1 //===-- DWARFUnit.cpp -----------------------------------------------------===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8
9 #include "DWARFUnit.h"
10
11 #include "lldb/Core/Module.h"
12 #include "lldb/Host/StringConvert.h"
13 #include "lldb/Symbol/ObjectFile.h"
14 #include "lldb/Utility/LLDBAssert.h"
15 #include "lldb/Utility/StreamString.h"
16 #include "lldb/Utility/Timer.h"
17 #include "llvm/Object/Error.h"
18
19 #include "DWARFCompileUnit.h"
20 #include "DWARFDebugAranges.h"
21 #include "DWARFDebugInfo.h"
22 #include "DWARFTypeUnit.h"
23 #include "LogChannelDWARF.h"
24 #include "SymbolFileDWARFDwo.h"
25
26 using namespace lldb;
27 using namespace lldb_private;
28 using namespace std;
29
30 extern int g_verbose;
31
DWARFUnit(SymbolFileDWARF & dwarf,lldb::user_id_t uid,const DWARFUnitHeader & header,const DWARFAbbreviationDeclarationSet & abbrevs,DIERef::Section section,bool is_dwo)32 DWARFUnit::DWARFUnit(SymbolFileDWARF &dwarf, lldb::user_id_t uid,
33 const DWARFUnitHeader &header,
34 const DWARFAbbreviationDeclarationSet &abbrevs,
35 DIERef::Section section, bool is_dwo)
36 : UserID(uid), m_dwarf(dwarf), m_header(header), m_abbrevs(&abbrevs),
37 m_cancel_scopes(false), m_section(section), m_is_dwo(is_dwo) {}
38
39 DWARFUnit::~DWARFUnit() = default;
40
41 // Parses first DIE of a compile unit.
ExtractUnitDIEIfNeeded()42 void DWARFUnit::ExtractUnitDIEIfNeeded() {
43 {
44 llvm::sys::ScopedReader lock(m_first_die_mutex);
45 if (m_first_die)
46 return; // Already parsed
47 }
48 llvm::sys::ScopedWriter lock(m_first_die_mutex);
49 if (m_first_die)
50 return; // Already parsed
51
52 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
53 Timer scoped_timer(func_cat, "%8.8x: DWARFUnit::ExtractUnitDIEIfNeeded()",
54 GetOffset());
55
56 // Set the offset to that of the first DIE and calculate the start of the
57 // next compilation unit header.
58 lldb::offset_t offset = GetFirstDIEOffset();
59
60 // We are in our compile unit, parse starting at the offset we were told to
61 // parse
62 const DWARFDataExtractor &data = GetData();
63 if (offset < GetNextUnitOffset() &&
64 m_first_die.Extract(data, this, &offset)) {
65 AddUnitDIE(m_first_die);
66 return;
67 }
68 }
69
70 // Parses a compile unit and indexes its DIEs if it hasn't already been done.
71 // It will leave this compile unit extracted forever.
ExtractDIEsIfNeeded()72 void DWARFUnit::ExtractDIEsIfNeeded() {
73 m_cancel_scopes = true;
74
75 {
76 llvm::sys::ScopedReader lock(m_die_array_mutex);
77 if (!m_die_array.empty())
78 return; // Already parsed
79 }
80 llvm::sys::ScopedWriter lock(m_die_array_mutex);
81 if (!m_die_array.empty())
82 return; // Already parsed
83
84 ExtractDIEsRWLocked();
85 }
86
87 // Parses a compile unit and indexes its DIEs if it hasn't already been done.
88 // It will clear this compile unit after returned instance gets out of scope,
89 // no other ScopedExtractDIEs instance is running for this compile unit
90 // and no ExtractDIEsIfNeeded() has been executed during this ScopedExtractDIEs
91 // lifetime.
ExtractDIEsScoped()92 DWARFUnit::ScopedExtractDIEs DWARFUnit::ExtractDIEsScoped() {
93 ScopedExtractDIEs scoped(*this);
94
95 {
96 llvm::sys::ScopedReader lock(m_die_array_mutex);
97 if (!m_die_array.empty())
98 return scoped; // Already parsed
99 }
100 llvm::sys::ScopedWriter lock(m_die_array_mutex);
101 if (!m_die_array.empty())
102 return scoped; // Already parsed
103
104 // Otherwise m_die_array would be already populated.
105 lldbassert(!m_cancel_scopes);
106
107 ExtractDIEsRWLocked();
108 scoped.m_clear_dies = true;
109 return scoped;
110 }
111
ScopedExtractDIEs(DWARFUnit & cu)112 DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(DWARFUnit &cu) : m_cu(&cu) {
113 m_cu->m_die_array_scoped_mutex.lock_shared();
114 }
115
~ScopedExtractDIEs()116 DWARFUnit::ScopedExtractDIEs::~ScopedExtractDIEs() {
117 if (!m_cu)
118 return;
119 m_cu->m_die_array_scoped_mutex.unlock_shared();
120 if (!m_clear_dies || m_cu->m_cancel_scopes)
121 return;
122 // Be sure no other ScopedExtractDIEs is running anymore.
