1//===-- DWARFUnit.cpp -------------------------------------------*- C++ -*-===//
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
26using namespace lldb;
27using namespace lldb_private;
28using namespace std;
29
30extern int g_verbose;
31
32DWARFUnit::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
39DWARFUnit::~DWARFUnit() = default;
40
41// Parses first DIE of a compile unit.
42void 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.
72void 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.
92DWARFUnit::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
112DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(DWARFUnit &cu) : m_cu(&cu) {
113  m_cu->m_die_array_scoped_mutex.lock_shared();
114}
115
116DWARFUnit::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
130DWARFUnit::ScopedExtractDIEs::ScopedExtractDIEs(ScopedExtractDIEs &&rhs)
131    : m_cu(rhs.m_cu), m_clear_dies(rhs.m_clear_dies) {
132  rhs.m_cu = nullptr;
133}
134
135DWARFUnit::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.
145void 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_symbol_file) {
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_symbol_file) {
253    DWARFUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
254    dwo_cu->ExtractDIEsIfNeeded();
255  }
256}
257
258// This is used when a split dwarf is enabled.
259// A skeleton compilation unit may contain the DW_AT_str_offsets_base attribute
260// that points to the first string offset of the CU contribution to the
261// .debug_str_offsets. At the same time, the corresponding split debug unit also
262// may use DW_FORM_strx* forms pointing to its own .debug_str_offsets.dwo and
263// for that case, we should find the offset (skip the section header).
264static void SetDwoStrOffsetsBase(DWARFUnit *dwo_cu) {
265  lldb::offset_t baseOffset = 0;
266
267  const DWARFDataExtractor &strOffsets =
268      dwo_cu->GetSymbolFileDWARF().GetDWARFContext().getOrLoadStrOffsetsData();
269  uint64_t length = strOffsets.GetU32(&baseOffset);
270  if (length == 0xffffffff)
271    length = strOffsets.GetU64(&baseOffset);
272
273  // Check version.
274  if (strOffsets.GetU16(&baseOffset) < 5)
275    return;
276
277  // Skip padding.
278  baseOffset += 2;
279
280  dwo_cu->SetStrOffsetsBase(baseOffset);
281}
282
283// m_die_array_mutex must be already held as read/write.
284void DWARFUnit::AddUnitDIE(const DWARFDebugInfoEntry &cu_die) {
285  llvm::Optional<uint64_t> addr_base, gnu_addr_base, ranges_base,
286      gnu_ranges_base;
287
288  DWARFAttributes attributes;
289  size_t num_attributes = cu_die.GetAttributes(this, attributes);
290
291  // Extract DW_AT_addr_base first, as other attributes may need it.
292  for (size_t i = 0; i < num_attributes; ++i) {
293    if (attributes.AttributeAtIndex(i) != DW_AT_addr_base)
294      continue;
295    DWARFFormValue form_value;
296    if (attributes.ExtractFormValueAtIndex(i, form_value)) {
297      addr_base = form_value.Unsigned();
298      SetAddrBase(*addr_base);
299      break;
300    }
301  }
302
303  for (size_t i = 0; i < num_attributes; ++i) {
304    dw_attr_t attr = attributes.AttributeAtIndex(i);
305    DWARFFormValue form_value;
306    if (!attributes.ExtractFormValueAtIndex(i, form_value))
307      continue;
308    switch (attr) {
309    case DW_AT_loclists_base:
310      SetLoclistsBase(form_value.Unsigned());
311      break;
312    case DW_AT_rnglists_base:
313      ranges_base = form_value.Unsigned();
314      SetRangesBase(*ranges_base);
315      break;
316    case DW_AT_str_offsets_base:
317      SetStrOffsetsBase(form_value.Unsigned());
318      break;
319    case DW_AT_low_pc:
320      SetBaseAddress(form_value.Address());
321      break;
322    case DW_AT_entry_pc:
323      // If the value was already set by DW_AT_low_pc, don't update it.
