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
GetDebugInfoSize() const395 size_t DWARFUnit::GetDebugInfoSize() const {
396 return GetLengthByteSize() + GetLength() - GetHeaderByteSize();
397 }
398
GetAbbreviations() const399 const DWARFAbbreviationDeclarationSet *DWARFUnit::GetAbbreviations() const {
400 return m_abbrevs;
401 }
402
GetAbbrevOffset() const403 dw_offset_t DWARFUnit::GetAbbrevOffset() const {
404 return m_abbrevs ? m_abbrevs->GetOffset() : DW_INVALID_OFFSET;
405 }
406
GetLineTableOffset()407 dw_offset_t DWARFUnit::GetLineTableOffset() {
408 ExtractUnitDIEIfNeeded();
409 return m_line_table_offset;
410 }
411
SetAddrBase(dw_addr_t addr_base)412 void DWARFUnit::SetAddrBase(dw_addr_t addr_base) { m_addr_base = addr_base; }
413
414 // Parse the rangelist table header, including the optional array of offsets
415 // following it (DWARF v5 and later).
416 template <typename ListTableType>
417 static llvm::Expected<ListTableType>
ParseListTableHeader(const llvm::DWARFDataExtractor & data,uint64_t offset,DwarfFormat format)418 ParseListTableHeader(const llvm::DWARFDataExtractor &data, uint64_t offset,
419 DwarfFormat format) {
420 // We are expected to be called with Offset 0 or pointing just past the table
421 // header. Correct Offset in the latter case so that it points to the start
422 // of the header.
423 if (offset > 0) {
424 uint64_t HeaderSize = llvm::DWARFListTableHeader::getHeaderSize(format);
425 if (offset < HeaderSize)
426 return llvm::createStringError(errc::invalid_argument,
427 "did not detect a valid"
428 " list table with base = 0x%" PRIx64 "\n",
429 offset);
430 offset -= HeaderSize;
431 }
432 ListTableType Table;
433 if (llvm::Error E = Table.extractHeaderAndOffsets(data, &offset))
434 return std::move(E);
435 return Table;
436 }
437
SetLoclistsBase(dw_addr_t loclists_base)438 void DWARFUnit::SetLoclistsBase(dw_addr_t loclists_base) {
439 m_loclists_base = loclists_base;
440
441 uint64_t header_size = llvm::DWARFListTableHeader::getHeaderSize(DWARF32);
442 if (loclists_base < header_size)
443 return;
444
445 m_loclist_table_header.emplace(".debug_loclists", "locations");
446 uint64_t offset = loclists_base - header_size;
447 if (llvm::Error E = m_loclist_table_header->extract(
448 m_dwarf.GetDWARFContext().getOrLoadLocListsData().GetAsLLVM(),
449 &offset)) {
450 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
451 "Failed to extract location list table at offset 0x%" PRIx64 ": %s",
452 loclists_base, toString(std::move(E)).c_str());
453 }
454 }
455
456 std::unique_ptr<llvm::DWARFLocationTable>
GetLocationTable(const DataExtractor & data) const457 DWARFUnit::GetLocationTable(const DataExtractor &data) const {
458 llvm::DWARFDataExtractor llvm_data(
459 data.GetData(), data.GetByteOrder() == lldb::eByteOrderLittle,
460 data.GetAddressByteSize());
461
462 if (m_is_dwo || GetVersion() >= 5)
463 return std::make_unique<llvm::DWARFDebugLoclists>(llvm_data, GetVersion());
464 return std::make_unique<llvm::DWARFDebugLoc>(llvm_data);
465 }
466
GetLocationData() const467 DWARFDataExtractor DWARFUnit::GetLocationData() const {
468 DWARFContext &Ctx = GetSymbolFileDWARF().GetDWARFContext();
469 const DWARFDataExtractor &data =
470 GetVersion() >= 5 ? Ctx.getOrLoadLocListsData() : Ctx.getOrLoadLocData();
471 if (const llvm::DWARFUnitIndex::Entry *entry = m_header.GetIndexEntry()) {
472 if (const auto *contribution = entry->getContribution(llvm::DW_SECT_EXT_LOC))
473 return DWARFDataExtractor(data, contribution->Offset,
474 contribution->Length);
475 return DWARFDataExtractor();
476 }
477 return data;
478 }
479
SetRangesBase(dw_addr_t ranges_base)480 void DWARFUnit::SetRangesBase(dw_addr_t ranges_base) {
481 m_ranges_base = ranges_base;
482
483 if (GetVersion() < 5)
484 return;
485
486 if (auto table_or_error = ParseListTableHeader<llvm::DWARFDebugRnglistTable>(
487 m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(),
488 ranges_base, DWARF32))
489 m_rnglist_table = std::move(table_or_error.get());
490 else
491 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
492 "Failed to extract range list table at offset 0x%" PRIx64 ": %s",
493 ranges_base, toString(table_or_error.takeError()).c_str());
494 }
495
SetStrOffsetsBase(dw_offset_t str_offsets_base)496 void DWARFUnit::SetStrOffsetsBase(dw_offset_t str_offsets_base) {
497 m_str_offsets_base = str_offsets_base;
498 }
499
500 // It may be called only with m_die_array_mutex held R/W.