123 llvm::sys::ScopedWriter lock_scoped(m_cu->m_die_array_scoped_mutex);
124 llvm::sys::ScopedWriter lock(m_cu->m_die_array_mutex);
125 if (m_cu->m_cancel_scopes)
126 return;
127 m_cu->ClearDIEsRWLocked();
128 }
129
ScopedExtractDIEs(ScopedExtractDIEs && rhs)130 DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(ScopedExtractDIEs &&rhs)
131 : m_cu(rhs.m_cu), m_clear_dies(rhs.m_clear_dies) {
132 rhs.m_cu = nullptr;
133 }
134
operator =(DWARFUnit::ScopedExtractDIEs && rhs)135 DWARFUnit::ScopedExtractDIEs &DWARFUnit::ScopedExtractDIEs::operator=(
136 DWARFUnit::ScopedExtractDIEs &&rhs) {
137 m_cu = rhs.m_cu;
138 rhs.m_cu = nullptr;
139 m_clear_dies = rhs.m_clear_dies;
140 return *this;
141 }
142
143 // Parses a compile unit and indexes its DIEs, m_die_array_mutex must be
144 // held R/W and m_die_array must be empty.
ExtractDIEsRWLocked()145 void DWARFUnit::ExtractDIEsRWLocked() {
146 llvm::sys::ScopedWriter first_die_lock(m_first_die_mutex);
147
148 static Timer::Category func_cat(LLVM_PRETTY_FUNCTION);
149 Timer scoped_timer(func_cat, "%8.8x: DWARFUnit::ExtractDIEsIfNeeded()",
150 GetOffset());
151
152 // Set the offset to that of the first DIE and calculate the start of the
153 // next compilation unit header.
154 lldb::offset_t offset = GetFirstDIEOffset();
155 lldb::offset_t next_cu_offset = GetNextUnitOffset();
156
157 DWARFDebugInfoEntry die;
158
159 uint32_t depth = 0;
160 // We are in our compile unit, parse starting at the offset we were told to
161 // parse
162 const DWARFDataExtractor &data = GetData();
163 std::vector<uint32_t> die_index_stack;
164 die_index_stack.reserve(32);
165 die_index_stack.push_back(0);
166 bool prev_die_had_children = false;
167 while (offset < next_cu_offset && die.Extract(data, this, &offset)) {
168 const bool null_die = die.IsNULL();
169 if (depth == 0) {
170 assert(m_die_array.empty() && "Compile unit DIE already added");
171
172 // The average bytes per DIE entry has been seen to be around 14-20 so
173 // lets pre-reserve half of that since we are now stripping the NULL
174 // tags.
175
176 // Only reserve the memory if we are adding children of the main
177 // compile unit DIE. The compile unit DIE is always the first entry, so
178 // if our size is 1, then we are adding the first compile unit child
179 // DIE and should reserve the memory.
180 m_die_array.reserve(GetDebugInfoSize() / 24);
181 m_die_array.push_back(die);
182
183 if (!m_first_die)
184 AddUnitDIE(m_die_array.front());
185
186 // With -fsplit-dwarf-inlining, clang will emit non-empty skeleton compile
187 // units. We are not able to access these DIE *and* the dwo file
188 // simultaneously. We also don't need to do that as the dwo file will
189 // contain a superset of information. So, we don't even attempt to parse
190 // any remaining DIEs.
191 if (m_dwo) {
192 m_die_array.front().SetHasChildren(false);
193 break;
194 }
195
196 } else {
197 if (null_die) {
198 if (prev_die_had_children) {
199 // This will only happen if a DIE says is has children but all it
200 // contains is a NULL tag. Since we are removing the NULL DIEs from
201 // the list (saves up to 25% in C++ code), we need a way to let the
202 // DIE know that it actually doesn't have children.
203 if (!m_die_array.empty())
204 m_die_array.back().SetHasChildren(false);
205 }
206 } else {
207 die.SetParentIndex(m_die_array.size() - die_index_stack[depth - 1]);
208
209 if (die_index_stack.back())
210 m_die_array[die_index_stack.back()].SetSiblingIndex(
211 m_die_array.size() - die_index_stack.back());
212
213 // Only push the DIE if it isn't a NULL DIE
214 m_die_array.push_back(die);
215 }
216 }
217
218 if (null_die) {
219 // NULL DIE.
220 if (!die_index_stack.empty())
221 die_index_stack.pop_back();
222
223 if (depth > 0)
224 --depth;
225 prev_die_had_children = false;
226 } else {
227 die_index_stack.back() = m_die_array.size() - 1;
228 // Normal DIE
229 const bool die_has_children = die.HasChildren();
230 if (die_has_children) {
231 die_index_stack.push_back(0);
232 ++depth;
233 }
234 prev_die_had_children = die_has_children;
235 }
236
237 if (depth == 0)
238 break; // We are done with this compile unit!
239 }
240
241 if (!m_die_array.empty()) {
242 if (m_first_die) {
243 // Only needed for the assertion.
244 m_first_die.SetHasChildren(m_die_array.front().HasChildren());
245 lldbassert(m_first_die == m_die_array.front());
246 }
247 m_first_die = m_die_array.front();
248 }
249
250 m_die_array.shrink_to_fit();
251
252 if (m_dwo)
253 m_dwo->ExtractDIEsIfNeeded();
254 }
255
256 // This is used when a split dwarf is enabled.
257 // A skeleton compilation unit may contain the DW_AT_str_offsets_base attribute
258 // that points to the first string offset of the CU contribution to the
259 // .debug_str_offsets. At the same time, the corresponding split debug unit also
260 // may use DW_FORM_strx* forms pointing to its own .debug_str_offsets.dwo and
261 // for that case, we should find the offset (skip the section header).
