an address for the target architecture. | |
861 | unsigned int address_size_; | |
862 | // Compilation unit version number. | |
863 | unsigned int cu_version_; | |
864 | // Type signature (for a type unit). | |
865 | uint64_t type_signature_; | |
866 | // Offset from the type unit header to the type DIE (for a type unit). | |
867 | off_t type_offset_; | |
868 | // Abbreviations table for current compilation unit. | |
869 | Dwarf_abbrev_table abbrev_table_; | |
870 | // Ranges table for the current compilation unit. | |
871 | Dwarf_ranges_table ranges_table_; | |
872 | // Relocation mapper for the section. | |
873 | Elf_reloc_mapper* reloc_mapper_; | |
874 | // The buffer for the debug string table. | |
875 | const char* string_buffer_; | |
876 | const char* string_buffer_end_; | |
877 | // True if this object owns the buffer and needs to delete it. | |
878 | bool owns_string_buffer_; | |
879 | // For incremental update links, this will hold the offset of the | |
880 | // input .debug_str section within the output section. Offsets read | |
881 | // from relocated data will be relative to the output section, and need | |
882 | // to be corrected before reading data from the input section. | |
883 | uint64_t string_output_section_offset_; | |
884 | }; | |
885 | ||
79e052ea ILT |
886 | // We can't do better than to keep the offsets in a sorted vector. |
887 | // Here, offset is the key, and file_num/line_num is the value. | |
888 | struct Offset_to_lineno_entry | |
889 | { | |
890 | off_t offset; | |
891 | int header_num; // which file-list to use (i.e. which .o file are we in) | |
71ff8986 ILT |
892 | // A pointer into files_. |
893 | unsigned int file_num : sizeof(int) * CHAR_BIT - 1; | |
894 | // True if this was the last entry for the current offset, meaning | |
895 | // it's the line that actually applies. | |
896 | unsigned int last_line_for_offset : 1; | |
897 | // The line number in the source file. -1 to indicate end-of-function. | |
898 | int line_num; | |
899 | ||
900 | // This sorts by offsets first, and then puts the correct line to | |
901 | // report for a given offset at the beginning of the run of equal | |
902 | // offsets (so that asking for 1 line gives the best answer). This | |
903 | // is not a total ordering. | |
79e052ea | 904 | bool operator<(const Offset_to_lineno_entry& that) const |
71ff8986 ILT |
905 | { |
906 | if (this->offset != that.offset) | |
907 | return this->offset < that.offset; | |
908 | // Note the '>' which makes this sort 'true' first. | |
909 | return this->last_line_for_offset > that.last_line_for_offset; | |
910 | } | |
79e052ea ILT |
911 | }; |
912 | ||
5c2c6c95 ILT |
913 | // This class is used to read the line information from the debugging |
914 | // section of an object file. | |
915 | ||
916 | class Dwarf_line_info | |
917 | { | |
918 | public: | |
a55ce7fe ILT |
919 | Dwarf_line_info() |
920 | { } | |
921 | ||
922 | virtual | |
923 | ~Dwarf_line_info() | |
924 | { } | |
5c2c6c95 ILT |
925 | |
926 | // Given a section number and an offset, returns the associated | |
927 | // file and line-number, as a string: "file:lineno". If unable | |
928 | // to do the mapping, returns the empty string. You must call | |
71ff8986 ILT |
929 | // read_line_mappings() before calling this function. If |
930 | // 'other_lines' is non-NULL, fills that in with other line | |
931 | // numbers assigned to the same offset. | |
5c2c6c95 | 932 | std::string |
71ff8986 ILT |
933 | addr2line(unsigned int shndx, off_t offset, |
934 | std::vector<std::string>* other_lines) | |
935 | { return this->do_addr2line(shndx, offset, other_lines); } | |