324      if (m_base_addr == LLDB_INVALID_ADDRESS)
325        SetBaseAddress(form_value.Address());
326      break;
327    case DW_AT_stmt_list:
328      m_line_table_offset = form_value.Unsigned();
329      break;
330    case DW_AT_GNU_addr_base:
331      gnu_addr_base = form_value.Unsigned();
332      break;
333    case DW_AT_GNU_ranges_base:
334      gnu_ranges_base = form_value.Unsigned();
335      break;
336    }
337  }
338
339  if (m_is_dwo)
340    return;
341
342  std::unique_ptr<SymbolFileDWARFDwo> dwo_symbol_file =
343      m_dwarf.GetDwoSymbolFileForCompileUnit(*this, cu_die);
344  if (!dwo_symbol_file)
345    return;
346
347  DWARFUnit *dwo_cu = dwo_symbol_file->GetCompileUnit();
348  if (!dwo_cu)
349    return; // Can't fetch the compile unit from the dwo file.
350
351  DWARFBaseDIE dwo_cu_die = dwo_cu->GetUnitDIEOnly();
352  if (!dwo_cu_die.IsValid())
353    return; // Can't fetch the compile unit DIE from the dwo file.
354
355  uint64_t main_dwo_id =
356      cu_die.GetAttributeValueAsUnsigned(this, DW_AT_GNU_dwo_id, 0);
357  uint64_t sub_dwo_id =
358      dwo_cu_die.GetAttributeValueAsUnsigned(DW_AT_GNU_dwo_id, 0);
359  if (main_dwo_id != sub_dwo_id)
360    return; // The 2 dwo ID isn't match. Don't use the dwo file as it belongs to
361  // a differectn compilation.
362
363  m_dwo_symbol_file = std::move(dwo_symbol_file);
364
365  // Here for DWO CU we want to use the address base set in the skeleton unit
366  // (DW_AT_addr_base) if it is available and use the DW_AT_GNU_addr_base
367  // otherwise. We do that because pre-DWARF v5 could use the DW_AT_GNU_*
368  // attributes which were applicable to the DWO units. The corresponding
369  // DW_AT_* attributes standardized in DWARF v5 are also applicable to the main
370  // unit in contrast.
371  if (addr_base)
372    dwo_cu->SetAddrBase(*addr_base);
373  else if (gnu_addr_base)
374    dwo_cu->SetAddrBase(*gnu_addr_base);
375
376  if (GetVersion() <= 4 && gnu_ranges_base)
377    dwo_cu->SetRangesBase(*gnu_ranges_base);
378  else if (m_dwo_symbol_file->GetDWARFContext()
379               .getOrLoadRngListsData()
380               .GetByteSize() > 0)
381    dwo_cu->SetRangesBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
382
383  if (GetVersion() >= 5 && m_dwo_symbol_file->GetDWARFContext()
384                                   .getOrLoadLocListsData()
385                                   .GetByteSize() > 0)
386    dwo_cu->SetLoclistsBase(llvm::DWARFListTableHeader::getHeaderSize(DWARF32));
387  dwo_cu->SetBaseAddress(GetBaseAddress());
388
389  for (size_t i = 0; i < m_dwo_symbol_file->DebugInfo()->GetNumUnits(); ++i) {
390    DWARFUnit *unit = m_dwo_symbol_file->DebugInfo()->GetUnitAtIndex(i);
391    SetDwoStrOffsetsBase(unit);
392  }
393}
394
395DWARFDIE 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
406size_t DWARFUnit::GetDebugInfoSize() const {
407  return GetLengthByteSize() + GetLength() - GetHeaderByteSize();
408}
409
410const DWARFAbbreviationDeclarationSet *DWARFUnit::GetAbbreviations() const {
411  return m_abbrevs;
412}
413
414dw_offset_t DWARFUnit::GetAbbrevOffset() const {
415  return m_abbrevs ? m_abbrevs->GetOffset() : DW_INVALID_OFFSET;
416}
417
418dw_offset_t DWARFUnit::GetLineTableOffset() {
419  ExtractUnitDIEIfNeeded();
420  return m_line_table_offset;
421}
422
423void 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).