SetDwoStrOffsetsBase()262 void DWARFUnit::SetDwoStrOffsetsBase() {
263 lldb::offset_t baseOffset = 0;
264
265 if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
266 if (const auto *contribution =
267 entry->getContribution(llvm::DW_SECT_STR_OFFSETS))
268 baseOffset = contribution->Offset;
269 else
270 return;
271 }
272
273 if (GetVersion() >= 5) {
274 const DWARFDataExtractor &strOffsets =
275 GetSymbolFileDWARF().GetDWARFContext().getOrLoadStrOffsetsData();
276 uint64_t length = strOffsets.GetU32(&baseOffset);
277 if (length == 0xffffffff)
278 length = strOffsets.GetU64(&baseOffset);
279
280 // Check version.
281 if (strOffsets.GetU16(&baseOffset) < 5)
282 return;
283
284 // Skip padding.
285 baseOffset += 2;
286 }
287
288 SetStrOffsetsBase(baseOffset);
289 }
290
291 // m_die_array_mutex must be already held as read/write.
AddUnitDIE(const DWARFDebugInfoEntry & cu_die)292 void DWARFUnit::AddUnitDIE(const DWARFDebugInfoEntry &cu_die) {
293 llvm::Optional<uint64_t> addr_base, gnu_addr_base, ranges_base,
294 gnu_ranges_base;
295
296 DWARFAttributes attributes;
297 size_t num_attributes = cu_die.GetAttributes(this, attributes);
298
299 // Extract DW_AT_addr_base first, as other attributes may need it.
300 for (size_t i = 0; i < num_attributes; ++i) {
301 if (attributes.AttributeAtIndex(i) != DW_AT_addr_base)
302 continue;
303 DWARFFormValue form_value;
304 if (attributes.ExtractFormValueAtIndex(i, form_value)) {
305 addr_base = form_value.Unsigned();
306 SetAddrBase(*addr_base);
307 break;
308 }
309 }
310
311 for (size_t i = 0; i < num_attributes; ++i) {
312 dw_attr_t attr = attributes.AttributeAtIndex(i);
313 DWARFFormValue form_value;
314 if (!attributes.ExtractFormValueAtIndex(i, form_value))
315 continue;
316 switch (attr) {
317 case DW_AT_loclists_base:
318 SetLoclistsBase(form_value.Unsigned());
319 break;
320 case DW_AT_rnglists_base:
321 ranges_base = form_value.Unsigned();
322 SetRangesBase(*ranges_base);
323 break;
324 case DW_AT_str_offsets_base:
325 SetStrOffsetsBase(form_value.Unsigned());
326 break;
327 case DW_AT_low_pc:
328 SetBaseAddress(form_value.Address());
329 break;
330 case DW_AT_entry_pc:
331 // If the value was already set by DW_AT_low_pc, don't update it.
332 if (m_base_addr == LLDB_INVALID_ADDRESS)
333 SetBaseAddress(form_value.Address());
334 break;
335 case DW_AT_stmt_list:
336 m_line_table_offset = form_value.Unsigned();
337 break;
338 case DW_AT_GNU_addr_base:
339 gnu_addr_base = form_value.Unsigned();
340 break;
341 case DW_AT_GNU_ranges_base:
342 gnu_ranges_base = form_value.Unsigned();
343 break;
344 }
345 }
346
347 if (m_is_dwo) {
348 SetDwoStrOffsetsBase();
349 return;
350 }
351
352 std::shared_ptr<SymbolFileDWARFDwo> dwo_symbol_file =
353 m_dwarf.GetDwoSymbolFileForCompileUnit(*this, cu_die);
354 if (!dwo_symbol_file)
355 return;
356
357 uint64_t main_dwo_id =
358 cu_die.GetAttributeValueAsUnsigned(this, DW_AT_GNU_dwo_id, 0);
359 DWARFUnit *dwo_cu = dwo_symbol_file->GetDWOCompileUnitForHash(main_dwo_id);
360 if (!dwo_cu)
361 return; // Can't fetch the compile unit from the dwo file.
362 dwo_cu->SetUserData(this);
363
364 DWARFBaseDIE dwo_cu_die = dwo_cu->GetUnitDIEOnly();
365 if (!dwo_cu_die.IsValid())
366 return; // Can't fetch the compile unit DIE from the dwo file.
367
368 // Here for DWO CU we want to use the address base set in the skeleton unit
369 // (DW_AT_addr_base) if it is available and use the DW_AT_GNU_addr_base
370 // otherwise. We do that because pre-DWARF v5 could use the DW_AT_GNU_*
371 // attributes which were applicable to the DWO units. The corresponding
372 // DW_AT_* attributes standardized in DWARF v5 are also applicable to the main
373 // unit in contrast.