a55ce7fe | 936 | |
e4e5049b CS |
937 | // A helper function for a single addr2line lookup. It also keeps a |
938 | // cache of the last CACHE_SIZE Dwarf_line_info objects it created; | |
939 | // set to 0 not to cache at all. The larger CACHE_SIZE is, the more | |
940 | // chance this routine won't have to re-create a Dwarf_line_info | |
941 | // object for its addr2line computation; such creations are slow. | |
942 | // NOTE: Not thread-safe, so only call from one thread at a time. | |
a55ce7fe | 943 | static std::string |
e4e5049b | 944 | one_addr2line(Object* object, unsigned int shndx, off_t offset, |
71ff8986 | 945 | size_t cache_size, std::vector<std::string>* other_lines); |
e4e5049b CS |
946 | |
947 | // This reclaims all the memory that one_addr2line may have cached. | |
948 | // Use this when you know you will not be calling one_addr2line again. | |
949 | static void | |
950 | clear_addr2line_cache(); | |
5c2c6c95 ILT |
951 | |
952 | private: | |
a55ce7fe | 953 | virtual std::string |
71ff8986 ILT |
954 | do_addr2line(unsigned int shndx, off_t offset, |
955 | std::vector<std::string>* other_lines) = 0; | |
a55ce7fe ILT |
956 | }; |
957 | ||
958 | template<int size, bool big_endian> | |
a18f2bd6 | 959 | class Sized_dwarf_line_info : public Dwarf_line_info |
a55ce7fe ILT |
960 | { |
961 | public: | |
962 | // Initializes a .debug_line reader for a given object file. | |
9430daf8 ILT |
963 | // If SHNDX is specified and non-negative, only read the debug |
964 | // information that pertains to the specified section. | |
75aea3d0 | 965 | Sized_dwarf_line_info(Object* object, unsigned int read_shndx = -1U); |
a55ce7fe | 966 | |
5dd8762a CC |
967 | virtual |
968 | ~Sized_dwarf_line_info() | |
969 | { | |
970 | if (this->buffer_start_ != NULL) | |
971 | delete[] this->buffer_start_; | |
972 | } | |
973 | ||
a55ce7fe ILT |
974 | private: |
975 | std::string | |
71ff8986 ILT |
976 | do_addr2line(unsigned int shndx, off_t offset, |
977 | std::vector<std::string>* other_lines); | |
978 | ||
979 | // Formats a file and line number to a string like "dirname/filename:lineno". | |
980 | std::string | |
981 | format_file_lineno(const Offset_to_lineno_entry& lineno) const; | |
a55ce7fe | 982 | |
24badc65 | 983 | // Start processing line info, and populates the offset_map_. |
9430daf8 ILT |
984 | // If SHNDX is non-negative, only store debug information that |
985 | // pertains to the specified section. | |
24badc65 | 986 | void |
c1027032 | 987 | read_line_mappings(unsigned int shndx); |
24badc65 | 988 | |
4c50553d ILT |
989 | // Reads the relocation section associated with .debug_line and |
990 | // stores relocation information in reloc_map_. | |
991 | void | |
c1027032 | 992 | read_relocs(); |
4c50553d | 993 | |
5c2c6c95 ILT |
994 | // Reads the DWARF2/3 header for this line info. Each takes as input |
995 | // a starting buffer position, and returns the ending position. | |
5c2c6c95 ILT |
996 | const unsigned char* |
997 | read_header_prolog(const unsigned char* lineptr); | |
998 | ||
999 | const unsigned char* | |
1000 | read_header_tables(const unsigned char* lineptr); | |
1001 | ||
9430daf8 ILT |
1002 | // Reads the DWARF2/3 line information. If shndx is non-negative, |
1003 | // discard all line information that doesn't pertain to the given | |
1004 | // section. | |
5c2c6c95 | 1005 | const unsigned char* |
75aea3d0 | 1006 | read_lines(const unsigned char* lineptr, unsigned int shndx); |
5c2c6c95 ILT |
1007 | |
1008 | // Process a single line info opcode at START using the state | |
1009 | // machine at LSM. Return true if we should define a line using the | |