427template <typename ListTableType>
428static llvm::Expected<ListTableType>
429ParseListTableHeader(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
449void 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
467std::unique_ptr<llvm::DWARFLocationTable>
468DWARFUnit::GetLocationTable(const DataExtractor &data) const {
469  llvm::DWARFDataExtractor llvm_data(
470      toStringRef(data.GetData()),
471      data.GetByteOrder() == lldb::eByteOrderLittle, 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
478const DWARFDataExtractor &DWARFUnit::GetLocationData() const {
479  DWARFContext &Ctx = GetSymbolFileDWARF().GetDWARFContext();
480  return GetVersion() >= 5 ? Ctx.getOrLoadLocListsData()
481                           : Ctx.getOrLoadLocData();
482}
483
484void DWARFUnit::SetRangesBase(dw_addr_t ranges_base) {
485  m_ranges_base = ranges_base;
486
487  if (GetVersion() < 5)
488    return;
489
490  if (auto table_or_error = ParseListTableHeader<llvm::DWARFDebugRnglistTable>(
491          m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(),
492          ranges_base, DWARF32))
493    m_rnglist_table = std::move(table_or_error.get());
494  else
495    GetSymbolFileDWARF().GetObjectFile()->GetModule()->ReportError(
496        "Failed to extract range list table at offset 0x%" PRIx64 ": %s",
497        ranges_base, toString(table_or_error.takeError()).c_str());
498}
499
500void DWARFUnit::SetStrOffsetsBase(dw_offset_t str_offsets_base) {
501  m_str_offsets_base = str_offsets_base;
502}
503
504// It may be called only with m_die_array_mutex held R/W.
505void DWARFUnit::ClearDIEsRWLocked() {
506  m_die_array.clear();
507  m_die_array.shrink_to_fit();
508
509  if (m_dwo_symbol_file)
510    m_dwo_symbol_file->GetCompileUnit()->ClearDIEsRWLocked();
511}
512
513lldb::ByteOrder DWARFUnit::GetByteOrder() const {
514  return m_dwarf.GetObjectFile()->GetByteOrder();
515}
516
517llvm::Expected<TypeSystem &> DWARFUnit::GetTypeSystem() {
518  return m_dwarf.GetTypeSystemForLanguage(GetLanguageType());
519}
520
521void DWARFUnit::SetBaseAddress(dw_addr_t base_addr) { m_base_addr = base_addr; }
522
523// Compare function DWARFDebugAranges::Range structures
524static bool CompareDIEOffset(const DWARFDebugInfoEntry &die,
525                             const dw_offset_t die_offset) {
526  return die.GetOffset() < die_offset;
527}
528
529// GetDIE()
530//
531// Get the DIE (Debug Information Entry) with the specified offset by first
532// checking if the DIE is contained within this compile unit and grabbing the
533// DIE from this compile unit. Otherwise we grab the DIE from the DWARF file.
534DWARFDIE
535DWARFUnit::GetDIE(dw_offset_t die_offset) {
536  if (die_offset != DW_INVALID_OFFSET) {
537    if (GetDwoSymbolFile())
538      return GetDwoSymbolFile()->GetCompileUnit()->GetDIE(die_offset);
539
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
557DWARFUnit &DWARFUnit::GetNonSkeletonUnit() {
558  if (SymbolFileDWARFDwo *dwo = GetDwoSymbolFile())
559    return *dwo->GetCompileUnit();
560  return *this;
561}
562
563uint8_t DWARFUnit::GetAddressByteSize(const DWARFUnit *cu) {
564  if (cu)
565    return cu->GetAddressByteSize();
566  return DWARFUnit::GetDefaultAddressSize();
567}
568
569uint8_t DWARFUnit::GetDefaultAddressSize() { return 4; }
570
571void *DWARFUnit::GetUserData() const { return m_user_data; }
572
573void DWARFUnit::SetUserData(void *d) {
574  m_user_data = d;
575  if (m_dwo_symbol_file)
576    m_dwo_symbol_file->GetCompileUnit()->SetUserData(d);
577}
578
579bool DWARFUnit::Supports_DW_AT_APPLE_objc_complete_type() {
580  return GetProducer() != eProducerLLVMGCC;
581}
582
583bool DWARFUnit::DW_AT_decl_file_attributes_are_invalid() {
584  // llvm-gcc makes completely invalid decl file attributes and won't ever be
585  // fixed, so we need to know to ignore these.