374 if (addr_base)
375 dwo_cu->SetAddrBase(*addr_base);
376 else if (gnu_addr_base)
377 dwo_cu->SetAddrBase(*gnu_addr_base);
378
379 if (GetVersion() <= 4 && gnu_ranges_base)
380 dwo_cu->SetRangesBase(*gnu_ranges_base);
381 else if (dwo_symbol_file->GetDWARFContext()
382 .getOrLoadRngListsData()
383 .GetByteSize() > 0)
384 dwo_cu->SetRangesBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
385
386 if (GetVersion() >= 5 &&
387 dwo_symbol_file->GetDWARFContext().getOrLoadLocListsData().GetByteSize() >
388 0)
389 dwo_cu->SetLoclistsBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
390 dwo_cu->SetBaseAddress(GetBaseAddress());
391
392 m_dwo = std::shared_ptr<DWARFUnit>(std::move(dwo_symbol_file), dwo_cu);
393 }
394
LookupAddress(const dw_addr_t address)395 DWARFDIE DWARFUnit::LookupAddress(const dw_addr_t address) {
396 if (DIE()) {
397 const DWARFDebugAranges &func_aranges = GetFunctionAranges();
398
399 // Re-check the aranges auto pointer contents in case it was created above
400 if (!func_aranges.IsEmpty())
401 return GetDIE(func_aranges.FindAddress(address));
402 }
403 return DWARFDIE();
404 }
405
GetDebugInfoSize() const406 size_t DWARFUnit::GetDebugInfoSize() const {
407 return GetLengthByteSize() + GetLength() - GetHeaderByteSize();
408 }
409
GetAbbreviations() const410 const DWARFAbbreviationDeclarationSet *DWARFUnit::GetAbbreviations() const {
411 return m_abbrevs;
412 }
413
GetAbbrevOffset() const414 dw_offset_t DWARFUnit::GetAbbrevOffset() const {
415 return m_abbrevs ? m_abbrevs->GetOffset() : DW_INVALID_OFFSET;
416 }
417
GetLineTableOffset()418 dw_offset_t DWARFUnit::GetLineTableOffset() {
419 ExtractUnitDIEIfNeeded();
420 return m_line_table_offset;
421 }
422
SetAddrBase(dw_addr_t addr_base)423 void DWARFUnit::SetAddrBase(dw_addr_t addr_base) { m_addr_base = addr_base; }
424
425 // Parse the rangelist table header, including the optional array of offsets
426 // following it (DWARF v5 and later).
427 template <typename ListTableType>
428 static llvm::Expected<ListTableType>
ParseListTableHeader(const llvm::DWARFDataExtractor & data,uint64_t offset,DwarfFormat format)429 ParseListTableHeader(const llvm::DWARFDataExtractor &data, uint64_t offset,
430 DwarfFormat format) {
431 // We are expected to be called with Offset 0 or pointing just past the table
432 // header. Correct Offset in the latter case so that it points to the start
433 // of the header.
434 if (offset > 0) {
435 uint64_t HeaderSize = llvm::DWARFListTableHeader::getHeaderSize(format);
436 if (offset < HeaderSize)
437 return llvm::createStringError(errc::invalid_argument,
438 "did not detect a valid"
439 " list table with base = 0x%" PRIx64 "\n",
440 offset);
441 offset -= HeaderSize;
442 }
443 ListTableType Table;
444 if (llvm::Error E = Table.extractHeaderAndOffsets(data, &offset))
445 return std::move(E);
446 return Table;
447 }
448
SetLoclistsBase(dw_addr_t loclists_base)449 void DWARFUnit::SetLoclistsBase(dw_addr_t loclists_base) {
450 m_loclists_base = loclists_base;
451
452 uint64_t header_size = llvm::DWARFListTableHeader::getHeaderSize(DWARF32);
453 if (loclists_base < header_size)
454 return;
455
456 m_loclist_table_header.emplace(".debug_loclists", "locations");
457 uint64_t offset = loclists_base - header_size;
458 if (llvm::Error E = m_loclist_table_header->extract(
459 m_dwarf.GetDWARFContext().getOrLoadLocListsData().GetAsLLVM(),
460 &offset)) {
461 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
462 "Failed to extract location list table at offset 0x%" PRIx64 ": %s",
463 loclists_base, toString(std::move(E)).c_str());
464 }
465 }
466
467 std::unique_ptr<llvm::DWARFLocationTable>
GetLocationTable(const DataExtractor & data) const468 DWARFUnit::GetLocationTable(const DataExtractor &data) const {
469 llvm::DWARFDataExtractor llvm_data(
470 data.GetData(), data.GetByteOrder() == lldb::eByteOrderLittle,
471 data.GetAddressByteSize());
472
473 if (m_is_dwo || GetVersion() >= 5)
474 return std::make_unique<llvm::DWARFDebugLoclists>(llvm_data, GetVersion());
475 return std::make_unique<llvm::DWARFDebugLoc>(llvm_data);
476 }
477
GetLocationData() const478 DWARFDataExtractor DWARFUnit::GetLocationData() const {
479 DWARFContext &Ctx = GetSymbolFileDWARF().GetDWARFContext();
480 const DWARFDataExtractor &data =
481 GetVersion() >= 5 ? Ctx.getOrLoadLocListsData() : Ctx.getOrLoadLocData();
482 if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
483 if (const auto *contribution = entry->getContribution(llvm::DW_SECT_EXT_LOC))
484 return DWARFDataExtractor(data, contribution->Offset,
485 contribution->Length);
486 return DWARFDataExtractor();
487 }
488 return data;
489 }
490
SetRangesBase(dw_addr_t ranges_base)491 void DWARFUnit::SetRangesBase(dw_addr_t ranges_base) {
492 m_ranges_base = ranges_base;
493
494 if (GetVersion() < 5)
495 return;
496
497 if (auto table_or_error = ParseListTableHeader<llvm::DWARFDebugRnglistTable>(
498 m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(),
499 ranges_base, DWARF32))
500 m_rnglist_table = std::move(table_or_error.get());
501 else
502 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
503 "Failed to extract range list table at offset 0x%" PRIx64 ": %s",
504 ranges_base, toString(table_or_error.takeError()).c_str());
505 }
506
SetStrOffsetsBase(dw_offset_t str_offsets_base)507 void DWARFUnit::SetStrOffsetsBase(dw_offset_t str_offsets_base) {
508 m_str_offsets_base = str_offsets_base;
509 }
510
511 // It may be called only with m_die_array_mutex held R/W.