1010 | // current state of the line state machine. Place the length of the | |
1011 | // opcode in LEN. | |
1012 | bool | |
e43872e9 | 1013 | process_one_opcode(const unsigned char* start, |
5c2c6c95 ILT |
1014 | struct LineStateMachine* lsm, size_t* len); |
1015 | ||
af674d1d ILT |
1016 | // Some parts of processing differ depending on whether the input |
1017 | // was a .o file or not. | |
1018 | bool input_is_relobj(); | |
1019 | ||
4c50553d ILT |
1020 | // If we saw anything amiss while parsing, we set this to false. |
1021 | // Then addr2line will always fail (rather than return possibly- | |
1022 | // corrupt data). | |
1023 | bool data_valid_; | |
1024 | ||
5c2c6c95 ILT |
1025 | // A DWARF2/3 line info header. This is not the same size as in the |
1026 | // actual file, as the one in the file may have a 32 bit or 64 bit | |
1027 | // lengths. | |
1028 | ||
1029 | struct Dwarf_line_infoHeader | |
1030 | { | |
1031 | off_t total_length; | |
1032 | int version; | |
1033 | off_t prologue_length; | |
1034 | int min_insn_length; // insn stands for instructin | |
1035 | bool default_is_stmt; // stmt stands for statement | |
1036 | signed char line_base; | |
1037 | int line_range; | |
1038 | unsigned char opcode_base; | |
1039 | std::vector<unsigned char> std_opcode_lengths; | |
1040 | int offset_size; | |
1041 | } header_; | |
1042 | ||
1043 | // buffer is the buffer for our line info, starting at exactly where | |
1044 | // the line info to read is. | |
1045 | const unsigned char* buffer_; | |
24badc65 | 1046 | const unsigned char* buffer_end_; |
5dd8762a CC |
1047 | // If the buffer was allocated temporarily, and therefore must be |
1048 | // deallocated in the dtor, this contains a pointer to the start | |
1049 | // of the buffer. | |
1050 | const unsigned char* buffer_start_; | |
5c2c6c95 | 1051 | |
4c50553d | 1052 | // This has relocations that point into buffer. |
c1027032 | 1053 | Sized_elf_reloc_mapper<size, big_endian>* reloc_mapper_; |
4dbfafcc ILT |
1054 | // The type of the reloc section in track_relocs_--SHT_REL or SHT_RELA. |
1055 | unsigned int track_relocs_type_; | |
4c50553d ILT |
1056 | |
1057 | // This is used to figure out what section to apply a relocation to. | |
24badc65 | 1058 | const unsigned char* symtab_buffer_; |
8383303e | 1059 | section_size_type symtab_buffer_size_; |
4c50553d | 1060 | |
af674d1d ILT |
1061 | // Holds the directories and files as we see them. We have an array |
1062 | // of directory-lists, one for each .o file we're reading (usually | |
1063 | // there will just be one, but there may be more if input is a .so). | |
1064 | std::vector<std::vector<std::string> > directories_; | |
5c2c6c95 | 1065 | // The first part is an index into directories_, the second the filename. |
af674d1d ILT |
1066 | std::vector<std::vector< std::pair<int, std::string> > > files_; |
1067 | ||
1068 | // An index into the current directories_ and files_ vectors. | |
1069 | int current_header_index_; | |
5c2c6c95 | 1070 | |
af674d1d ILT |
1071 | // A sorted map from offset of the relocation target to the shndx |
1072 | // and addend for the relocation. | |
c1027032 | 1073 | typedef std::map<off_t, std::pair<unsigned int, off_t> > |
4c50553d ILT |
1074 | Reloc_map; |
1075 | Reloc_map reloc_map_; | |
1076 | ||
5c2c6c95 ILT |
1077 | // We have a vector of offset->lineno entries for every input section. |
1078 | typedef Unordered_map<unsigned int, std::vector<Offset_to_lineno_entry> > | |
1079 | Lineno_map; | |
1080 | ||
1081 | Lineno_map line_number_map_; | |
1082 | }; | |
1083 | ||
1084 | } // End namespace gold. | |
1085 | ||
1086 | #endif // !defined(GOLD_DWARF_READER_H) |