586  return GetProducer() == eProducerLLVMGCC;
587}
588
589bool DWARFUnit::Supports_unnamed_objc_bitfields() {
590  if (GetProducer() == eProducerClang) {
591    const uint32_t major_version = GetProducerVersionMajor();
592    return major_version > 425 ||
593           (major_version == 425 && GetProducerVersionUpdate() >= 13);
594  }
595  return true; // Assume all other compilers didn't have incorrect ObjC bitfield
596               // info
597}
598
599void DWARFUnit::ParseProducerInfo() {
600  m_producer_version_major = UINT32_MAX;
601  m_producer_version_minor = UINT32_MAX;
602  m_producer_version_update = UINT32_MAX;
603
604  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
605  if (die) {
606
607    const char *producer_cstr =
608        die->GetAttributeValueAsString(this, DW_AT_producer, nullptr);
609    if (producer_cstr) {
610      RegularExpression llvm_gcc_regex(
611          llvm::StringRef("^4\\.[012]\\.[01] \\(Based on Apple "
612                          "Inc\\. build [0-9]+\\) \\(LLVM build "
613                          "[\\.0-9]+\\)$"));
614      if (llvm_gcc_regex.Execute(llvm::StringRef(producer_cstr))) {
615        m_producer = eProducerLLVMGCC;
616      } else if (strstr(producer_cstr, "clang")) {
617        static RegularExpression g_clang_version_regex(
618            llvm::StringRef("clang-([0-9]+)\\.([0-9]+)\\.([0-9]+)"));
619        llvm::SmallVector<llvm::StringRef, 4> matches;
620        if (g_clang_version_regex.Execute(llvm::StringRef(producer_cstr),
621                                          &matches)) {
622          m_producer_version_major =
623              StringConvert::ToUInt32(matches[1].str().c_str(), UINT32_MAX, 10);
624          m_producer_version_minor =
625              StringConvert::ToUInt32(matches[2].str().c_str(), UINT32_MAX, 10);
626          m_producer_version_update =
627              StringConvert::ToUInt32(matches[3].str().c_str(), UINT32_MAX, 10);
628        }
629        m_producer = eProducerClang;
630      } else if (strstr(producer_cstr, "GNU"))
631        m_producer = eProducerGCC;
632    }
633  }
634  if (m_producer == eProducerInvalid)
635    m_producer = eProcucerOther;
636}
637
638DWARFProducer DWARFUnit::GetProducer() {
639  if (m_producer == eProducerInvalid)
640    ParseProducerInfo();
641  return m_producer;
642}
643
644uint32_t DWARFUnit::GetProducerVersionMajor() {
645  if (m_producer_version_major == 0)
646    ParseProducerInfo();
647  return m_producer_version_major;
648}
649
650uint32_t DWARFUnit::GetProducerVersionMinor() {
651  if (m_producer_version_minor == 0)
652    ParseProducerInfo();
653  return m_producer_version_minor;
654}
655
656uint32_t DWARFUnit::GetProducerVersionUpdate() {
657  if (m_producer_version_update == 0)
658    ParseProducerInfo();
659  return m_producer_version_update;
660}
661LanguageType DWARFUnit::LanguageTypeFromDWARF(uint64_t val) {
662  // Note: user languages between lo_user and hi_user must be handled
663  // explicitly here.