ClearDIEsRWLocked()501 void DWARFUnit::ClearDIEsRWLocked() {
502 m_die_array.clear();
503 m_die_array.shrink_to_fit();
504
505 if (m_dwo)
506 m_dwo->ClearDIEsRWLocked();
507 }
508
GetByteOrder() const509 lldb::ByteOrder DWARFUnit::GetByteOrder() const {
510 return m_dwarf.GetObjectFile()->GetByteOrder();
511 }
512
SetBaseAddress(dw_addr_t base_addr)513 void DWARFUnit::SetBaseAddress(dw_addr_t base_addr) { m_base_addr = base_addr; }
514
515 // Compare function DWARFDebugAranges::Range structures
CompareDIEOffset(const DWARFDebugInfoEntry & die,const dw_offset_t die_offset)516 static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
517 const dw_offset_t die_offset) {
518 return die.GetOffset() < die_offset;
519 }
520
521 // GetDIE()
522 //
523 // Get the DIE (Debug Information Entry) with the specified offset by first
524 // checking if the DIE is contained within this compile unit and grabbing the
525 // DIE from this compile unit. Otherwise we grab the DIE from the DWARF file.
526 DWARFDIE
GetDIE(dw_offset_t die_offset)527 DWARFUnit::GetDIE(dw_offset_t die_offset) {
528 if (die_offset == DW_INVALID_OFFSET)
529 return DWARFDIE(); // Not found
530
531 if (!ContainsDIEOffset(die_offset)) {
532 GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
533 "GetDIE for DIE 0x%" PRIx32 " is outside of its CU 0x%" PRIx32,
534 die_offset, GetOffset());
535 return DWARFDIE(); // Not found
536 }
537
538 ExtractDIEsIfNeeded();
539 DWARFDebugInfoEntry::const_iterator end = m_die_array.cend();
540 DWARFDebugInfoEntry::const_iterator pos =
541 lower_bound(m_die_array.cbegin(), end, die_offset, CompareDIEOffset);
542
543 if (pos != end && die_offset == (*pos).GetOffset())
544 return DWARFDIE(this, &(*pos));
545 return DWARFDIE(); // Not found
546 }
547
GetNonSkeletonUnit()548 DWARFUnit &DWARFUnit::GetNonSkeletonUnit() {
549 ExtractUnitDIEIfNeeded();
550 if (m_dwo)
551 return *m_dwo;
552 return *this;
553 }
554
GetAddressByteSize(const DWARFUnit * cu)555 uint8_t DWARFUnit::GetAddressByteSize(const DWARFUnit *cu) {
556 if (cu)
557 return cu->GetAddressByteSize();
558 return DWARFUnit::GetDefaultAddressSize();
559 }
560
GetDefaultAddressSize()561 uint8_t DWARFUnit::GetDefaultAddressSize() { return 4; }
562
GetUserData() const563 void *DWARFUnit::GetUserData() const { return m_user_data; }
564
SetUserData(void * d)565 void DWARFUnit::SetUserData(void *d) { m_user_data = d; }
566
Supports_DW_AT_APPLE_objc_complete_type()567 bool DWARFUnit::Supports_DW_AT_APPLE_objc_complete_type() {
568 return GetProducer() != eProducerLLVMGCC;
569 }
570
DW_AT_decl_file_attributes_are_invalid()571 bool DWARFUnit::DW_AT_decl_file_attributes_are_invalid() {
572 // llvm-gcc makes completely invalid decl file attributes and won't ever be
573 // fixed, so we need to know to ignore these.