ClearDIEsRWLocked()512 void DWARFUnit::ClearDIEsRWLocked() {
513 m_die_array.clear();
514 m_die_array.shrink_to_fit();
515
516 if (m_dwo)
517 m_dwo->ClearDIEsRWLocked();
518 }
519
GetByteOrder() const520 lldb::ByteOrder DWARFUnit::GetByteOrder() const {
521 return m_dwarf.GetObjectFile()->GetByteOrder();
522 }
523
SetBaseAddress(dw_addr_t base_addr)524 void DWARFUnit::SetBaseAddress(dw_addr_t base_addr) { m_base_addr = base_addr; }
525
526 // Compare function DWARFDebugAranges::Range structures
CompareDIEOffset(const DWARFDebugInfoEntry & die,const dw_offset_t die_offset)527 static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
528 const dw_offset_t die_offset) {
529 return die.GetOffset() < die_offset;
530 }
531
532 // GetDIE()
533 //
534 // Get the DIE (Debug Information Entry) with the specified offset by first
535 // checking if the DIE is contained within this compile unit and grabbing the
536 // DIE from this compile unit. Otherwise we grab the DIE from the DWARF file.
537 DWARFDIE
GetDIE(dw_offset_t die_offset)538 DWARFUnit::GetDIE(dw_offset_t die_offset) {
539 if (die_offset != DW_INVALID_OFFSET) {
540 if (ContainsDIEOffset(die_offset)) {
541 ExtractDIEsIfNeeded();
542 DWARFDebugInfoEntry::const_iterator end = m_die_array.cend();
543 DWARFDebugInfoEntry::const_iterator pos =
544 lower_bound(m_die_array.cbegin(), end, die_offset, CompareDIEOffset);
545 if (pos != end) {
546 if (die_offset == (*pos).GetOffset())
547 return DWARFDIE(this, &(*pos));
548 }
549 } else
550 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
551 "GetDIE for DIE 0x%" PRIx32 " is outside of its CU 0x%" PRIx32,
552 die_offset, GetOffset());
553 }
554 return DWARFDIE(); // Not found
555 }
556
GetNonSkeletonUnit()557 DWARFUnit &DWARFUnit::GetNonSkeletonUnit() {
558 ExtractUnitDIEIfNeeded();
559 if (m_dwo)
560 return *m_dwo;
561 return *this;
562 }
563
GetAddressByteSize(const DWARFUnit * cu)564 uint8_t DWARFUnit::GetAddressByteSize(const DWARFUnit *cu) {
565 if (cu)
566 return cu->GetAddressByteSize();
567 return DWARFUnit::GetDefaultAddressSize();
568 }
569
GetDefaultAddressSize()570 uint8_t DWARFUnit::GetDefaultAddressSize() { return 4; }
571
GetUserData() const572 void *DWARFUnit::GetUserData() const { return m_user_data; }
573
SetUserData(void * d)574 void DWARFUnit::SetUserData(void *d) { m_user_data = d; }
575
Supports_DW_AT_APPLE_objc_complete_type()576 bool DWARFUnit::Supports_DW_AT_APPLE_objc_complete_type() {
577 return GetProducer() != eProducerLLVMGCC;
578 }
579
DW_AT_decl_file_attributes_are_invalid()580 bool DWARFUnit::DW_AT_decl_file_attributes_are_invalid() {
581 // llvm-gcc makes completely invalid decl file attributes and won't ever be
582 // fixed, so we need to know to ignore these.