664  switch (val) {
665  case DW_LANG_Mips_Assembler:
666    return eLanguageTypeMipsAssembler;
667  case DW_LANG_GOOGLE_RenderScript:
668    return eLanguageTypeExtRenderScript;
669  default:
670    return static_cast<LanguageType>(val);
671  }
672}
673
674LanguageType DWARFUnit::GetLanguageType() {
675  if (m_language_type != eLanguageTypeUnknown)
676    return m_language_type;
677
678  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
679  if (die)
680    m_language_type = LanguageTypeFromDWARF(
681        die->GetAttributeValueAsUnsigned(this, DW_AT_language, 0));
682  return m_language_type;
683}
684
685bool DWARFUnit::GetIsOptimized() {
686  if (m_is_optimized == eLazyBoolCalculate) {
687    const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
688    if (die) {
689      m_is_optimized = eLazyBoolNo;
690      if (die->GetAttributeValueAsUnsigned(this, DW_AT_APPLE_optimized, 0) ==
691          1) {
692        m_is_optimized = eLazyBoolYes;
693      }
694    }
695  }
696  return m_is_optimized == eLazyBoolYes;
697}
698
699FileSpec::Style DWARFUnit::GetPathStyle() {
700  if (!m_comp_dir)
701    ComputeCompDirAndGuessPathStyle();
702  return m_comp_dir->GetPathStyle();
703}
704
705const FileSpec &DWARFUnit::GetCompilationDirectory() {
706  if (!m_comp_dir)
707    ComputeCompDirAndGuessPathStyle();
708  return *m_comp_dir;
709}
710
711const FileSpec &DWARFUnit::GetAbsolutePath() {
712  if (!m_file_spec)
713    ComputeAbsolutePath();
714  return *m_file_spec;
715}
716
717FileSpec DWARFUnit::GetFile(size_t file_idx) {
718  return m_dwarf.GetFile(*this, file_idx);
719}
720
721// DWARF2/3 suggests the form hostname:pathname for compilation directory.
722// Remove the host part if present.
723static llvm::StringRef
724removeHostnameFromPathname(llvm::StringRef path_from_dwarf) {
725  if (!path_from_dwarf.contains(':'))
726    return path_from_dwarf;
727  llvm::StringRef host, path;
728  std::tie(host, path) = path_from_dwarf.split(':');
729
730  if (host.contains('/'))
731    return path_from_dwarf;
732
733  // check whether we have a windows path, and so the first character is a
734  // drive-letter not a hostname.
735  if (host.size() == 1 && llvm::isAlpha(host[0]) && path.startswith("\\"))
736    return path_from_dwarf;
737
738  return path;
739}
740
741static FileSpec resolveCompDir(const FileSpec &path) {
742  bool is_symlink = SymbolFileDWARF::GetSymlinkPaths().FindFileIndex(
743                        0, path, /*full*/ true) != UINT32_MAX;
744
745  if (!is_symlink)
746    return path;
747
748  namespace fs = llvm::sys::fs;
749  if (fs::get_file_type(path.GetPath(), false) != fs::file_type::symlink_file)
750    return path;
751
752  FileSpec resolved_symlink;
753  const auto error = FileSystem::Instance().Readlink(path, resolved_symlink);
754  if (error.Success())
755    return resolved_symlink;
756
757  return path;
758}
759
760void DWARFUnit::ComputeCompDirAndGuessPathStyle() {
761  m_comp_dir = FileSpec();
762  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
763  if (!die)
764    return;
765
766  llvm::StringRef comp_dir = removeHostnameFromPathname(
767      die->GetAttributeValueAsString(this, DW_AT_comp_dir, nullptr));
768  if (!comp_dir.empty()) {
769    FileSpec::Style comp_dir_style =
770        FileSpec::GuessPathStyle(comp_dir).getValueOr(FileSpec::Style::native);
771    m_comp_dir = resolveCompDir(FileSpec(comp_dir, comp_dir_style));
772  } else {
773    // Try to detect the style based on the DW_AT_name attribute, but just store
774    // the detected style in the m_comp_dir field.