574 return GetProducer() == eProducerLLVMGCC;
575 }
576
Supports_unnamed_objc_bitfields()577 bool DWARFUnit::Supports_unnamed_objc_bitfields() {
578 if (GetProducer() == eProducerClang) {
579 const uint32_t major_version = GetProducerVersionMajor();
580 return major_version > 425 ||
581 (major_version == 425 && GetProducerVersionUpdate() >= 13);
582 }
583 return true; // Assume all other compilers didn't have incorrect ObjC bitfield
584 // info
585 }
586
ParseProducerInfo()587 void DWARFUnit::ParseProducerInfo() {
588 m_producer_version_major = UINT32_MAX;
589 m_producer_version_minor = UINT32_MAX;
590 m_producer_version_update = UINT32_MAX;
591
592 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
593 if (die) {
594
595 const char *producer_cstr =
596 die->GetAttributeValueAsString(this, DW_AT_producer, nullptr);
597 if (producer_cstr) {
598 RegularExpression llvm_gcc_regex(
599 llvm::StringRef("^4\\.[012]\\.[01] \\(Based on Apple "
600 "Inc\\. build [0-9]+\\) \\(LLVM build "
601 "[\\.0-9]+\\)$"));
602 if (llvm_gcc_regex.Execute(llvm::StringRef(producer_cstr))) {
603 m_producer = eProducerLLVMGCC;
604 } else if (strstr(producer_cstr, "clang")) {
605 static RegularExpression g_clang_version_regex(
606 llvm::StringRef("clang-([0-9]+)\\.([0-9]+)\\.([0-9]+)"));
607 llvm::SmallVector<llvm::StringRef, 4> matches;
608 if (g_clang_version_regex.Execute(llvm::StringRef(producer_cstr),
609 &matches)) {
610 m_producer_version_major =
611 StringConvert::ToUInt32(matches[1].str().c_str(), UINT32_MAX, 10);
612 m_producer_version_minor =
613 StringConvert::ToUInt32(matches[2].str().c_str(), UINT32_MAX, 10);
614 m_producer_version_update =
615 StringConvert::ToUInt32(matches[3].str().c_str(), UINT32_MAX, 10);
616 }
617 m_producer = eProducerClang;
618 } else if (strstr(producer_cstr, "GNU"))
619 m_producer = eProducerGCC;
620 }
621 }
622 if (m_producer == eProducerInvalid)
623 m_producer = eProcucerOther;
624 }
625
GetProducer()626 DWARFProducer DWARFUnit::GetProducer() {
627 if (m_producer == eProducerInvalid)
628 ParseProducerInfo();
629 return m_producer;
630 }
631
GetProducerVersionMajor()632 uint32_t DWARFUnit::GetProducerVersionMajor() {
633 if (m_producer_version_major == 0)
634 ParseProducerInfo();
635 return m_producer_version_major;
636 }
637
GetProducerVersionMinor()638 uint32_t DWARFUnit::GetProducerVersionMinor() {
639 if (m_producer_version_minor == 0)
640 ParseProducerInfo();
641 return m_producer_version_minor;
642 }
643
GetProducerVersionUpdate()644 uint32_t DWARFUnit::GetProducerVersionUpdate() {
645 if (m_producer_version_update == 0)
646 ParseProducerInfo();
647 return m_producer_version_update;
648 }
649
GetDWARFLanguageType()650 uint64_t DWARFUnit::GetDWARFLanguageType() {
651 if (m_language_type)
652 return *m_language_type;
653
654 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
655 if (!die)
656 m_language_type = 0;
657 else
658 m_language_type = die->GetAttributeValueAsUnsigned(this, DW_AT_language, 0);
659 return *m_language_type;
660 }
661
GetIsOptimized()662 bool DWARFUnit::GetIsOptimized() {
663 if (m_is_optimized == eLazyBoolCalculate) {
664 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
665 if (die) {
666 m_is_optimized = eLazyBoolNo;
667 if (die->GetAttributeValueAsUnsigned(this, DW_AT_APPLE_optimized, 0) ==
668 1) {
669 m_is_optimized = eLazyBoolYes;
670 }
671 }
672 }
673 return m_is_optimized == eLazyBoolYes;
674 }
675
GetPathStyle()676 FileSpec::Style DWARFUnit::GetPathStyle() {
677 if (!m_comp_dir)
678 ComputeCompDirAndGuessPathStyle();
679 return m_comp_dir->GetPathStyle();
680 }
681
GetCompilationDirectory()682 const FileSpec &DWARFUnit::GetCompilationDirectory() {
683 if (!m_comp_dir)
684 ComputeCompDirAndGuessPathStyle();
685 return *m_comp_dir;
686 }
687
GetAbsolutePath()688 const FileSpec &DWARFUnit::GetAbsolutePath() {
689 if (!m_file_spec)
690 ComputeAbsolutePath();
691 return *m_file_spec;
692 }
693
GetFile(size_t file_idx)694 FileSpec DWARFUnit::GetFile(size_t file_idx) {
695 return m_dwarf.GetFile(*this, file_idx);
696 }
697
698 // DWARF2/3 suggests the form hostname:pathname for compilation directory.