583 return GetProducer() == eProducerLLVMGCC;
584 }
585
Supports_unnamed_objc_bitfields()586 bool DWARFUnit::Supports_unnamed_objc_bitfields() {
587 if (GetProducer() == eProducerClang) {
588 const uint32_t major_version = GetProducerVersionMajor();
589 return major_version > 425 ||
590 (major_version == 425 && GetProducerVersionUpdate() >= 13);
591 }
592 return true; // Assume all other compilers didn't have incorrect ObjC bitfield
593 // info
594 }
595
ParseProducerInfo()596 void DWARFUnit::ParseProducerInfo() {
597 m_producer_version_major = UINT32_MAX;
598 m_producer_version_minor = UINT32_MAX;
599 m_producer_version_update = UINT32_MAX;
600
601 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
602 if (die) {
603
604 const char *producer_cstr =
605 die->GetAttributeValueAsString(this, DW_AT_producer, nullptr);
606 if (producer_cstr) {
607 RegularExpression llvm_gcc_regex(
608 llvm::StringRef("^4\\.[012]\\.[01] \\(Based on Apple "
609 "Inc\\. build [0-9]+\\) \\(LLVM build "
610 "[\\.0-9]+\\)$"));
611 if (llvm_gcc_regex.Execute(llvm::StringRef(producer_cstr))) {
612 m_producer = eProducerLLVMGCC;
613 } else if (strstr(producer_cstr, "clang")) {
614 static RegularExpression g_clang_version_regex(
615 llvm::StringRef("clang-([0-9]+)\\.([0-9]+)\\.([0-9]+)"));
616 llvm::SmallVector<llvm::StringRef, 4> matches;
617 if (g_clang_version_regex.Execute(llvm::StringRef(producer_cstr),
618 &matches)) {
619 m_producer_version_major =
620 StringConvert::ToUInt32(matches[1].str().c_str(), UINT32_MAX, 10);
621 m_producer_version_minor =
622 StringConvert::ToUInt32(matches[2].str().c_str(), UINT32_MAX, 10);
623 m_producer_version_update =
624 StringConvert::ToUInt32(matches[3].str().c_str(), UINT32_MAX, 10);
625 }
626 m_producer = eProducerClang;
627 } else if (strstr(producer_cstr, "GNU"))
628 m_producer = eProducerGCC;
629 }
630 }
631 if (m_producer == eProducerInvalid)
632 m_producer = eProcucerOther;
633 }
634
GetProducer()635 DWARFProducer DWARFUnit::GetProducer() {
636 if (m_producer == eProducerInvalid)
637 ParseProducerInfo();
638 return m_producer;
639 }
640
GetProducerVersionMajor()641 uint32_t DWARFUnit::GetProducerVersionMajor() {
642 if (m_producer_version_major == 0)
643 ParseProducerInfo();
644 return m_producer_version_major;
645 }
646
GetProducerVersionMinor()647 uint32_t DWARFUnit::GetProducerVersionMinor() {
648 if (m_producer_version_minor == 0)
649 ParseProducerInfo();
650 return m_producer_version_minor;
651 }
652
GetProducerVersionUpdate()653 uint32_t DWARFUnit::GetProducerVersionUpdate() {
654 if (m_producer_version_update == 0)
655 ParseProducerInfo();
656 return m_producer_version_update;
657 }
658
GetDWARFLanguageType()659 uint64_t DWARFUnit::GetDWARFLanguageType() {
660 if (m_language_type)
661 return *m_language_type;
662
663 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
664 if (!die)
665 m_language_type = 0;
666 else
667 m_language_type = die->GetAttributeValueAsUnsigned(this, DW_AT_language, 0);
668 return *m_language_type;
669 }
670
GetIsOptimized()671 bool DWARFUnit::GetIsOptimized() {
672 if (m_is_optimized == eLazyBoolCalculate) {
673 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
674 if (die) {
675 m_is_optimized = eLazyBoolNo;
676 if (die->GetAttributeValueAsUnsigned(this, DW_AT_APPLE_optimized, 0) ==
677 1) {
678 m_is_optimized = eLazyBoolYes;
679 }
680 }
681 }
682 return m_is_optimized == eLazyBoolYes;
683 }
684
GetPathStyle()685 FileSpec::Style DWARFUnit::GetPathStyle() {
686 if (!m_comp_dir)
687 ComputeCompDirAndGuessPathStyle();
688 return m_comp_dir->GetPathStyle();
689 }
690
GetCompilationDirectory()691 const FileSpec &DWARFUnit::GetCompilationDirectory() {
692 if (!m_comp_dir)
693 ComputeCompDirAndGuessPathStyle();
694 return *m_comp_dir;
695 }
696
GetAbsolutePath()697 const FileSpec &DWARFUnit::GetAbsolutePath() {
698 if (!m_file_spec)
699 ComputeAbsolutePath();
700 return *m_file_spec;
701 }
702
GetFile(size_t file_idx)703 FileSpec DWARFUnit::GetFile(size_t file_idx) {
704 return m_dwarf.GetFile(*this, file_idx);
705 }
706
707 // DWARF2/3 suggests the form hostname:pathname for compilation directory.
708 // Remove the host part if present.
709 static llvm::StringRef
removeHostnameFromPathname(llvm::StringRef path_from_dwarf)710 removeHostnameFromPathname(llvm::StringRef path_from_dwarf) {
711 if (!path_from_dwarf.contains(':'))
712 return path_from_dwarf;
713 llvm::StringRef host, path;
714 std::tie(host, path) = path_from_dwarf.split(':');
715
716 if (host.contains('/'))
717 return path_from_dwarf;
718
719 // check whether we have a windows path, and so the first character is a
720 // drive-letter not a hostname.
721 if (host.size() == 1 && llvm::isAlpha(host[0]) && path.startswith("\\"))
722 return path_from_dwarf;
723
724 return path;
725 }
726
ComputeCompDirAndGuessPathStyle()727 void DWARFUnit::ComputeCompDirAndGuessPathStyle() {
728 m_comp_dir = FileSpec();
729 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
730 if (!die)
731 return;
732
733 llvm::StringRef comp_dir = removeHostnameFromPathname(
734 die->GetAttributeValueAsString(this, DW_AT_comp_dir, nullptr));
735 if (!comp_dir.empty()) {
736 FileSpec::Style comp_dir_style =
737 FileSpec::GuessPathStyle(comp_dir).getValueOr(FileSpec::Style::native);
738 m_comp_dir = FileSpec(comp_dir, comp_dir_style);
739 } else {
740 // Try to detect the style based on the DW_AT_name attribute, but just store
741 // the detected style in the m_comp_dir field.