775    const char *name =
776        die->GetAttributeValueAsString(this, DW_AT_name, nullptr);
777    m_comp_dir = FileSpec(
778        "", FileSpec::GuessPathStyle(name).getValueOr(FileSpec::Style::native));
779  }
780}
781
782void DWARFUnit::ComputeAbsolutePath() {
783  m_file_spec = FileSpec();
784  const DWARFDebugInfoEntry *die = GetUnitDIEPtrOnly();
785  if (!die)
786    return;
787
788  m_file_spec =
789      FileSpec(die->GetAttributeValueAsString(this, DW_AT_name, nullptr),
790               GetPathStyle());
791
792  if (m_file_spec->IsRelative())
793    m_file_spec->MakeAbsolute(GetCompilationDirectory());
794}
795
796SymbolFileDWARFDwo *DWARFUnit::GetDwoSymbolFile() {
797  ExtractUnitDIEIfNeeded();
798  return m_dwo_symbol_file.get();
799}
800
801const DWARFDebugAranges &DWARFUnit::GetFunctionAranges() {
802  if (m_func_aranges_up == nullptr) {
803    m_func_aranges_up.reset(new DWARFDebugAranges());
804    const DWARFDebugInfoEntry *die = DIEPtr();
805    if (die)
806      die->BuildFunctionAddressRangeTable(this, m_func_aranges_up.get());
807
808    if (m_dwo_symbol_file) {
809      DWARFUnit *dwo_cu = m_dwo_symbol_file->GetCompileUnit();
810      const DWARFDebugInfoEntry *dwo_die = dwo_cu->DIEPtr();
811      if (dwo_die)
812        dwo_die->BuildFunctionAddressRangeTable(dwo_cu,
813                                                m_func_aranges_up.get());
814    }
815
816    const bool minimize = false;
817    m_func_aranges_up->Sort(minimize);
818  }
819  return *m_func_aranges_up;
820}
821
822llvm::Expected<DWARFUnitHeader>
823DWARFUnitHeader::extract(const DWARFDataExtractor &data, DIERef::Section section,
824                         lldb::offset_t *offset_ptr) {
825  DWARFUnitHeader header;
826  header.m_offset = *offset_ptr;
827  header.m_length = data.GetDWARFInitialLength(offset_ptr);
828  header.m_version = data.GetU16(offset_ptr);
829  if (header.m_version == 5) {
830    header.m_unit_type = data.GetU8(offset_ptr);
831    header.m_addr_size = data.GetU8(offset_ptr);
832    header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
833    if (header.m_unit_type == llvm::dwarf::DW_UT_skeleton)
834      header.m_dwo_id = data.GetU64(offset_ptr);
835  } else {
836    header.m_abbr_offset = data.GetDWARFOffset(offset_ptr);
837    header.m_addr_size = data.GetU8(offset_ptr);
838    header.m_unit_type =
839        section == DIERef::Section::DebugTypes ? DW_UT_type : DW_UT_compile;
840  }
841
842  if (header.IsTypeUnit()) {
843    header.m_type_hash = data.GetU64(offset_ptr);
844    header.m_type_offset = data.GetDWARFOffset(offset_ptr);
845  }
846
847  bool length_OK = data.ValidOffset(header.GetNextUnitOffset() - 1);
848  bool version_OK = SymbolFileDWARF::SupportedVersion(header.m_version);
849  bool addr_size_OK = (header.m_addr_size == 4) || (header.m_addr_size == 8);
850  bool type_offset_OK =
851      !header.IsTypeUnit() || (header.m_type_offset <= header.GetLength());
852
853  if (!length_OK)
854    return llvm::make_error<llvm::object::GenericBinaryError>(
855        "Invalid unit length");
856  if (!version_OK)
857    return llvm::make_error<llvm::object::GenericBinaryError>(
858        "Unsupported unit version");
859  if (!addr_size_OK)
860    return llvm::make_error<llvm::object::GenericBinaryError>(
861        "Invalid unit address size");
862  if (!type_offset_OK)
863    return llvm::make_error<llvm::object::GenericBinaryError>(
864        "Type offset out of range");
865
866  return header;
867}
868
869llvm::Expected<DWARFUnitSP>
870DWARFUnit::extract(SymbolFileDWARF &dwarf, user_id_t uid,
871                   const DWARFDataExtractor &debug_info,
872                   DIERef::Section section, lldb::offset_t *offset_ptr) {
873  assert(debug_info.ValidOffset(*offset_ptr));
874
875  auto expected_header =
876      DWARFUnitHeader::extract(debug_info, section, offset_ptr);
877  if (!expected_header)
878    return expected_header.takeError();
879
880  const DWARFDebugAbbrev *abbr = dwarf.DebugAbbrev();
881  if (!abbr)
882    return llvm::make_error<llvm::object::GenericBinaryError>(