699 // Remove the host part if present.
700 static llvm::StringRef
removeHostnameFromPathname(llvm::StringRef path_from_dwarf)701 removeHostnameFromPathname(llvm::StringRef path_from_dwarf) {
702 if (!path_from_dwarf.contains(':'))
703 return path_from_dwarf;
704 llvm::StringRef host, path;
705 std::tie(host, path) = path_from_dwarf.split(':');
706
707 if (host.contains('/'))
708 return path_from_dwarf;
709
710 // check whether we have a windows path, and so the first character is a
711 // drive-letter not a hostname.
712 if (host.size() == 1 && llvm::isAlpha(host[0]) && path.startswith("\\"))
713 return path_from_dwarf;
714
715 return path;
716 }
717
ComputeCompDirAndGuessPathStyle()718 void DWARFUnit::ComputeCompDirAndGuessPathStyle() {
719 m_comp_dir = FileSpec();
720 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
721 if (!die)
722 return;
723
724 llvm::StringRef comp_dir = removeHostnameFromPathname(
725 die->GetAttributeValueAsString(this, DW_AT_comp_dir, nullptr));
726 if (!comp_dir.empty()) {
727 FileSpec::Style comp_dir_style =
728 FileSpec::GuessPathStyle(comp_dir).getValueOr(FileSpec::Style::native);
729 m_comp_dir = FileSpec(comp_dir, comp_dir_style);
730 } else {
731 // Try to detect the style based on the DW_AT_name attribute, but just store
732 // the detected style in the m_comp_dir field.
733 const char *name =
734 die->GetAttributeValueAsString(this, DW_AT_name, nullptr);
735 m_comp_dir = FileSpec(
736 "", FileSpec::GuessPathStyle(name).getValueOr(FileSpec::Style::native));
737 }
738 }
739
ComputeAbsolutePath()740 void DWARFUnit::ComputeAbsolutePath() {
741 m_file_spec = FileSpec();
742 const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
743 if (!die)
744 return;
745
746 m_file_spec =
747 FileSpec(die->GetAttributeValueAsString(this, DW_AT_name, nullptr),
748 GetPathStyle());
749
750 if (m_file_spec->IsRelative())
751 m_file_spec->MakeAbsolute(GetCompilationDirectory());
752 }
753
GetDwoSymbolFile()754 SymbolFileDWARFDwo *DWARFUnit::GetDwoSymbolFile() {
755 ExtractUnitDIEIfNeeded();
756 if (m_dwo)
757 return &llvm::cast<SymbolFileDWARFDwo>(m_dwo->GetSymbolFileDWARF());
758 return nullptr;
759 }
760
GetFunctionAranges()761 const DWARFDebugAranges &DWARFUnit::GetFunctionAranges() {
762 if (m_func_aranges_up == nullptr) {
763 m_func_aranges_up = std::make_unique<DWARFDebugAranges>();
764 const DWARFDebugInfoEntry *die = DIEPtr();
765 if (die)
766 die->BuildFunctionAddressRangeTable(this, m_func_aranges_up.get());
767
768 if (m_dwo) {
769 const DWARFDebugInfoEntry *dwo_die = m_dwo->DIEPtr();
770 if (dwo_die)
771 dwo_die->BuildFunctionAddressRangeTable(m_dwo.get(),
772 m_func_aranges_up.get());
773 }
774
775 const bool minimize = false;
776 m_func_aranges_up->Sort(minimize);
777 }
778 return *m_func_aranges_up;
779 }
780
781 llvm::Expected<DWARFUnitHeader>
extract(const DWARFDataExtractor & data,DIERef::Section section,lldb::offset_t * offset_ptr,const llvm::DWARFUnitIndex * index)782 DWARFUnitHeader::extract(const DWARFDataExtractor &data,
783 DIERef::Section section, lldb::offset_t *offset_ptr,
784 const llvm::DWARFUnitIndex *index) {
785 DWARFUnitHeader header;
786 header.m_offset = *offset_ptr;
787 if (index)
788 header.m_index_entry = index->getFromOffset(*offset_ptr);
789 header.m_length = data.GetDWARFInitialLength(offset_ptr);
790 header.m_version = data.GetU16(offset_ptr);
791 if (header.m_version == 5) {