742 const char *name =
743 die->GetAttributeValueAsString(this, DW_AT_name, nullptr);
744 m_comp_dir = FileSpec(
745 "", FileSpec::GuessPathStyle(name).getValueOr(FileSpec::Style::native));
746 }
747 }
748
ComputeAbsolutePath()749 void DWARFUnit::ComputeAbsolutePath() {
750 m_file_spec = FileSpec();
751 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
752 if (!die)
753 return;
754
755 m_file_spec =
756 FileSpec(die->GetAttributeValueAsString(this, DW_AT_name, nullptr),
757 GetPathStyle());
758
759 if (m_file_spec->IsRelative())
760 m_file_spec->MakeAbsolute(GetCompilationDirectory());
761 }
762
GetDwoSymbolFile()763 SymbolFileDWARFDwo *DWARFUnit::GetDwoSymbolFile() {
764 ExtractUnitDIEIfNeeded();
765 if (m_dwo)
766 return &llvm::cast<SymbolFileDWARFDwo>(m_dwo->GetSymbolFileDWARF());
767 return nullptr;
768 }
769
GetFunctionAranges()770 const DWARFDebugAranges &DWARFUnit::GetFunctionAranges() {
771 if (m_func_aranges_up == nullptr) {
772 m_func_aranges_up = std::make_unique<DWARFDebugAranges>();
773 const DWARFDebugInfoEntry *die = DIEPtr();
774 if (die)
775 die->BuildFunctionAddressRangeTable(this, m_func_aranges_up.get());
776
777 if (m_dwo) {
778 const DWARFDebugInfoEntry *dwo_die = m_dwo->DIEPtr();
779 if (dwo_die)
780 dwo_die->BuildFunctionAddressRangeTable(m_dwo.get(),
781 m_func_aranges_up.get());
782 }
783
784 const bool minimize = false;
785 m_func_aranges_up->Sort(minimize);
786 }
787 return *m_func_aranges_up;
788 }
789
790 llvm::Expected<DWARFUnitHeader>
extract(const DWARFDataExtractor & data,DIERef::Section section,lldb::offset_t * offset_ptr,const llvm::DWARFUnitIndex * index)791 DWARFUnitHeader::extract(const DWARFDataExtractor &data,
792 DIERef::Section section, lldb::offset_t *offset_ptr,
793 const llvm::DWARFUnitIndex *index) {
794 DWARFUnitHeader header;
795 header.m_offset = *offset_ptr;
796 if (index)
797 header.m_index_entry = index->getFromOffset(*offset_ptr);
798 header.m_length = data.GetDWARFInitialLength(offset_ptr);
799 header.m_version = data.GetU16(offset_ptr);
800 if (header.m_version == 5) {
801 header.m_unit_type = data.GetU8(offset_ptr);
802 header.m_addr_size = data.GetU8(offset_ptr);
803 header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
804 if (header.m_unit_type == llvm::dwarf::DW_UT_skeleton)
805 header.m_dwo_id = data.GetU64(offset_ptr);
806 } else {
807 header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
808 header.m_addr_size = data.GetU8(offset_ptr);
809 header.m_unit_type =
810 section == DIERef::Section::DebugTypes ? DW_UT_type : DW_UT_compile;
811 }
812
813 if (header.m_index_entry) {
814 if (header.m_abbr_offset) {
815 return llvm::createStringError(
816 llvm::inconvertibleErrorCode(),
817 "Package unit with a non-zero abbreviation offset");
818 }
819 auto *unit_contrib = header.m_index_entry->getContribution();
820 if (!unit_contrib || unit_contrib->Length != header.m_length + 4) {
821 return llvm::createStringError(llvm::inconvertibleErrorCode(),
822 "Inconsistent DWARF package unit index");
823 }
824 auto *abbr_entry =
825 header.m_index_entry->getContribution(llvm::DW_SECT_ABBREV);
826 if (!abbr_entry) {
827 return llvm::createStringError(
828 llvm::inconvertibleErrorCode(),
829 "DWARF package index missing abbreviation column");
830 }
831 header.m_abbr_offset = abbr_entry->Offset;
832 }
833 if (header.IsTypeUnit()) {
834 header.m_type_hash = data.GetU64(offset_ptr);
835 header.m_type_offset = data.GetDWARFOffset(offset_ptr);
836 }
837
838 bool length_OK = data.ValidOffset(header.GetNextUnitOffset() - 1);
839 bool version_OK = SymbolFileDWARF::SupportedVersion(header.m_version);
840 bool addr_size_OK = (header.m_addr_size == 4) || (header.m_addr_size == 8);
841 bool type_offset_OK =
842 !header.IsTypeUnit() || (header.m_type_offset <= header.GetLength());
843
844 if (!length_OK)
845 return llvm::make_error<llvm::object::GenericBinaryError>(
846 "Invalid unit length");
847 if (!version_OK)
848 return llvm::make_error<llvm::object::GenericBinaryError>(
849 "Unsupported unit version");
850 if (!addr_size_OK)
851 return llvm::make_error<llvm::object::GenericBinaryError>(
852 "Invalid unit address size");
853 if (!type_offset_OK)
854 return llvm::make_error<llvm::object::GenericBinaryError>(
855 "Type offset out of range");
856
857 return header;
858 }
859
860 llvm::Expected<DWARFUnitSP>
extract(SymbolFileDWARF & dwarf,user_id_t uid,const DWARFDataExtractor & debug_info,DIERef::Section section,lldb::offset_t * offset_ptr,const llvm::DWARFUnitIndex * index)861 DWARFUnit::extract(SymbolFileDWARF &dwarf, user_id_t uid,