883        "No debug_abbrev data");
884
885  bool abbr_offset_OK =
886      dwarf.GetDWARFContext().getOrLoadAbbrevData().ValidOffset(
887          expected_header->GetAbbrOffset());
888  if (!abbr_offset_OK)
889    return llvm::make_error<llvm::object::GenericBinaryError>(
890        "Abbreviation offset for unit is not valid");
891
892  const DWARFAbbreviationDeclarationSet *abbrevs =
893      abbr->GetAbbreviationDeclarationSet(expected_header->GetAbbrOffset());
894  if (!abbrevs)
895    return llvm::make_error<llvm::object::GenericBinaryError>(
896        "No abbrev exists at the specified offset.");
897
898  bool is_dwo = dwarf.GetDWARFContext().isDwo();
899  if (expected_header->IsTypeUnit())
900    return DWARFUnitSP(new DWARFTypeUnit(dwarf, uid, *expected_header, *abbrevs,
901                                         section, is_dwo));
902  return DWARFUnitSP(new DWARFCompileUnit(dwarf, uid, *expected_header,
903                                          *abbrevs, section, is_dwo));
904}
905
906const lldb_private::DWARFDataExtractor &DWARFUnit::GetData() const {
907  return m_section == DIERef::Section::DebugTypes
908             ? m_dwarf.GetDWARFContext().getOrLoadDebugTypesData()
909             : m_dwarf.GetDWARFContext().getOrLoadDebugInfoData();
910}
911
912uint32_t DWARFUnit::GetHeaderByteSize() const {
913  switch (m_header.GetUnitType()) {
914  case llvm::dwarf::DW_UT_compile:
915  case llvm::dwarf::DW_UT_partial:
916    return GetVersion() < 5 ? 11 : 12;
917  case llvm::dwarf::DW_UT_skeleton:
918  case llvm::dwarf::DW_UT_split_compile:
919    return 20;
920  case llvm::dwarf::DW_UT_type:
921  case llvm::dwarf::DW_UT_split_type:
922    return GetVersion() < 5 ? 23 : 24;
923  }
924  llvm_unreachable("invalid UnitType.");
925}
926
927llvm::Expected<DWARFRangeList>
928DWARFUnit::FindRnglistFromOffset(dw_offset_t offset) {
929  if (GetVersion() <= 4) {
930    const DWARFDebugRanges *debug_ranges = m_dwarf.GetDebugRanges();
931    if (!debug_ranges)
932      return llvm::make_error<llvm::object::GenericBinaryError>(
933          "No debug_ranges section");
934    DWARFRangeList ranges;
935    debug_ranges->FindRanges(this, offset, ranges);
936    return ranges;
937  }
938
939  if (!m_rnglist_table)
940    return llvm::createStringError(errc::invalid_argument,
941                                   "missing or invalid range list table");
942
943  auto range_list_or_error = m_rnglist_table->findList(
944      m_dwarf.GetDWARFContext().getOrLoadRngListsData().GetAsLLVM(), offset);
945  if (!range_list_or_error)
946    return range_list_or_error.takeError();
947
948  llvm::Expected<llvm::DWARFAddressRangesVector> llvm_ranges =
949      range_list_or_error->getAbsoluteRanges(
950          llvm::object::SectionedAddress{GetBaseAddress()},
951          [&](uint32_t index) {
952            uint32_t index_size = GetAddressByteSize();
953            dw_offset_t addr_base = GetAddrBase();
954            lldb::offset_t offset = addr_base + index * index_size;
955            return llvm::object::SectionedAddress{
956                m_dwarf.GetDWARFContext().getOrLoadAddrData().GetMaxU64(
957                    &offset, index_size)};
958          });
959  if (!llvm_ranges)
960    return llvm_ranges.takeError();
961
962  DWARFRangeList ranges;
963  for (const llvm::DWARFAddressRange &llvm_range : *llvm_ranges) {
964    ranges.Append(DWARFRangeList::Entry(llvm_range.LowPC,
965                                        llvm_range.HighPC - llvm_range.LowPC));
966  }
967  return ranges;
968}
969
970llvm::Expected<DWARFRangeList>
971DWARFUnit::FindRnglistFromIndex(uint32_t index) {
972  if (llvm::Optional<uint64_t> offset = GetRnglistOffset(index))
973    return FindRnglistFromOffset(*offset);
974  if (m_rnglist_table)
975    return llvm::createStringError(errc::invalid_argument,
976                                   "invalid range list table index %d", index);
977
978  return llvm::createStringError(errc::invalid_argument,
979                                 "missing or invalid range list table");
980}
981