792 header.m_unit_type = data.GetU8(offset_ptr);
793 header.m_addr_size = data.GetU8(offset_ptr);
794 header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
795 if (header.m_unit_type == llvm::dwarf::DW_UT_skeleton)
796 header.m_dwo_id = data.GetU64(offset_ptr);
797 } else {
798 header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
799 header.m_addr_size = data.GetU8(offset_ptr);
800 header.m_unit_type =
801 section == DIERef::Section::DebugTypes ? DW_UT_type : DW_UT_compile;
802 }
803
804 if (header.m_index_entry) {
805 if (header.m_abbr_offset) {
806 return llvm::createStringError(
807 llvm::inconvertibleErrorCode(),
808 "Package unit with a non-zero abbreviation offset");
809 }
810 auto *unit_contrib = header.m_index_entry->getContribution();
811 if (!unit_contrib || unit_contrib->Length != header.m_length + 4) {
812 return llvm::createStringError(llvm::inconvertibleErrorCode(),
813 "Inconsistent DWARF package unit index");
814 }
815 auto *abbr_entry =
816 header.m_index_entry->getContribution(llvm::DW_SECT_ABBREV);
817 if (!abbr_entry) {
818 return llvm::createStringError(
819 llvm::inconvertibleErrorCode(),
820 "DWARF package index missing abbreviation column");
821 }
822 header.m_abbr_offset = abbr_entry->Offset;
823 }
824 if (header.IsTypeUnit()) {
825 header.m_type_hash = data.GetU64(offset_ptr);
826 header.m_type_offset = data.GetDWARFOffset(offset_ptr);
827 }
828
829 bool length_OK = data.ValidOffset(header.GetNextUnitOffset() - 1);
830 bool version_OK = SymbolFileDWARF::SupportedVersion(header.m_version);
831 bool addr_size_OK = (header.m_addr_size == 4) || (header.m_addr_size == 8);
832 bool type_offset_OK =
833 !header.IsTypeUnit() || (header.m_type_offset <= header.GetLength());
834
835 if (!length_OK)
836 return llvm::make_error<llvm::object::GenericBinaryError>(
837 "Invalid unit length");
838 if (!version_OK)
839 return llvm::make_error<llvm::object::GenericBinaryError>(
840 "Unsupported unit version");
841 if (!addr_size_OK)
842 return llvm::make_error<llvm::object::GenericBinaryError>(
843 "Invalid unit address size");
844 if (!type_offset_OK)
845 return llvm::make_error<llvm::object::GenericBinaryError>(
846 "Type offset out of range");
847
848 return header;
849 }
850
851 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)852 DWARFUnit::extract(SymbolFileDWARF &dwarf, user_id_t uid,
853 const DWARFDataExtractor &debug_info,
854 DIERef::Section section, lldb::offset_t *offset_ptr,
855 const llvm::DWARFUnitIndex *index) {
856 assert(debug_info.ValidOffset(*offset_ptr));
857
858 auto expected_header =
859 DWARFUnitHeader::extract(debug_info, section, offset_ptr, index);
860 if (!expected_header)
861 return expected_header.takeError();
862
863 const DWARFDebugAbbrev *abbr = dwarf.DebugAbbrev();
864 if (!abbr)
865 return llvm::make_error<llvm::object::GenericBinaryError>(
866 "No debug_abbrev data");
867
868 bool abbr_offset_OK =
869 dwarf.GetDWARFContext().getOrLoadAbbrevData().ValidOffset(
870 expected_header->GetAbbrOffset());
871 if (!abbr_offset_OK)
872 return llvm::make_error<llvm::object::GenericBinaryError>(
873 "Abbreviation offset for unit is not valid");
874
875 const DWARFAbbreviationDeclarationSet *abbrevs =
876 abbr->GetAbbreviationDeclarationSet(expected_header->GetAbbrOffset());
877 if (!abbrevs)
878 return llvm::make_error<llvm::object::GenericBinaryError>(
879 "No abbrev exists at the specified offset.");
880