862 const DWARFDataExtractor &debug_info,
863 DIERef::Section section, lldb::offset_t *offset_ptr,
864 const llvm::DWARFUnitIndex *index) {
865 assert(debug_info.ValidOffset(*offset_ptr));
866
867 auto expected_header =
868 DWARFUnitHeader::extract(debug_info, section, offset_ptr, index);
869 if (!expected_header)
870 return expected_header.takeError();
871
872 const DWARFDebugAbbrev *abbr = dwarf.DebugAbbrev();
873 if (!abbr)
874 return llvm::make_error<llvm::object::GenericBinaryError>(
875 "No debug_abbrev data");
876
877 bool abbr_offset_OK =
878 dwarf.GetDWARFContext().getOrLoadAbbrevData().ValidOffset(
879 expected_header->GetAbbrOffset());
880 if (!abbr_offset_OK)
881 return llvm::make_error<llvm::object::GenericBinaryError>(
882 "Abbreviation offset for unit is not valid");
883
884 const DWARFAbbreviationDeclarationSet *abbrevs =
885 abbr->GetAbbreviationDeclarationSet(expected_header->GetAbbrOffset());
886 if (!abbrevs)
887 return llvm::make_error<llvm::object::GenericBinaryError>(
888 "No abbrev exists at the specified offset.");
889
890 bool is_dwo = dwarf.GetDWARFContext().isDwo();
891 if (expected_header->IsTypeUnit())
892 return DWARFUnitSP(new DWARFTypeUnit(dwarf, uid, *expected_header, *abbrevs,
893 section, is_dwo));
894 return DWARFUnitSP(new DWARFCompileUnit(dwarf, uid, *expected_header,
895 *abbrevs, section, is_dwo));
896 }
897
GetData() const898 const lldb_private::DWARFDataExtractor &DWARFUnit::GetData() const {
899 return m_section == DIERef::Section::DebugTypes
900 ? m_dwarf.GetDWARFContext().getOrLoadDebugTypesData()
901 : m_dwarf.GetDWARFContext().getOrLoadDebugInfoData();
902 }
903
GetHeaderByteSize() const904 uint32_t DWARFUnit::GetHeaderByteSize() const {
905 switch (m_header.GetUnitType()) {
906 case llvm::dwarf::DW_UT_compile:
907 case llvm::dwarf::DW_UT_partial:
908 return GetVersion() < 5 ? 11 : 12;
909 case llvm::dwarf::DW_UT_skeleton:
910 case llvm::dwarf::DW_UT_split_compile:
911 return 20;
912 case llvm::dwarf::DW_UT_type:
913 case llvm::dwarf::DW_UT_split_type:
914 return GetVersion() < 5 ? 23 : 24;
915 }
916 llvm_unreachable("invalid UnitType.");
917 }
918
919 llvm::Optional<uint64_t>
GetStringOffsetSectionItem(uint32_t index) const920 DWARFUnit::GetStringOffsetSectionItem(uint32_t index) const {
921 offset_t offset = GetStrOffsetsBase() + index * 4;
922 return m_dwarf.GetDWARFContext().getOrLoadStrOffsetsData().GetU32(&offset);
923 }
924
925 llvm::Expected<DWARFRangeList>
FindRnglistFromOffset(dw_offset_t offset)926 DWARFUnit::FindRnglistFromOffset(dw_offset_t offset) {
927 if (GetVersion() <= 4) {
928 const DWARFDebugRanges *debug_ranges = m_dwarf.GetDebugRanges();
929 if (!debug_ranges)
930 return llvm::make_error<llvm::object::GenericBinaryError>(
931 "No debug_ranges section");
932 DWARFRangeList ranges;
933 debug_ranges->FindRanges(this, offset, ranges);
934 return ranges;
935 }
936
937 if (!m_rnglist_table)
938 return llvm::createStringError(errc::invalid_argument,
939 "missing or invalid range list table");
940
941 auto range_list_or_error = m_rnglist_table->findList(
942 m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(), offset);
943 if (!range_list_or_error)
944 return range_list_or_error.takeError();
945
946 llvm::Expected<llvm::DWARFAddressRangesVector> llvm_ranges =
947 range_list_or_error->getAbsoluteRanges(
948 llvm::object::SectionedAddress{GetBaseAddress()},
949 [&](uint32_t index) {
950 uint32_t index_size = GetAddressByteSize();
951 dw_offset_t addr_base = GetAddrBase();
952 lldb::offset_t offset = addr_base + index * index_size;
953 return llvm::object::SectionedAddress{
954 m_dwarf.GetDWARFContext().getOrLoadAddrData().GetMaxU64(
955 &offset, index_size)};
956 });
957 if (!llvm_ranges)
958 return llvm_ranges.takeError();
959
960 DWARFRangeList ranges;
961 for (const llvm::DWARFAddressRange &llvm_range : *llvm_ranges) {
962 ranges.Append(DWARFRangeList::Entry(llvm_range.LowPC,
963 llvm_range.HighPC - llvm_range.LowPC));
964 }
965 return ranges;
966 }
967
968 llvm::Expected<DWARFRangeList>
FindRnglistFromIndex(uint32_t index)969 DWARFUnit::FindRnglistFromIndex(uint32_t index) {
970 if (llvm::Optional<uint64_t> offset = GetRnglistOffset(index))
971 return FindRnglistFromOffset(*offset);
972 if (m_rnglist_table)
973 return llvm::createStringError(errc::invalid_argument,
974 "invalid range list table index %d", index);
975
976 return llvm::createStringError(errc::invalid_argument,
977 "missing or invalid range list table");
978 }
979