881 bool is_dwo = dwarf.GetDWARFContext().isDwo();
882 if (expected_header->IsTypeUnit())
883 return DWARFUnitSP(new DWARFTypeUnit(dwarf, uid, *expected_header, *abbrevs,
884 section, is_dwo));
885 return DWARFUnitSP(new DWARFCompileUnit(dwarf, uid, *expected_header,
886 *abbrevs, section, is_dwo));
887 }
888
GetData() const889 const lldb_private::DWARFDataExtractor &DWARFUnit::GetData() const {
890 return m_section == DIERef::Section::DebugTypes
891 ? m_dwarf.GetDWARFContext().getOrLoadDebugTypesData()
892 : m_dwarf.GetDWARFContext().getOrLoadDebugInfoData();
893 }
894
GetHeaderByteSize() const895 uint32_t DWARFUnit::GetHeaderByteSize() const {
896 switch (m_header.GetUnitType()) {
897 case llvm::dwarf::DW_UT_compile:
898 case llvm::dwarf::DW_UT_partial:
899 return GetVersion() < 5 ? 11 : 12;
900 case llvm::dwarf::DW_UT_skeleton:
901 case llvm::dwarf::DW_UT_split_compile:
902 return 20;
903 case llvm::dwarf::DW_UT_type:
904 case llvm::dwarf::DW_UT_split_type:
905 return GetVersion() < 5 ? 23 : 24;
906 }
907 llvm_unreachable("invalid UnitType.");
908 }
909
910 llvm::Optional<uint64_t>
GetStringOffsetSectionItem(uint32_t index) const911 DWARFUnit::GetStringOffsetSectionItem(uint32_t index) const {
912 offset_t offset = GetStrOffsetsBase() + index * 4;
913 return m_dwarf.GetDWARFContext().getOrLoadStrOffsetsData().GetU32(&offset);
914 }
915
916 llvm::Expected<DWARFRangeList>
FindRnglistFromOffset(dw_offset_t offset)917 DWARFUnit::FindRnglistFromOffset(dw_offset_t offset) {
918 if (GetVersion() <= 4) {
919 const DWARFDebugRanges *debug_ranges = m_dwarf.GetDebugRanges();
920 if (!debug_ranges)
921 return llvm::make_error<llvm::object::GenericBinaryError>(
922 "No debug_ranges section");
923 DWARFRangeList ranges;
924 debug_ranges->FindRanges(this, offset, ranges);
925 return ranges;
926 }
927
928 if (!m_rnglist_table)
929 return llvm::createStringError(errc::invalid_argument,
930 "missing or invalid range list table");
931
932 auto range_list_or_error = m_rnglist_table->findList(
933 m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(), offset);
934 if (!range_list_or_error)
935 return range_list_or_error.takeError();
936
937 llvm::Expected<llvm::DWARFAddressRangesVector> llvm_ranges =
938 range_list_or_error->getAbsoluteRanges(
939 llvm::object::SectionedAddress{GetBaseAddress()},
940 GetAddressByteSize(), [&](uint32_t index) {
941 uint32_t index_size = GetAddressByteSize();
942 dw_offset_t addr_base = GetAddrBase();
943 lldb::offset_t offset = addr_base + index * index_size;
944 return llvm::object::SectionedAddress{
945 m_dwarf.GetDWARFContext().getOrLoadAddrData().GetMaxU64(
946 &offset, index_size)};
947 });
948 if (!llvm_ranges)
949 return llvm_ranges.takeError();
950
951 DWARFRangeList ranges;
952 for (const llvm::DWARFAddressRange &llvm_range : *llvm_ranges) {
953 ranges.Append(DWARFRangeList::Entry(llvm_range.LowPC,
954 llvm_range.HighPC - llvm_range.LowPC));
955 }
956 return ranges;
957 }
958
959 llvm::Expected<DWARFRangeList>
FindRnglistFromIndex(uint32_t index)960 DWARFUnit::FindRnglistFromIndex(uint32_t index) {
961 if (llvm::Optional<uint64_t> offset = GetRnglistOffset(index))
962 return FindRnglistFromOffset(*offset);
963 if (m_rnglist_table)
964 return llvm::createStringError(errc::invalid_argument,
965 "invalid range list table index %d", index);
966
967 return llvm::createStringError(errc::invalid_argument,
968 "missing or invalid range list table");
969 }
970